Home Prediction and monitoring model for farmland environmental system using soil sensor and neural network algorithm
Article Open Access

Prediction and monitoring model for farmland environmental system using soil sensor and neural network algorithm

  • Tao Song , Yulong Si , Jie Gao , Wei Wang EMAIL logo , Congwei Nie and Jiří Jaromír Klemeš
Published/Copyright: February 4, 2023
Download Article (PDF)
Open Physics
From the journal Open Physics Volume 21 Issue 1

Abstract

In this study, data fusion algorithm is used to classify the soil species and calibrate the soil humidity sensor, and by using edge computing and a wireless sensor network, farmland environment monitoring system with a two-stage calibration function of frequency domain reflectometer (FDR) is established. Edge computing is used in system nodes, including the saturation value of the soil humidity sensor, the calculated soil hardness, the calculation process of the neural network, and the model of soil classification. A bagged tree is adopted to avoid over-fitting to reduce the prediction variance of the decision tree. A decision tree model is established on each training set, and the C4.5 algorithm is adopted to construct each decision tree. After primary calibration, the root mean squared error (RMSE) between the measured and standard values is reduced to less than 0.0849%. The mean squared error (MSE) and mean absolute error (MAE) are reduced to less than 0.7208 and 0.6929%. The bagged tree model and backpropagation neural network are used to classify the soil and train the dynamic soil dataset. The output of the trained neural network is closer to the actual soil humidity than that of the FDR soil humidity sensor. The MAE, the MSE, and the RMSE decrease by 1.37%, 3.79, and 1.86%. With accurate measurements of soil humidity, this research shows an important guiding significance for improving the utilization efficiency of agricultural water, saving agricultural water, and formulating the crop irrigation process.

Keywords: humidity sensor; sensor calibration; backpropagation; bagged tree; neural network

1 Introduction

Abuse of earth resources results in unprecedented environmental crises. The development of resource utilization mainly focuses on non-renewable fossil resources and mineral resources currently, such as the replacement of conventional fluid with nanofluids [1], energy storage with nanocomposite phase change materials [2], and increase in heat transfer with magnetic fields [3]. Plate heat exchangers [4] and double tube heat exchangers [5] with novel structures have been introduced in distributed energy systems for energy saving. However, water resources have received less attention than fossil and mineral resources. Water resource is indispensable for social and economic development [6]. About 85% of the global freshwater is used for agricultural irrigation [7], and about 53% of Chinese freshwater resource is used for agriculture applications [8]. Only 40–50% of agricultural water in developing countries has been effectively utilized [9], and the utilization efficiency of agricultural water is 70–80% in developed countries [10]. The accurate measurement of soil humidity is of great significance for improving the utilization efficiency of agricultural water, saving agricultural water, and formulating the crop irrigation process [11].

The frequency domain reflectometer (FDR) method is commonly used to measure soil humidity [12]. The FDR method is used to obtain soil humidity by measuring the dielectric constant of the soil [13]. As the FDR soil humidity sensor is buried in soil for a long time [14], it is aged or corroded due to soil texture and other factors. The data drift occurs due to the influence of soil temperature and soil hardness [15]. In addition, the physical properties of soil are affected by nanoparticles due to the application of nano-agriculture [16].

First, linear data drift over the full humidity range is caused by aged and corroded sensors. According to this feature, the data characteristics measured by standard sensors are recorded first [17]. Then, the offset value is calculated according to the data characteristics after aging or corrosion. Second, nonlinear drift in the output of sensors is caused because of the difference in soil temperature, hardness, and soil type [18]. Based on the characteristics of nonlinear data drift of sensors [19], the FDR soil humidity sensor shows the soil humidity output value [20], to establish the neural network relation among soil temperature, soil hardness, and the actual soil humidity value [21].

A novel soil classification method based on data fusion of soil environmental parameters is proposed. A two-stage calibration method is designed to improve the measurement accuracy of the FDR soil humidity sensor. A farmland environment monitoring system with a two-stage calibration function of the FDR sensor is established using edge computing and a wireless sensor network.

2 Experimental investigations

In order to verify a linear rule of data drift due to aging and corrosion problems of FDR soil humidity sensors, the soil samples were collected on the Beichen Campus of Hebei University of Technology in Tianjin, China.

2.1 Primary calibration of FDR soil humidity sensor

The standard FDR soil humidity sensor marked A was accurate, and the other sensors marked B and C were used for 1 year and 2 years, respectively. In order to verify the linear drift characteristics of the sensors, three sensors were used for the experimental analysis of the soil at the same humidity [22].

Three sensors were buried in an even array in a plastic vessel, which was filled with dry soil samples. Add water evenly, so that the soil humidity inside the vessel gradually increases to the maximum, and a full humidity range was measured simultaneously with these three sensors. Test results are presented in Figure 1. It is found that the growth rate of the sensor output is slowing down with the increase in water injection. When its output reaches the saturation of soil humidity sensors, the output values of the sensors no longer increase with the amount of water injected. The output saturation values of sensors A, B, and C are 46.63, 55.79, and 58.93%. Sensors B and C, which are buried in soil for a few years, provide a large data offset than the standard sensor A. This part of the data offset is linear, and the curves show a larger overall value within the whole range of humidity.

Figure 1 Relation curves between sensor outputs and water injection time.
Figure 1

Relation curves between sensor outputs and water injection time.

According to the characteristics of sensors, the sensor with numerical fault can be calibrated by data compensation. As long as the output values of the standard sensor and the error sensor under a different soil humidity are measured, the data compensation can be carried out according to the difference.

In the actual greenhouse or farmland, it is difficult to use standard sensors and non-standard sensors to measure the output data under the same soil humidity conditions without manual operation [23]. The data compensation is carried out by measuring the saturation process curves of the non-standard soil sensors and the standard soil sensor. The soil humidity sensor will reach its output saturation value during irrigation time or rainy season, which corresponds to the highest points of sensors A and B as shown in Figure 1. The compensation value for the linear data is shown in Eq. (1) as follows:

(1) V = VS i − VS n ,

where V represents the compensation value for linear data, VS i represents the saturation value from a non-standard humidity sensor, and VS n represents the saturation value from the standard soil humidity sensor.

Aging and corrosion on humidity sensors are slow processes. The compensation value V is calculated according to Eq. (1) after the output saturation value of the non-standard sensor is monitored by the agricultural personnel for irrigation or in the rainy season. The sensor is calibrated according to the following equation:

(2) E = M − V ,

where E represents the calibrated soil humidity, M represents measured values from the non-standard soil humidity sensor, and V is the linear data compensation.

In Eq. (1), the output saturation value of the standard soil moisture sensor is unique and measured by the standard soil moisture sensor, while the output saturation value of each non-standard soil moisture sensor is different. The compensation value for linear data of each non-standard soil moisture sensor is obtained by subtracting the two. For a specific non-standard sensor, Eq. (2) is used to subtract the measured value from the compensation value of linear data to carry out data compensation for the non-standard soil moisture sensor. The data of the non-standard soil moisture sensor that is closer to the actual soil moisture after linear data compensation are obtained, namely, the E value.

2.2 Secondary calibration of FDR soil humidity sensor

Because only the linear compensation of the sensor cannot meet the requirements of agricultural utilizations, it is necessary to carry out the secondary calibration. Soil distributed in China is roughly divided into loam, brown soil, and sandy soil. Three soil samples were used in this experiment. For the same soil humidity in actual farmland, the FDR humidity sensor is affected by soil temperature, soil hardness, and soil type [24]. In the secondary calibration, the data set of corresponding relationships among soil temperature, soil hardness, soil type, output values of the humidity sensor, and actual soil humidity was obtained through many experiments. First, the soil was classified by the relationship among soil type, soil temperature, soil hardness, and output humidity of the FDR soil sensor after the primary calibration [25]. The backpropagation (BP) neural network algorithm was used to train the data set to get the characteristics of the neural network [26]. According to the network characteristics, the actual soil humidity value was obtained by inputting soil temperature, soil hardness, and output humidity of the FDR soil sensor.

In the experiment, the temperature range was from 10 to 50°C and the temperature gradient was 1°C. The soil hardness means the amount of soil pressure per cm2. The pressure value was measured by pressure sensors, and the surface area of the sensor-sensitive element was measured. The soil hardness is calculated by Eq. (3) as follows:

(3) H = F / S ,

where H represents the soil hardness, F represents the pressure value measured by the pressure sensor, and S represents the surface area of the sensor-sensitive element. The suitable range of soil hardness for crop growth in this experiment was 20–80 N/cm2.

A number of soil samples were thoroughly dried and divided into 24 parts, and each piece weighed 2,400 g marked m 0 . Each sample of soil was placed in a plastic cup. The 24 cups were labeled as Ai ∼ Fi (i is 1, 2, 3, and 4). An FDR soil humidity sensor, a soil temperature sensor, and pressure sensors were buried in each cup. Table 1 shows the weight of soil in each plastic cup, the weight of water injected, and the soil humidity. The plastic cup was sealed with plastic wrap to prevent water volatilization in the soil, and water was distributed evenly in the soil for 24 h in the plastic containers. Every accurate soil humidity in 24 plastic containers was calculated by Eq. (4) as follows:

(4) Hu = m i m 0 + m i ,

where Hu represents the actual soil humidity, m i represents the weight of water added into the cup, and m 0 represents the weight of soil.

Table 1

Labels and soil humidity

Vessel number (i = 1, 2, 3, 4) Ai Bi Ci Di Ei Fi
Weight of water (unit: g) The first measurement 125.3 262.8 429.1 608.3 800.7 1028.0
The second measurement 126.2 264.8 423.0 600.3 796.3 1027.6
The third measurement 126.5 273.4 419.9 591.4 803.0 1032.6
Average weight of water (unit: g) 126.0 267.0 424.0 600.0 800.0 1029.0
Soil humidity (unit: %) 5 10 15 20 25 30

The weight of water injected into the soil was measured three times, and the average value was taken to calculate the soil moisture value after different weights of water were injected into the dry soil through Eq. (4). After calculation, the soil moisture of six groups of containers was 5, 10, 15, 20, 25, and 30%, respectively. The output of the soil humidity sensor shows deviations from the actual soil humidity due to effects of soil temperature, hardness, and type. The secondary calibration of the FDR soil humidity sensor is used to eliminate the measurement deviations caused by soil temperature, soil hardness, and soil type. The measured results are improved using the secondary calibration, closer to the actual measurement accuracy.

The testing system was established to measure and store experimental data as shown in Figure 2.

Figure 2 Experimental testing system.
Figure 2

Experimental testing system.

The small experimental measurement system consisted of an FDR soil humidity sensor, soil temperature sensor, pressure sensor, data transmission unit, and cloud monitoring platform. The sensor whose output was deviated from the actual soil humidity was the sensor to be calibrated at the second stage. The data transmission unit USR-G781 is produced by Jinan YouRen Technology Company, China. The soil in each vessel was extruded to improve the soil hardness in each vessel as shown in Table 2.

Table 2

Correspondence between labels and soil hardness

Vessel number (X = A, B, C, D, E, F) X1 X2 X3 X4
Soil hardness (unit: N/cm2) 20 40 60 80

The output of the FDR soil humidity sensor in 24 vessels was measured with a temperature gradient of 1°C from 0 to 50°C. In Table 1, the accurate soil humidity calculated in the 24 vessels is 5, 10, 15, 20, 25, and 30%. The output of the FDR soil humidity sensor is affected by soil temperature, soil hardness, and soil types [27]. The output measured is shown in Figures 3–5.

Figure 3 Output of FDR soil humidity sensor at various soil temperatures and hardness in loam.
Figure 3

Output of FDR soil humidity sensor at various soil temperatures and hardness in loam.

Figure 4 Output of FDR soil humidity sensor at various soil temperatures and hardness in brown soil.
Figure 4

Output of FDR soil humidity sensor at various soil temperatures and hardness in brown soil.

Figure 5 Output of FDR soil humidity sensor at various soil temperatures and hardness in sandy soil.
Figure 5

Output of FDR soil humidity sensor at various soil temperatures and hardness in sandy soil.

The experimental process was repeated using samples of brown soil and sandy soil. The output of FDR soil humidity sensors is shown in Figures 4 and 5. From Figures 3–5, it is found that the output of the FDR soil humidity sensor is significantly affected by soil temperature, soil hardness, and soil type. In Figure 3, the error reaches its maximum value of 5.28% at 50°C. The soil hardness is 80 N/cm2, and the soil humidity is 20%.

In Figure 4, the error reaches its maximum value of 7.79% at 50°C. The soil hardness is 80 N/cm2, and the soil humidity is 20%. The error reaches its maximum value of 5.25% at 50°C. The soil hardness is 80 N/cm2, and the soil humidity is 20%. In soil humidity measurement, it is necessary to eliminate the influence of soil temperature, soil hardness, and soil type on soil humidity sensor.

The output of the FDR soil humidity sensors buried in three soils is used to build 984 groups of data set. Each group of data set is composed of soil temperature, soil hardness, FDR soil humidity sensor output, actual soil humidity, and the type of soil.

In order to eliminate the influence of soil type on the output of the FDR soil humidity sensor, a decision tree is a popular model for the classification of soil [28]. Bagged tree and random forests are the most famous and competitive extensions of decision trees. Li et al. [29] pointed out that the bagged tree combined with C4.5 is an excellent performance integration approach for classification problems [30]. A bagged tree that reduces the prediction variance of the decision tree was adopted to avoid over-fitting.

The bagged tree involves three steps. The original data are separated into subset data in step 1. Step 2 is to build a classifier on each data subset. Step 3 is the majority voting, which selects the best classifier from all the classifiers [31].

In the model of bagged tree, the C4.5 algorithm was used to construct trees based on a subset. The data set included soil temperature, soil hardness, and the output of the FDR soil humidity sensor. There were three labels: loam, brown soil, and sandy soil. In the C4.5 algorithm, the gain rate of the attribute needs to be calculated, and Eqs (5)–(8) show the calculation method. The basic information entropy represents the degree of chaos in the data set as Eq. (5). Information entropy is a basic concept of information theory, which describes the uncertainty of the occurrence of each possible event of an information source.

(5) entropy ( D ) = − ∑ j = 1 | C | Pr ( c j ) log 2 Pr ( c j ) ,

where D represents a data set, Pr ( c j ) represents the probability of category c j in data set D , C represents the number of categories in data set D , and entropy ( D ) represents the entropy of the data set D . After the data set is divided into v subsets according to attributes, the entropy of the data set is calculated as the following equation:

(6) entropy ( A i , D ) = ∑ j = 1 v D j D * entropy ( D j ) ,

where A i is an attribute of data set D , D j is a subset of data set D divided by attributes, v is the number of attributes in the data set, and entropy ( D j ) is the entropy of the subset D j and entropy ( D j ) . The information gain of attribute A i is calculated by the following equation:

(7) gain ( D , A i ) = entropy ( D ) − entropy ( A i , D ) ,

where gain ( D , A i ) represents the information gain of attribute A i . The information gained is biased to select attributes with more values. In order to correct this bias, the entropy of the data set relative to the distribution of attribute values is used to normalize the information gain, which called the information gain rate and calculated as follows:

(8) entropy ( A i , D ) = ∑ j = 1 v D j D * entropy ( D j ) ,

where gainRatio ( D , A i ) represents the information gain rate of attribute A i .

Figure 6 shows the flowchart of the C4.5 algorithm. The process of building a decision tree using the C4.5 algorithm is divided into four steps. In step 1, the total data sample MP is input. In step 2, the entropy of MP and the entropy of each attribute are calculated. In step 3, the information gain rate of each attribute is calculated, and the data set is divided into the categories with the maximum information gain rate to generate nodes. In step 4, the decision tree and mapping rules are obtained by repeating step 1–step 3.

Figure 6 Flowchart of the C4.5 algorithm.
Figure 6

Flowchart of the C4.5 algorithm.

Figure 7 shows the three steps of the construction of the bagged tree. In Figure 7, MP dataset is the entire dataset, and n samples are selected from MP dataset as the training set by placing back samples. A decision tree model was established on each training set, and the C4.5 algorithm was adopted to construct each decision tree. The models were used to vote on the results of the test samples, so as to obtain the final prediction results.

Figure 7 Model of the bagged tree.
Figure 7

Model of the bagged tree.

The bagged tree algorithm is not dependent on one decision tree, but by many decision trees to jointly determine the prediction results, which can improve the accuracy and stability of the results, but also avoid the problem of over-fitting.

In this study, BP neural network was used for data fusion. The mathematical equation of the mapping relationship between input and output in advance is not necessary to determine in the artificial neural network [32]. Learning specific rules through training obtained the result closest to the expected output value [33]. The BP neural network is a multi-layer feed-forward network trained according to error BP. With the gradient descent and searching gradient, this BP algorithm [34] minimized the mean squared error (MSE) of the actual output value and the expected output value of the network [35].

The basic BP algorithm includes two processes: forward propagation of signal and backward propagation of error [36]. The error output, the adjustment weight, and threshold are calculated from input to output. In forward propagation, the input signal acts on the output node through the hidden layer, which generates the output signal through nonlinear transformation. If the actual output is inconsistent with the expected output, the process of backward propagation is to be carried out. Backward propagation of error is apportioned to all units of each layer, which means that the output error is transmitted back to the input through the hidden layer. The error is lowered along the gradient direction by adjusting the connection strengths between the input node and the hidden layer node and between the hidden layer node and the output node [37]. After repeated learning and training, the network parameters (weights and thresholds) corresponding to the minimum error are determined [38]. In this case, the trained neural network processes the input information of similar samples with the least output error after nonlinear transformation.

The topology of the three-layer neural network is shown in Figure 8.

Figure 8 Topology of the three-layer neural network.
Figure 8

Topology of the three-layer neural network.

There are n nodes in the input layer of BP network, q nodes in the hidden layer, and m nodes in the output layer. The weight between the input layer and the hidden layer is v ki , and the weight between the hidden layer and the output layer is w jk . The transfer function of the hidden layer is f 1 ( · ) , and the transfer function of the output layer is f 2 ( · ) ; then, the output of the node of the hidden layer is shown as follows:

(9) z k = f 1 ∑ i = 0 n v ki x i ,

where the range of k is 1, 2, … , q . The output of the node in the output layer is

(10) y j = f 2 ∑ k = 0 q w jk z k .

At this point, BP network has completed the mapping from N-dimensional space vector to M-dimensional space. p learning samples are input into the BP network, and they are denoted by x 1 , x 2 , x 3 , … , x p . The p th sample is input to the network, and the output y j p ( j = 1 , 2 , … , m ) is obtained. The squared error function is used, so the error E p of the p th is obtained.

(11) E p = 1 2 ∑ j = 1 m ( t j p − y j p ) 2 ,

where t j p is the expected output. For p samples, the global error is shown as follows:

(12) E = 1 2 ∑ p = 1 P ∑ j = 1 m ( t j p − y j p ) = ∑ p = 1 P E p .

The cumulative error BP algorithm is used to adjust w jk to make the global error E smaller.

(13) ∆ w jk = − η ∂ E ∂ w jk = ∑ p = 1 P − η ∂ E p ∂ w jk ,

where η is the learning rate. The error signal is δ yj .

(14) δ yj = − ∂ E p ∂ S j = − ∂ E p ∂ y j ∂ y j ∂ S j ,

where the first term is shown in Eq. (15) and the second term is the partial of the transfer function of the output layer, which is shown in the following equation:

(15) ∂ E p ∂ y j = ∂ ∂ y j 1 2 ∑ j = 1 m ( t j p − y j p ) = − ∑ j = 1 m ( t j p − y j p ) ,

(16) ∂ y j ∂ S j = f 2 ′ ( S j ) .

So, δ y j can be expressed as follows:

(17) δ y j = ∑ j = 1 m ( t j p − y j p ) f 2 ′ ( S j ) .

According to the chain rule,

(18) ∂ E p ∂ w jk = ∂ E p ∂ S j ∂ S j ∂ w jk = − δ y j z k = − ∑ j = 1 m ( t j p − y j p ) f 2 ′ ( S j ) · z k .

Therefore, the weight adjustment formula of each neuron in the output layer is as follows:

(19) ∆ w jk = ∑ p = 1 P ∑ j = 1 m η ( t j p − y j p ) f 2 ′ ( S j ) · z k .

The change of hidden layer weight is given as follows:

(20) ∆ v ki = ∑ p = 1 P ( − η ∂ E p v ki ) .

The error signal is given as follows :

(21) δ zk = − ∂ E P ∂ S k = − ∂ E p ∂ z k ∂ z k ∂ S k ,

where the first term is shown in the following equation:

(22) ∂ E p ∂ z k = ∂ ∂ z k 1 2 ∑ j = 1 m ( t j p − y j p ) = − ∑ j = 1 m ( t j p − y j p ) ∂ y j ∂ z k .

Eq. (23) is obtained by the chain theorem.

(23) ∂ y i ∂ z k = ∂ y j ∂ S j ∂ S j ∂ z k = f 2 ′ ( S j ) w jk .

The second term in Eq. (21) is the partial derivative of the transfer function of the hidden layer, which is shown in the following equation:

(24) ∂ z k ∂ S k = f 1 ′ ( S k ) .

So, Eq. (25) is given as follows:

(25) δ z k = ∑ j = 1 m ( t j p − y j p ) f 2 ′ ( S j ) w jk f 1 ′ ( S k ) .

According to the chain theorem,

(26) ∂ E p ∂ v ki = ∂ E p ∂ S k ∂ S k ∂ v ki = − δ zk x i = − ∑ j = 1 m ( t j p − y j p ) f 2 ′ ( S j ) w jk f 1 ′ ( S k ) x i .

Thus, the weight adjustment formula of each neuron in the hidden layer is as follows:

(27) ∆ v ki = ∑ p = 1 P ∑ j = 1 m η ( t j p − y j p ) f 2 ′ ( S j ) w jk f 1 ′ ( S k ) x i .

In this study, the BP neural network was adopted for data fusion, and the data from Figures 3 –5 were used as the data set of the BP neural network [39]. The structured setting of the neural network is shown in Figure 9.

Figure 9 Structure of neural network.
Figure 9

Structure of neural network.

In Figure 9, neurons i 1 , i 2 , i 3 , and i 4 in input layer L 1 of the neural network are soil temperature, soil hardness, FDR soil humidity sensor, and soil type. There are eight neurons in hidden layer L 2 , and neuron Q in output layer L 3 is the output of the network, which is the soil humidity value after two-stage calibration.

The neural network was trained by the established data set to determine the weights from the input layer to the hidden layer and from the hidden layer to the output layer and the thresholds of each neuron in the hidden layer and output layer [40]. Figure 10 shows the flowchart of the BP neural network based on the algorithm of error reverse propagation.

Figure 10 Flowchart of BP neural network.
Figure 10

Flowchart of BP neural network.

After the BP neural network was trained, the weights of the input layer to the hidden layer, the hidden layer to the output layer, and the thresholds of neurons of the hidden layer and neurons of the output layer were obtained [41]. Table 3 shows the weights from input to hidden layers and the thresholds of neurons of the hidden layer.

Table 3

Weights and thresholds from input layer to hidden layer

Neuron h 1 h 2 h 3 h 4 h 5 h 6 h 7 h 8
Threshold −2.3545 1.6818 −1.0091 0.3364 −0.3364 −1.0091 0.5343 −1.1011
i 1 1.4268 −1.3495 1.6309 −1.4247 −0.5631 −1.5113 0.9078 1.5212
i 2 −1.4192 1.2833 −0.5860 0.8284 −0.5172 0.4666 −0.9872 1.0212
i 3 0.9376 1.4137 1.2027 −0.6019 0.5223 0.2766 0.5612 −0.8967
i 4 −0.7841 −0.2781 −1.0459 1.5702 2.1648 −1.7220 −1.0539 0.3436

Table 4 shows the weights from hidden to output layers and the thresholds of neurons of the output layer.

Table 4

Weights and thresholds from hidden layer to output layer

Neuron h1 h2 h3 h4 h5 h6 h7 h8
Q Threshold 0.6042 0.9783 −0.866 0.8788 −0.964 0.3677 0.5675 0.0683
0.7707

Tables 3 and 4 are deployed to the following system for secondary calibration of the FDR soil humidity sensor.

3 System structure

3.1 System overview

Based on the experimental analysis, a farmland environment monitoring system with two-stage calibration function was constructed.

The primary calibration is performed in this system by measuring historical soil humidity. In the secondary calibration, accurate soil humidity value is obtained through the acquisition of the current soil temperature, soil hardness, humidity output, and network calculations. The soil humidity value calibrated is closer to the actual value. It has a real guiding significance for crop growth monitor and water-saving. The system can also be used to monitor other environmental parameters of crop growth, including ambient temperature, air humidity, carbon dioxide concentration, and light intensity. These environmental parameters also showed vital guiding significance for crop growth and other agricultural activities [42].

3.2 Edge computing

As an extension and supplement of cloud computing, edge computing is considered a new computing paradigm that performs computing tasks at the edge of the network [43]. The core concept of edge computing is to make computing closer to the edge of the network and data sources to provide edge services nearby to meet the needs of users in agile connection [44], real-time data transmission, data optimization, application intelligence, security, and privacy protection [45]. The edge computing consortium defines four domains for edge computing, that is, device domain, network domain, data domain, and application domain, as shown in Figure 11. The application domain, data domain, network domain, and equipment domain are computing objects of edge computing. In the device domain, edge computing can directly process the perceived information. In the network domain, the automatic conversion of every network protocol is implemented, and the data format is standardized. Computing at the edge of the data domain makes data management more intelligent and flexible. Edge computing can provide localized controlling logic and application intelligence [46].

Figure 11 Field of edge computing.
Figure 11

Field of edge computing.

In a traditional cloud computing environment, only the cloud has a computing function [47]. However, due to the large amount of data collected by the agricultural monitoring system, network delays and congestion will occur if all collected data are uploaded to the cloud [48]. The system constructed in this study adopts the method of lightweight edge computing. The edge computing in this system is in the data domain layer in Figure 11. The humidity sensor saturation value, the calculated soil hardness, the calculation process of the neural network, and the model of soil classification were deployed to the edge of the system to finish the process of two-stage calibration of the FDR soil humidity sensor.

The 32-bit ST microelectronics (STM32) was adopted as the microcontroller unit of edge calculation, and Figure 12 shows the connection between STM32 and the sensors.

Figure 12 Connection between STM32 and sensors.
Figure 12

Connection between STM32 and sensors.

The program flowchart of STM32 is shown in Figure 13. The algorithm of model of bagged tree and the algorithm of BP neural network were implanted into STM32.

Figure 13 Controller program flowchart.
Figure 13

Controller program flowchart.

The STM32 collects data from the FDR soil humidity sensor, soil temperature sensor, and pressure sensor, which converts pressure data into soil hardness data by Eq. (3), and stores soil humidity data. Eq. (2) is used to obtain the soil humidity value after primary calibration. The model of a bagged tree is used to classify the soil. The soil temperature, soil hardness, soil humidity after the primary calibration, and kind of soil are used as a group of input data for the neural network. The neural network calculates the accurate soil humidity after the second calibration. When the STM32 receives the command for data collecting, it sends the soil humidity data to the long-range radio (LoRa) sub-nodes in the agriculture Internet of things (IoT).

3.3 System composition

The system consists of the perception, network, and application layers [49]. The overall architecture of the system is shown in Figure 14. The perception layer consists of five sensors and a microprocessor STM32, in which three sensors are connected to STM32. Modbus remote terminal unit (RTU) protocol is adopted for communication between LoRa sub-nodes and the perception layer, and Modbus RTU protocol is also adopted for light, temperature, humidity, and CO2 concentration sensors. When STM32 receives the data transmission command from LoRa sub-nodes, the soil humidity data calibrated are packaged and sent to LoRa sub-nodes by Modbus RTU protocol [50].

Figure 14 Overall architecture of the system.
Figure 14

Overall architecture of the system.

LoRa technology is used for data transmission at the network layer. LoRa spreads farther than another wireless mode under the same power consumption, realizing the unification of low power consumption and long distance [51]. Compared with traditional wireless communication, the distance of LoRa is 3–5 times larger than that of conventional radio frequency communication, up to 2–5 km [52].

In addition, LoRa technology supports the topology of the star network [53]. LoRa technology is very suitable for agricultural IOT [54]. The network layer consists of LoRa sub-nodes and LoRa gateway. LoRa sub-nodes send the perception layer data to the LoRa gateway, and then, the data are transferred to the cloud monitoring platform by the 4G network, which receives the data from the sub-nodes and sends the data to the cloud monitoring platform through the SIM card.

The cloud monitoring platform adopts the cloud platform of the Youren Cloud, Ltd., Shandong Province, China. The data transmission mode between the cloud monitoring platform and LoRa gateway is network pass-through. On the cloud monitoring platform, users can check the growing environment of crops in the farmland at any time to take related agricultural activities.

Figure 15 shows the data transmission modes among the cloud monitoring platform, LoRa gateway, LoRa sub-nodes, and sensors. After completing the network construction, users log in to the cloud monitoring platform to view the environmental conditions of the farmland. At this time, the soil humidity displayed on the platform is the soil humidity value calibrated by the two-stage calibration method in this study. Farmers send a command to update the calibration information in the cloud every time when they irrigate their fields or during the rainy season. When the STM32 receives the command, it will update primary calibration data according to the flowchart in Figure 13. The secondary calibration is performed automatically in the STM32 without the farmer’s involvement. In addition, light intensity, carbon dioxide concentration, ambient temperature, and humidity in the farm are also displayed to the user in the cloud monitoring platform. According to the environmental information, users can take related agricultural activities to improve the growth condition and yield of crops.

Figure 15 Way in which parts of a system communication.
Figure 15

Way in which parts of a system communication.

4 Results and discussion

This chapter shows discussions about the results of the primary calibration of soil humidity sensor, the soil classification model, and the performance of the BP neural network.

4.1 Primary calibration

If there are m groups of data, the calculation method of MSE is shown in the following equation:

(28) MAE = 1 m ∑ i = 1 m | f i − y i | ,

where m is the number of groups of data, f i is the predicted value, and y i is the real value. The calculation method of RMSE is shown in the following equation:

(29) RMSE = ∑ i = 1 N ( f i − y i ) 2 N .

The calculation method of MSE is shown in the following equation:

(30) MSE = 1 m ∑ i = 1 m ( f i − y i ) 2 .

Table 5 shows the MSE, the root mean squared error (RMSE), and the mean absolute error (MAE) and those of sensor B and sensor A and the error of sensor C and sensor A before and after calibration in Eqs (1) and (2).

Table 5

Error comparison of sensors before and after first-stage calibration

Error types MSE RMSE MAE
Errors between sensor B and sensor A Before calibration 73.4681 0.8571% 8.5509%
After calibration 0.7208 0.0849% 0.6929%
Errors between sensor C and sensor A Before calibration 145.4828 1.2062% 12.0460%
After calibration 0.5139 0.0717% 0.5102%

Compared with sensor A, the RMSE of sensor B and sensor C is reduced to a value below 0.0849% after first-stage calibration. For the MSE and the MAE, sensors B and C show the highest values of 0.7208 and 0.6929%. After first calibration, the output value of soil sensor is closer to the standard soil sensor output value.

4.2 Soil classification model

The accuracy of data classification reaches 90.0% using the bagged tree model with the training set. Regarding the data characteristics in the dataset, the bagged tree classification model is much more accurate than other methods.

There are some accuracies of some other methods. The linear discriminant, the support vector machine, the decision tree, the k nearest neighbors, and the bagged tree algorithms show classification accuracies of 34.4, 33.1, 42.6, and 52.8%. Figure 16 shows the confusion matrix and receiver operating characteristic (ROC) curves using the bagged tree training dataset.

Figure 16 Confusion matrix and ROC curves for the bagged tree model. (a) The confusion matrix of bagged tree and (b) the ROC curves of the bagged tree.
Figure 16

Confusion matrix and ROC curves for the bagged tree model. (a) The confusion matrix of bagged tree and (b) the ROC curves of the bagged tree.

A total of 2,952 groups of test samples were tested for the bagged tree classification model. In confusion matrix of bagged tree classification model, a total of 2,657 groups of test samples were accurately classified and a total of 295 groups of test samples were inaccurately classified. The classification accuracy of bagged tree classification model is 90%. In ROC curves, false-positive rate is the probability that a wrong sample is predicted to be positive and true-positive rate is the probability that a correct sample is predicted to be positive. Area under curve (AUC) is the area under the ROC curve. AUC is an evaluation index of classification model, and the closer it is to the 1, the better the effect of classification model. Among the above classification algorithms, the bag tree algorithm has the largest AUC value, which is as high as 0.98.

The bagged tree is the most suitable algorithm for this data set. Using the bagged tree algorithm, soil classification prepares for the BP neural network and secondary calibration of the FDR soil humidity sensor.

4.3 Secondary calibration

After the neural network training was completed, 20 sets of data were verified. Figure 17 shows the output error between the network and the actual soil humidity.

Figure 17 Output error between the neural network and actual soil humidity.
Figure 17

Output error between the neural network and actual soil humidity.

The absolute error between the output of neural network and actual soil humidity is controlled within 0.13% in Figure 17. This absolute error meets the needs of agricultural soil humidity measurement.

Table 6 shows the error between output of the network and the actual soil humidity and the error between input of the network and the actual soil humidity.

Table 6

Error among network input, output, and real soil humidity

Error classification MAE MSE RMSE
Error between output and actual value 0.08% 0.01 0.09%
Error between input and actual value 1.45% 3.80 1.95%
difference value 1.37% 3.79 1.86%

MAE, MSE, and RMSE between the output of the network and actual soil humidity are 1.37%, 3.79, and 1.86%. less than those between the input of network and actual soil humidity. The output of the neural network almost eliminates the effects of soil temperature, soil hardness, and soil type on the output of the FDR soil humidity sensor. The accuracy of soil humidity measurement is improved by using the FDR soil humidity sensor in agriculture. This result improves the utilization rate of agricultural water and saves agricultural water resources, which accurately determines the process of crop irrigation.

Various algorithms have been tried for the soil classification models, but only the bagged tree algorithm accurately classifies soil based on the mechanism of voting. For the BP neural network model, the value of soil humidity is closer to the real soil humidity according to the input, and its training results meet the expected expectations. The effects of soil temperature, soil hardness, and soil type on the FDR soil humidity sensor are removed by the calibration method in this study, and the accurate soil humidity value is obtained.

5 Conclusions

A two-stage calibration method for the measurement of FDR soil humidity sensor is proposed in this study. Linear and nonlinear data drifts of the FDR soil humidity sensor are analyzed.

The RMSE between measured values of sensor B and sensor C after first-stage calibration and that of standard sensor A is reduced to less than 0.0849%. Both the MSE and the MAE are reduced to less than 0.7208 and 0.6929%. In the secondary calibration, the data sets of soil temperature, soil hardness, soil types, soil humidity sensor output, and actual soil humidity are established experimentally. The data set trains the bagged tree for soil classification model and the BP neural network. The accuracy of the soil classification model reaches 90%. The output of the trained neural network is closer to the actual soil humidity than that of the FDR soil humidity sensor. The MAE, MSE, and RMSE decrease by 1.37%, 3.79, and 1.86%. The calibrated soil humidity value is indispensable for agricultural activities and saving agricultural water.

The dynamic monitoring system of the crop growth environment is built based on the two-stage calculation. The process of sensor secondary calibration is deployed to the edge of the network so that the calibrated soil humidity is collected in real time to guide the user to carry out corresponding agricultural irrigation activities in time, which reduces unnecessary water consumption in agriculture. The system adds the function of collecting the growth environment parameters of crops, which is convenient for users to carry out other agricultural activities according to environmental changes.

A novel idea for water-saving irrigation and water-resource saving globally is provided in this study, but some shortcomings exist in this study. The hardness measurement of soil in this experiment was insufficient. The soil was only roughly divided into three categories, and the measurement range of soil hardness was limited. In the subsequent work, soil classifications in both China and around the world will be shown in detail, and the measurement range of soil hardness will be further expanded.

  1. Funding information: This work was supported by the Natural Science Foundation of Hebei Province (Grant No. 19227212D), the Natural Science Foundation of Hebei Province of China (Grant No. E2021202163), and the Special Project of Science and Technology Winter Olympics in the Hebei Technology Innovation Guidance Plan (Grant No. 21474501D). This research is also supported by a project “Sustainable Process Integration Laboratory – SPIL,” project No. CZ.02.1.01/0.0/0.0/15_003/0000456 funded by EU as “CZ Operational Programme Research, Development and Education,” Priority 1: Strengthening capacity for quality research under the collaboration agreement with Hebei University of Technology, Tianjin, China.

  2. Author contributions: All authors have accepted responsibility for the entire content of this manuscript and approved its submission.

  3. Conflict of interest: The authors state no conflict of interest.

References

[1] Zhao M, Zhao L, Xiong X, He Y, Huang W, Liu Z, et al. Performance analysis of a plate heat exchanger using various nanofluids. Int J Heat Mass Transf. 2020;158:119993–2281.10.1016/j.ijheatmasstransfer.2020.119993Search in Google Scholar

[2] Wang J, Li Y, Zheng D, Mikulčić H, Vujanović M, Sundén B. Preparation and thermophysical property analysis of nanocomposite phase change materials for energy storage. Renew Sustain Energy Rev. 2021;151:111541.10.1016/j.rser.2021.111541Search in Google Scholar

[3] Wang J, Li G, Zhu H, Luo J, Sundén B. Experimental investigation on convective heat transfer of ferrofluids inside a pipe under various magnet orientations. Int J Heat Mass Transf. 2019;132:407–19.10.1016/j.ijheatmasstransfer.2018.12.023Search in Google Scholar

[4] Zheng D, Yang J, Wang J, Kabelac S, Sundén B. Analyses of thermal performance and pressure drop in a plate heat exchanger filled with ferrofluids under a magnetic field. Fuel. 2021;293:120432.10.1016/j.fuel.2021.120432Search in Google Scholar

[5] Zheng D, Du J, Wang W, Klemeš JJ, Wang J, Sundén B. Analysis of thermal efficiency of a corrugated double-tube heat exchanger with nanofluids. Energy. 2022;256:124522.10.1016/j.energy.2022.124522Search in Google Scholar

[6] Béné C, Prager SD, Achicanoy HAE, Toro PA, Lamotte L, Cedrez CB, et al. Understanding food systems drivers: A critical review of the literature. Glob Food Security. 2019;23:149–59.10.1016/j.gfs.2019.04.009Search in Google Scholar

[7] Zhang B, AghaKouchak A, Yang Y, Wei J, Wang G. A water-energy balance approach for multi-category drought assessment across globally diverse hydrological basins. Agric For Meteorol. 2019;264:247–65.10.1016/j.agrformet.2018.10.010Search in Google Scholar

[8] Bai Z, Caspari T, Gonzalez MR, Batjes NH, Mäder P, Bünemann EK, et al. Effects of agricultural management practices on soil quality: A review of long-term experiments for Europe and China. Agric Ecosyst Environ. 2018;265:1–7.10.1016/j.agee.2018.05.028Search in Google Scholar

[9] Rigby D, Cáceres D. Organic farming and the sustainability of agricultural systems. Agric Syst. 2001;68(1):21–40.10.1016/S0308-521X(00)00060-3Search in Google Scholar

[10] Nimmagadda SM, Harish KS. Review paper on technology adoption and sustainability in India towards smart cities. Multimed Tools Appl. 2022;81:27217–45.10.1007/s11042-022-12885-1Search in Google Scholar

[11] Fritz S, See L, Bayas JCL, Waldner F, Jacques D, Becker-Reshef I, et al. A comparison of global agricultural monitoring systems and current gaps. Agric Syst. 2019;168:258–72.10.1016/j.agsy.2018.05.010Search in Google Scholar

[12] Rajendran S, Priya AK, Senthil Kumar P, Hoang T, Sekar K, Chong KY, et al. A critical and recent developments on adsorption technique for removal of heavy metals from wastewater-A review. Chemosphere. 2022;303:135146.10.1016/j.chemosphere.2022.135146Search in Google Scholar PubMed

[13] Jiao W, Tian C, Chang Q, Novick KA, Wang L. A new multi-sensor integrated index for drought monitoring. Agric For Meteorol. 2019;268:74–85.10.1016/j.agrformet.2019.01.008Search in Google Scholar

[14] Wild J, Kopecký M, Macek M, Šanda M, Jankovec J, Haase T. Climate at ecologically relevant scales: A new temperature and soil humidity logger for long-term microclimate measurement. Agric For Meteorol. 2019;268:40–7.10.1016/j.agrformet.2018.12.018Search in Google Scholar

[15] Huuskonen J, Oksanen T. Soil sampling with drones and augmented reality in precision agriculture. Comput Electron Agric. 2018;154:25–35.10.1016/j.compag.2018.08.039Search in Google Scholar

[16] Wang J, Li G, Li T, Zeng M, Sundén B. Effect of various surfactants on stability and thermophysical properties of nanofluids. J Therm Anal Calorim. 2021;143:4057–70.10.1007/s10973-020-09381-9Search in Google Scholar

[17] Kamilaris A, Prenafeta-Boldú FX. Deep learning in agriculture: A survey. Comput Electron Agric. 2018;147:70–90.10.1016/j.compag.2018.02.016Search in Google Scholar

[18] Parsons DJ, Rey D, Tanguy M, Holman IP. Regional variations in the link between drought indices and reported agricultural impacts of drought. Agric Syst. 2019;173:119–29.10.1016/j.agsy.2019.02.015Search in Google Scholar

[19] Iizumi T, Ramankutty N. How do weather and climate influence cropping area and intensity? Glob Food Secur. 2015;4:46–50.10.1016/j.gfs.2014.11.003Search in Google Scholar

[20] Yu H, Zou W, Chen J, Chen H, Yu Z, Huang J, et al. Biochar amendment improves crop production in problem soils: A review. J Environ Manag. 2019;232:8–21.10.1016/j.jenvman.2018.10.117Search in Google Scholar PubMed

[21] Turner C, Aggarwal A, Walls H, Herforth A, Drewnowski A, Coates J, et al. Concepts and critical perspectives for food environment research: A global framework with implications for action in low-and middle-income countries. Glob Food Secur. 2018;18:93–101.10.1016/j.gfs.2018.08.003Search in Google Scholar

[22] Ramadan KM, Oates MJ, Molina-Martinez JM, Ruiz-Canales A. Design and implementation of a low cost photovoltaic soil humidity monitoring station for irrigation scheduling with different frequency domain analysis probe structures. Comput Electron Agric. 2018;148:148–59.10.1016/j.compag.2017.12.038Search in Google Scholar

[23] Javed M, Sajid M, Yousaf HMZ, Hassan G, Mahmood H. Facile and low cost temperature compensated humidity sensor and signal conditioning system. IEEE Sens J. 2021;21(13):14906–14.10.1109/JSEN.2021.3073957Search in Google Scholar

[24] Matthees HL, He Y, Owen RK, Hopkins D, Deutsch B, Lee J, et al. Predicting soil electrical conductivity of the saturation extract from a 1:1 soil to water ratio. Commun Soil Sci Plant Anal. 2017;48(18):2148–54.10.1080/00103624.2017.1407780Search in Google Scholar

[25] Louki II, Al‐Omran AM, Aly AA, Al‐Harbi AR. Sensor effectiveness for soil water content measurements under normal and extreme conditions. Irrig Drain. 2019;68(5):979–92.10.1002/ird.2377Search in Google Scholar

[26] Mosavi MR, Khishe M, Akbarisani M. Neural network trained by biogeography-based optimizer with chaos for sonar data set classification. Wirel Personal Commun. 2017;95:4623–42.10.1007/s11277-017-4110-xSearch in Google Scholar

[27] Lu S, Lu Y, Peng W, Ju Z, Ren T. A generalized relationship between thermal conductivity and matric suction of soils. Geoderma. 2019;337:491–7.10.1016/j.geoderma.2018.09.057Search in Google Scholar

[28] Ge Y, Shen L, Sun M. Temperature compensation for optical fiber graphene micro-pressure sensor using genetic wavelet neural networks. IEEE Sens J. 2021;21:24195–201.10.1109/JSEN.2021.3115810Search in Google Scholar

[29] Li Y, Chiou CT, Li H, Schnoor JL. Improved prediction of the bioconcentration factors of organic contaminants from soils into plant/crop roots by related physicochemical parameters. Environ Int. 2019;126:46–53.10.1016/j.envint.2019.02.020Search in Google Scholar PubMed PubMed Central

[30] Tao K, Zhao S, Gao P, Wang L, Jia H. Impacts of Pantoea agglomerans strain and cation-modified clay minerals on the adsorption and biodegradation of phenanthrene. Ecotoxicol Environ Saf. 2018;161:237–44.10.1016/j.ecoenv.2018.05.091Search in Google Scholar PubMed

[31] Liu Q, Liu Y, Zhang C, Ruan Z, Meng W, Cai Y, et al. sEMG-based dynamic muscle fatigue classification using SVM with improved whale optimization algorithm. IEEE Internet Things J. 2021;8(23):16835–44.10.1109/JIOT.2021.3056126Search in Google Scholar

[32] Le TTH, Kang H, Kim H. Household appliance classification using lower odd-numbered harmonics and the bagging decision tree. IEEE Access. 2020;8:55937–52.10.1109/ACCESS.2020.2981969Search in Google Scholar

[33] Iseki K, Konta T, Asahi K, Yamagata K, Fujimoto S, Tsuruya K, et al. Association of dipstick hematuria with all-cause mortality in the general population: results from the specific health check and guidance program in Japan. Nephrol Dialysis Transplant. 2018;33(5):825–32.10.1093/ndt/gfx213Search in Google Scholar PubMed

[34] Yang A, Zhuansun Y, Liu C, Li J, Zhang C. Design of intrusion detection system for internet of things based on improved BP neural network. IEEE Access. 2019;7:106043–52.10.1109/ACCESS.2019.2929919Search in Google Scholar

[35] Guo C, Cui Y, Dong B, Luo Y, Liu F, Zhao S, et al. Test study of the optimal design for hydraulic performance and treatment performance of free water surface flow constructed wetland. Bioresour Technol. 2017;238:461–71.10.1016/j.biortech.2017.03.163Search in Google Scholar PubMed

[36] Ghosh S, Bequette BW. Spectral graph theoretic analysis of process systems: An application to distillation columns. Comput Chem Eng. 2022;161:107748.10.1016/j.compchemeng.2022.107748Search in Google Scholar

[37] Jiang Q, Huang R, Huang Y, Chen S, He Y, Lan L, et al. Application of BP neural network based on genetic algorithm optimization in evaluation of power grid investment risk. IEEE Access. 2019;7:154827–35.10.1109/ACCESS.2019.2944609Search in Google Scholar

[38] Mosbah H, El-Hawary ME. Optimization of neural network parameters by Stochastic Fractal Search for dynamic state estimation under communication failure. Electr Power Syst Res. 2017;147:288–301.10.1016/j.epsr.2017.03.002Search in Google Scholar

[39] Rush B, Celi LA, Stone DJ. Applying machine learning to continuously monitored physiological data. J Clin Monit Comput. 2019;33:887–93.10.1007/s10877-018-0219-zSearch in Google Scholar PubMed PubMed Central

[40] Yang Y, Kim KR, Kou R, Li Y, Fu J, Zhao L, et al. Prediction of effluent quality in a wastewater treatment plant by dynamic neural network modeling. Process Saf Environ Prot. 2022;158:515–24.10.1016/j.psep.2021.12.034Search in Google Scholar

[41] Song S, Xiong X, Wu X, Xue Z. Modeling the SOFC by BP neural network algorithm. Int J Hydrog Energy. 2021;46(38):20065–77.10.1016/j.ijhydene.2021.03.132Search in Google Scholar

[42] Mahanty T, Bhattacharjee S, Goswami M, Bhattacharyya P, Das B, Ghosh A, et al. Biofertilizers: a potential approach for sustainable agriculture development. Environ Sci Pollut Res. 2017;24:3315–35.10.1007/s11356-016-8104-0Search in Google Scholar PubMed

[43] Köksal Ö, Tekinerdogan B. Architecture design approach for IoT-based farm management information systems. Precis Agric. 2019;20:926–58.10.1007/s11119-018-09624-8Search in Google Scholar

[44] Tian Y, Rao H, Zhang K, Tao S, Xue WT. Data security and privacy-preserving in edge computing paradigm: Survey and open issues. IEEE Access. 2018;6:18209–37.10.1109/ACCESS.2018.2820162Search in Google Scholar

[45] Khan WZ, Ahmed E, Hakak S, Yaqoob I, Ahmed A. Edge computing: A survey. Future Gener Comput Syst. 2019;97:219–35.10.1016/j.future.2019.02.050Search in Google Scholar

[46] Chen B, Wan J, Celesti A, Li D, Abbas H, Zhang Q. Edge computing in IoT-based manufacturing. IEEE Commun Mag. 2018;56(9):103–9.10.1109/MCOM.2018.1701231Search in Google Scholar

[47] Gochhayat SP, Kaliyar P, Conti M, Tiwari P, Prasath VB, Gupta D, et al. LISA: Lightweight context-aware IoT service architecture. J Clean Prod. 2019;212:1345–56.10.1016/j.jclepro.2018.12.096Search in Google Scholar

[48] Wang T, Cao Z, Wang S, Wang J, Qi L, Liu A, et al. Privacy-enhanced data collection based on deep learning for internet of vehicles. IEEE Trans Ind Inform. 2020;16(10):6663–72.10.1109/TII.2019.2962844Search in Google Scholar

[49] Franco P, Martinez JM, Kim YC, Ahmed MA. IoT based approach for load monitoring and activity recognition in smart homes. IEEE Access. 2021;9:45325–39.10.1109/ACCESS.2021.3067029Search in Google Scholar

[50] Park DS. Future computing with IoT and cloud computing. J Supercomput. 2018;74:6401–7.10.1007/s11227-018-2652-7Search in Google Scholar

[51] Cardenas AM, Nakamura Pinto MK, Pietrosemoli E, Zennaro M, Rainone M, Manzoni P. A low-cost and low-power messaging system based on the LoRa wireless technology. Mob Netw Appl. 2020;25:961–8.10.1007/s11036-019-01235-5Search in Google Scholar

[52] Liao CH, Zhu G, Kuwabara D, Suzuki M, Morikawa H. Multi-hop LoRa networks enabled by concurrent transmission. IEEE Access. 2017;5:21430–46.10.1109/ACCESS.2017.2755858Search in Google Scholar

[53] Zhang X, Cao Z, Dong W. Overview of edge computing in the agricultural internet of things: Key technologies, applications, challenges. IEEE Access. 2020;8:141748–61.10.1109/ACCESS.2020.3013005Search in Google Scholar

[54] Sodhro AH, Pirbhulal S, Luo Z, de Albuquerque VHC Towards an optimal resource management for IoT based green and sustainable smart cities. J Clean Prod. 2019;220:1167–79.10.1016/j.jclepro.2019.01.188Search in Google Scholar
Received: 2022-09-22
Revised: 2022-12-18
Accepted: 2023-01-02
Published Online: 2023-02-04

© 2023 the author(s), published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Articles in the same Issue

  1. Regular Articles
  2. Dynamic properties of the attachment oscillator arising in the nanophysics
  3. Parametric simulation of stagnation point flow of motile microorganism hybrid nanofluid across a circular cylinder with sinusoidal radius
  4. Fractal-fractional advection–diffusion–reaction equations by Ritz approximation approach
  5. Behaviour and onset of low-dimensional chaos with a periodically varying loss in single-mode homogeneously broadened laser
  6. Ammonia gas-sensing behavior of uniform nanostructured PPy film prepared by simple-straightforward in situ chemical vapor oxidation
  7. Analysis of the working mechanism and detection sensitivity of a flash detector
  8. Flat and bent branes with inner structure in two-field mimetic gravity
  9. Heat transfer analysis of the MHD stagnation-point flow of third-grade fluid over a porous sheet with thermal radiation effect: An algorithmic approach
  10. Weighted survival functional entropy and its properties
  11. Bioconvection effect in the Carreau nanofluid with Cattaneo–Christov heat flux using stagnation point flow in the entropy generation: Micromachines level study
  12. Study on the impulse mechanism of optical films formed by laser plasma shock waves
  13. Analysis of sweeping jet and film composite cooling using the decoupled model
  14. Research on the influence of trapezoidal magnetization of bonded magnetic ring on cogging torque
  15. Tripartite entanglement and entanglement transfer in a hybrid cavity magnomechanical system
  16. Compounded Bell-G class of statistical models with applications to COVID-19 and actuarial data
  17. Degradation of Vibrio cholerae from drinking water by the underwater capillary discharge
  18. Multiple Lie symmetry solutions for effects of viscous on magnetohydrodynamic flow and heat transfer in non-Newtonian thin film
  19. Thermal characterization of heat source (sink) on hybridized (Cu–Ag/EG) nanofluid flow via solid stretchable sheet
  20. Optimizing condition monitoring of ball bearings: An integrated approach using decision tree and extreme learning machine for effective decision-making
  21. Study on the inter-porosity transfer rate and producing degree of matrix in fractured-porous gas reservoirs
  22. Interstellar radiation as a Maxwell field: Improved numerical scheme and application to the spectral energy density
  23. Numerical study of hybridized Williamson nanofluid flow with TC4 and Nichrome over an extending surface
  24. Controlling the physical field using the shape function technique
  25. Significance of heat and mass transport in peristaltic flow of Jeffrey material subject to chemical reaction and radiation phenomenon through a tapered channel
  26. Complex dynamics of a sub-quadratic Lorenz-like system
  27. Stability control in a helicoidal spin–orbit-coupled open Bose–Bose mixture
  28. Research on WPD and DBSCAN-L-ISOMAP for circuit fault feature extraction
  29. Simulation for formation process of atomic orbitals by the finite difference time domain method based on the eight-element Dirac equation
  30. A modified power-law model: Properties, estimation, and applications
  31. Bayesian and non-Bayesian estimation of dynamic cumulative residual Tsallis entropy for moment exponential distribution under progressive censored type II
  32. Computational analysis and biomechanical study of Oldroyd-B fluid with homogeneous and heterogeneous reactions through a vertical non-uniform channel
  33. Predictability of machine learning framework in cross-section data
  34. Chaotic characteristics and mixing performance of pseudoplastic fluids in a stirred tank
  35. Isomorphic shut form valuation for quantum field theory and biological population models
  36. Vibration sensitivity minimization of an ultra-stable optical reference cavity based on orthogonal experimental design
  37. Effect of dysprosium on the radiation-shielding features of SiO2–PbO–B2O3 glasses
  38. Asymptotic formulations of anti-plane problems in pre-stressed compressible elastic laminates
  39. A study on soliton, lump solutions to a generalized (3+1)-dimensional Hirota--Satsuma--Ito equation
  40. Tangential electrostatic field at metal surfaces
  41. Bioconvective gyrotactic microorganisms in third-grade nanofluid flow over a Riga surface with stratification: An approach to entropy minimization
  42. Infrared spectroscopy for ageing assessment of insulating oils via dielectric loss factor and interfacial tension
  43. Influence of cationic surfactants on the growth of gypsum crystals
  44. Study on instability mechanism of KCl/PHPA drilling waste fluid
  45. Analytical solutions of the extended Kadomtsev–Petviashvili equation in nonlinear media
  46. A novel compact highly sensitive non-invasive microwave antenna sensor for blood glucose monitoring
  47. Inspection of Couette and pressure-driven Poiseuille entropy-optimized dissipated flow in a suction/injection horizontal channel: Analytical solutions
  48. Conserved vectors and solutions of the two-dimensional potential KP equation
  49. The reciprocal linear effect, a new optical effect of the Sagnac type
  50. Optimal interatomic potentials using modified method of least squares: Optimal form of interatomic potentials
  51. The soliton solutions for stochastic Calogero–Bogoyavlenskii Schiff equation in plasma physics/fluid mechanics
  52. Research on absolute ranging technology of resampling phase comparison method based on FMCW
  53. Analysis of Cu and Zn contents in aluminum alloys by femtosecond laser-ablation spark-induced breakdown spectroscopy
  54. Nonsequential double ionization channels control of CO2 molecules with counter-rotating two-color circularly polarized laser field by laser wavelength
  55. Fractional-order modeling: Analysis of foam drainage and Fisher's equations
  56. Thermo-solutal Marangoni convective Darcy-Forchheimer bio-hybrid nanofluid flow over a permeable disk with activation energy: Analysis of interfacial nanolayer thickness
  57. Investigation on topology-optimized compressor piston by metal additive manufacturing technique: Analytical and numeric computational modeling using finite element analysis in ANSYS
  58. Breast cancer segmentation using a hybrid AttendSeg architecture combined with a gravitational clustering optimization algorithm using mathematical modelling
  59. On the localized and periodic solutions to the time-fractional Klein-Gordan equations: Optimal additive function method and new iterative method
  60. 3D thin-film nanofluid flow with heat transfer on an inclined disc by using HWCM
  61. Numerical study of static pressure on the sonochemistry characteristics of the gas bubble under acoustic excitation
  62. Optimal auxiliary function method for analyzing nonlinear system of coupled Schrödinger–KdV equation with Caputo operator
  63. Analysis of magnetized micropolar fluid subjected to generalized heat-mass transfer theories
  64. Does the Mott problem extend to Geiger counters?
  65. Stability analysis, phase plane analysis, and isolated soliton solution to the LGH equation in mathematical physics
  66. Effects of Joule heating and reaction mechanisms on couple stress fluid flow with peristalsis in the presence of a porous material through an inclined channel
  67. Bayesian and E-Bayesian estimation based on constant-stress partially accelerated life testing for inverted Topp–Leone distribution
  68. Dynamical and physical characteristics of soliton solutions to the (2+1)-dimensional Konopelchenko–Dubrovsky system
  69. Study of fractional variable order COVID-19 environmental transformation model
  70. Sisko nanofluid flow through exponential stretching sheet with swimming of motile gyrotactic microorganisms: An application to nanoengineering
  71. Influence of the regularization scheme in the QCD phase diagram in the PNJL model
  72. Fixed-point theory and numerical analysis of an epidemic model with fractional calculus: Exploring dynamical behavior
  73. Computational analysis of reconstructing current and sag of three-phase overhead line based on the TMR sensor array
  74. Investigation of tripled sine-Gordon equation: Localized modes in multi-stacked long Josephson junctions
  75. High-sensitivity on-chip temperature sensor based on cascaded microring resonators
  76. Pathological study on uncertain numbers and proposed solutions for discrete fuzzy fractional order calculus
  77. Bifurcation, chaotic behavior, and traveling wave solution of stochastic coupled Konno–Oono equation with multiplicative noise in the Stratonovich sense
  78. Thermal radiation and heat generation on three-dimensional Casson fluid motion via porous stretching surface with variable thermal conductivity
  79. Numerical simulation and analysis of Airy's-type equation
  80. A homotopy perturbation method with Elzaki transformation for solving the fractional Biswas–Milovic model
  81. Heat transfer performance of magnetohydrodynamic multiphase nanofluid flow of Cu–Al2O3/H2O over a stretching cylinder
  82. ΛCDM and the principle of equivalence
  83. Axisymmetric stagnation-point flow of non-Newtonian nanomaterial and heat transport over a lubricated surface: Hybrid homotopy analysis method simulations
  84. HAM simulation for bioconvective magnetohydrodynamic flow of Walters-B fluid containing nanoparticles and microorganisms past a stretching sheet with velocity slip and convective conditions
  85. Coupled heat and mass transfer mathematical study for lubricated non-Newtonian nanomaterial conveying oblique stagnation point flow: A comparison of viscous and viscoelastic nanofluid model
  86. Power Topp–Leone exponential negative family of distributions with numerical illustrations to engineering and biological data
  87. Extracting solitary solutions of the nonlinear Kaup–Kupershmidt (KK) equation by analytical method
  88. A case study on the environmental and economic impact of photovoltaic systems in wastewater treatment plants
  89. Application of IoT network for marine wildlife surveillance
  90. Non-similar modeling and numerical simulations of microploar hybrid nanofluid adjacent to isothermal sphere
  91. Joint optimization of two-dimensional warranty period and maintenance strategy considering availability and cost constraints
  92. Numerical investigation of the flow characteristics involving dissipation and slip effects in a convectively nanofluid within a porous medium
  93. Spectral uncertainty analysis of grassland and its camouflage materials based on land-based hyperspectral images
  94. Application of low-altitude wind shear recognition algorithm and laser wind radar in aviation meteorological services
  95. Investigation of different structures of screw extruders on the flow in direct ink writing SiC slurry based on LBM
  96. Harmonic current suppression method of virtual DC motor based on fuzzy sliding mode
  97. Micropolar flow and heat transfer within a permeable channel using the successive linearization method
  98. Different lump k-soliton solutions to (2+1)-dimensional KdV system using Hirota binary Bell polynomials
  99. Investigation of nanomaterials in flow of non-Newtonian liquid toward a stretchable surface
  100. Weak beat frequency extraction method for photon Doppler signal with low signal-to-noise ratio
  101. Electrokinetic energy conversion of nanofluids in porous microtubes with Green’s function
  102. Examining the role of activation energy and convective boundary conditions in nanofluid behavior of Couette-Poiseuille flow
  103. Review Article
  104. Effects of stretching on phase transformation of PVDF and its copolymers: A review
  105. Special Issue on Transport phenomena and thermal analysis in micro/nano-scale structure surfaces - Part IV
  106. Prediction and monitoring model for farmland environmental system using soil sensor and neural network algorithm
  107. Special Issue on Advanced Topics on the Modelling and Assessment of Complicated Physical Phenomena - Part III
  108. Some standard and nonstandard finite difference schemes for a reaction–diffusion–chemotaxis model
  109. Special Issue on Advanced Energy Materials - Part II
  110. Rapid productivity prediction method for frac hits affected wells based on gas reservoir numerical simulation and probability method
  111. Special Issue on Novel Numerical and Analytical Techniques for Fractional Nonlinear Schrodinger Type - Part III
  112. Adomian decomposition method for solution of fourteenth order boundary value problems
  113. New soliton solutions of modified (3+1)-D Wazwaz–Benjamin–Bona–Mahony and (2+1)-D cubic Klein–Gordon equations using first integral method
  114. On traveling wave solutions to Manakov model with variable coefficients
  115. Rational approximation for solving Fredholm integro-differential equations by new algorithm
  116. Special Issue on Predicting pattern alterations in nature - Part I
  117. Modeling the monkeypox infection using the Mittag–Leffler kernel
  118. Spectral analysis of variable-order multi-terms fractional differential equations
  119. Special Issue on Nanomaterial utilization and structural optimization - Part I
  120. Heat treatment and tensile test of 3D-printed parts manufactured at different build orientations
https://doi.org/10.1515/phys-2022-0224
Keywords for this article
humidity sensor; sensor calibration; backpropagation; bagged tree; neural network
Creative Commons
BY 4.0

Articles in the same Issue

  1. Regular Articles
  2. Dynamic properties of the attachment oscillator arising in the nanophysics
  3. Parametric simulation of stagnation point flow of motile microorganism hybrid nanofluid across a circular cylinder with sinusoidal radius
  4. Fractal-fractional advection–diffusion–reaction equations by Ritz approximation approach
  5. Behaviour and onset of low-dimensional chaos with a periodically varying loss in single-mode homogeneously broadened laser
  6. Ammonia gas-sensing behavior of uniform nanostructured PPy film prepared by simple-straightforward in situ chemical vapor oxidation
  7. Analysis of the working mechanism and detection sensitivity of a flash detector
  8. Flat and bent branes with inner structure in two-field mimetic gravity
  9. Heat transfer analysis of the MHD stagnation-point flow of third-grade fluid over a porous sheet with thermal radiation effect: An algorithmic approach
  10. Weighted survival functional entropy and its properties
  11. Bioconvection effect in the Carreau nanofluid with Cattaneo–Christov heat flux using stagnation point flow in the entropy generation: Micromachines level study
  12. Study on the impulse mechanism of optical films formed by laser plasma shock waves
  13. Analysis of sweeping jet and film composite cooling using the decoupled model
  14. Research on the influence of trapezoidal magnetization of bonded magnetic ring on cogging torque
  15. Tripartite entanglement and entanglement transfer in a hybrid cavity magnomechanical system
  16. Compounded Bell-G class of statistical models with applications to COVID-19 and actuarial data
  17. Degradation of Vibrio cholerae from drinking water by the underwater capillary discharge
  18. Multiple Lie symmetry solutions for effects of viscous on magnetohydrodynamic flow and heat transfer in non-Newtonian thin film
  19. Thermal characterization of heat source (sink) on hybridized (Cu–Ag/EG) nanofluid flow via solid stretchable sheet
  20. Optimizing condition monitoring of ball bearings: An integrated approach using decision tree and extreme learning machine for effective decision-making
  21. Study on the inter-porosity transfer rate and producing degree of matrix in fractured-porous gas reservoirs
  22. Interstellar radiation as a Maxwell field: Improved numerical scheme and application to the spectral energy density
  23. Numerical study of hybridized Williamson nanofluid flow with TC4 and Nichrome over an extending surface
  24. Controlling the physical field using the shape function technique
  25. Significance of heat and mass transport in peristaltic flow of Jeffrey material subject to chemical reaction and radiation phenomenon through a tapered channel
  26. Complex dynamics of a sub-quadratic Lorenz-like system
  27. Stability control in a helicoidal spin–orbit-coupled open Bose–Bose mixture
  28. Research on WPD and DBSCAN-L-ISOMAP for circuit fault feature extraction
  29. Simulation for formation process of atomic orbitals by the finite difference time domain method based on the eight-element Dirac equation
  30. A modified power-law model: Properties, estimation, and applications
  31. Bayesian and non-Bayesian estimation of dynamic cumulative residual Tsallis entropy for moment exponential distribution under progressive censored type II
  32. Computational analysis and biomechanical study of Oldroyd-B fluid with homogeneous and heterogeneous reactions through a vertical non-uniform channel
  33. Predictability of machine learning framework in cross-section data
  34. Chaotic characteristics and mixing performance of pseudoplastic fluids in a stirred tank
  35. Isomorphic shut form valuation for quantum field theory and biological population models
  36. Vibration sensitivity minimization of an ultra-stable optical reference cavity based on orthogonal experimental design
  37. Effect of dysprosium on the radiation-shielding features of SiO2–PbO–B2O3 glasses
  38. Asymptotic formulations of anti-plane problems in pre-stressed compressible elastic laminates
  39. A study on soliton, lump solutions to a generalized (3+1)-dimensional Hirota--Satsuma--Ito equation
  40. Tangential electrostatic field at metal surfaces
  41. Bioconvective gyrotactic microorganisms in third-grade nanofluid flow over a Riga surface with stratification: An approach to entropy minimization
  42. Infrared spectroscopy for ageing assessment of insulating oils via dielectric loss factor and interfacial tension
  43. Influence of cationic surfactants on the growth of gypsum crystals
  44. Study on instability mechanism of KCl/PHPA drilling waste fluid
  45. Analytical solutions of the extended Kadomtsev–Petviashvili equation in nonlinear media
  46. A novel compact highly sensitive non-invasive microwave antenna sensor for blood glucose monitoring
  47. Inspection of Couette and pressure-driven Poiseuille entropy-optimized dissipated flow in a suction/injection horizontal channel: Analytical solutions
  48. Conserved vectors and solutions of the two-dimensional potential KP equation
  49. The reciprocal linear effect, a new optical effect of the Sagnac type
  50. Optimal interatomic potentials using modified method of least squares: Optimal form of interatomic potentials
  51. The soliton solutions for stochastic Calogero–Bogoyavlenskii Schiff equation in plasma physics/fluid mechanics
  52. Research on absolute ranging technology of resampling phase comparison method based on FMCW
  53. Analysis of Cu and Zn contents in aluminum alloys by femtosecond laser-ablation spark-induced breakdown spectroscopy
  54. Nonsequential double ionization channels control of CO2 molecules with counter-rotating two-color circularly polarized laser field by laser wavelength
  55. Fractional-order modeling: Analysis of foam drainage and Fisher's equations
  56. Thermo-solutal Marangoni convective Darcy-Forchheimer bio-hybrid nanofluid flow over a permeable disk with activation energy: Analysis of interfacial nanolayer thickness
  57. Investigation on topology-optimized compressor piston by metal additive manufacturing technique: Analytical and numeric computational modeling using finite element analysis in ANSYS
  58. Breast cancer segmentation using a hybrid AttendSeg architecture combined with a gravitational clustering optimization algorithm using mathematical modelling
  59. On the localized and periodic solutions to the time-fractional Klein-Gordan equations: Optimal additive function method and new iterative method
  60. 3D thin-film nanofluid flow with heat transfer on an inclined disc by using HWCM
  61. Numerical study of static pressure on the sonochemistry characteristics of the gas bubble under acoustic excitation
  62. Optimal auxiliary function method for analyzing nonlinear system of coupled Schrödinger–KdV equation with Caputo operator
  63. Analysis of magnetized micropolar fluid subjected to generalized heat-mass transfer theories
  64. Does the Mott problem extend to Geiger counters?
  65. Stability analysis, phase plane analysis, and isolated soliton solution to the LGH equation in mathematical physics
  66. Effects of Joule heating and reaction mechanisms on couple stress fluid flow with peristalsis in the presence of a porous material through an inclined channel
  67. Bayesian and E-Bayesian estimation based on constant-stress partially accelerated life testing for inverted Topp–Leone distribution
  68. Dynamical and physical characteristics of soliton solutions to the (2+1)-dimensional Konopelchenko–Dubrovsky system
  69. Study of fractional variable order COVID-19 environmental transformation model
  70. Sisko nanofluid flow through exponential stretching sheet with swimming of motile gyrotactic microorganisms: An application to nanoengineering
  71. Influence of the regularization scheme in the QCD phase diagram in the PNJL model
  72. Fixed-point theory and numerical analysis of an epidemic model with fractional calculus: Exploring dynamical behavior
  73. Computational analysis of reconstructing current and sag of three-phase overhead line based on the TMR sensor array
  74. Investigation of tripled sine-Gordon equation: Localized modes in multi-stacked long Josephson junctions
  75. High-sensitivity on-chip temperature sensor based on cascaded microring resonators
  76. Pathological study on uncertain numbers and proposed solutions for discrete fuzzy fractional order calculus
  77. Bifurcation, chaotic behavior, and traveling wave solution of stochastic coupled Konno–Oono equation with multiplicative noise in the Stratonovich sense
  78. Thermal radiation and heat generation on three-dimensional Casson fluid motion via porous stretching surface with variable thermal conductivity
  79. Numerical simulation and analysis of Airy's-type equation
  80. A homotopy perturbation method with Elzaki transformation for solving the fractional Biswas–Milovic model
  81. Heat transfer performance of magnetohydrodynamic multiphase nanofluid flow of Cu–Al2O3/H2O over a stretching cylinder
  82. ΛCDM and the principle of equivalence
  83. Axisymmetric stagnation-point flow of non-Newtonian nanomaterial and heat transport over a lubricated surface: Hybrid homotopy analysis method simulations
  84. HAM simulation for bioconvective magnetohydrodynamic flow of Walters-B fluid containing nanoparticles and microorganisms past a stretching sheet with velocity slip and convective conditions
  85. Coupled heat and mass transfer mathematical study for lubricated non-Newtonian nanomaterial conveying oblique stagnation point flow: A comparison of viscous and viscoelastic nanofluid model
  86. Power Topp–Leone exponential negative family of distributions with numerical illustrations to engineering and biological data
  87. Extracting solitary solutions of the nonlinear Kaup–Kupershmidt (KK) equation by analytical method
  88. A case study on the environmental and economic impact of photovoltaic systems in wastewater treatment plants
  89. Application of IoT network for marine wildlife surveillance
  90. Non-similar modeling and numerical simulations of microploar hybrid nanofluid adjacent to isothermal sphere
  91. Joint optimization of two-dimensional warranty period and maintenance strategy considering availability and cost constraints
  92. Numerical investigation of the flow characteristics involving dissipation and slip effects in a convectively nanofluid within a porous medium
  93. Spectral uncertainty analysis of grassland and its camouflage materials based on land-based hyperspectral images
  94. Application of low-altitude wind shear recognition algorithm and laser wind radar in aviation meteorological services
  95. Investigation of different structures of screw extruders on the flow in direct ink writing SiC slurry based on LBM
  96. Harmonic current suppression method of virtual DC motor based on fuzzy sliding mode
  97. Micropolar flow and heat transfer within a permeable channel using the successive linearization method
  98. Different lump k-soliton solutions to (2+1)-dimensional KdV system using Hirota binary Bell polynomials
  99. Investigation of nanomaterials in flow of non-Newtonian liquid toward a stretchable surface
  100. Weak beat frequency extraction method for photon Doppler signal with low signal-to-noise ratio
  101. Electrokinetic energy conversion of nanofluids in porous microtubes with Green’s function
  102. Examining the role of activation energy and convective boundary conditions in nanofluid behavior of Couette-Poiseuille flow
  103. Review Article
  104. Effects of stretching on phase transformation of PVDF and its copolymers: A review
  105. Special Issue on Transport phenomena and thermal analysis in micro/nano-scale structure surfaces - Part IV
  106. Prediction and monitoring model for farmland environmental system using soil sensor and neural network algorithm
  107. Special Issue on Advanced Topics on the Modelling and Assessment of Complicated Physical Phenomena - Part III
  108. Some standard and nonstandard finite difference schemes for a reaction–diffusion–chemotaxis model
  109. Special Issue on Advanced Energy Materials - Part II
  110. Rapid productivity prediction method for frac hits affected wells based on gas reservoir numerical simulation and probability method
  111. Special Issue on Novel Numerical and Analytical Techniques for Fractional Nonlinear Schrodinger Type - Part III
  112. Adomian decomposition method for solution of fourteenth order boundary value problems
  113. New soliton solutions of modified (3+1)-D Wazwaz–Benjamin–Bona–Mahony and (2+1)-D cubic Klein–Gordon equations using first integral method
  114. On traveling wave solutions to Manakov model with variable coefficients
  115. Rational approximation for solving Fredholm integro-differential equations by new algorithm
  116. Special Issue on Predicting pattern alterations in nature - Part I
  117. Modeling the monkeypox infection using the Mittag–Leffler kernel
  118. Spectral analysis of variable-order multi-terms fractional differential equations
  119. Special Issue on Nanomaterial utilization and structural optimization - Part I
  120. Heat treatment and tensile test of 3D-printed parts manufactured at different build orientations
Sign up now to receive a 20% welcome discount
Institutional Access
How does access work?
Have an idea on how to improve our website?
Please write us.
© 2025 De Gruyter Brill
Downloaded on 11.9.2025 from https://www.degruyterbrill.com/document/doi/10.1515/phys-2022-0224/html
Scroll to top button