Next Article in Journal
Numerical Investigation of the Effect of Sloped Terrain on Wind-Driven Surface Fire and Its Impact on Idealized Structures
Next Article in Special Issue
Mechanical and Impact Properties of Engineered Cementitious Composites Reinforced with PP Fibers at Elevated Temperatures
Previous Article in Journal
Long Term Post-Fire Vegetation Dynamics in North-East Mediterranean Ecosystems. The Case of Mount Athos Greece
Previous Article in Special Issue
Residual Stress-Strain Relationship of Scoria Aggregate Concrete with the Addition of PP Fiber after Fire Exposure
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Thermal–Mechanical Coupling Evaluation of the Panel Performance of a Prefabricated Cabin-Type Substation Based on Machine Learning

1
Power Grid Planning Research Center of Guangdong Power Grid Limited Liability Company, Guangzhou 510699, China
2
Guangdong Engineering Research Centre for Major Infrastructure Safety, Sun Yat-Sen University, Guangzhou 510275, China
3
China Energy Construction Group Guangdong Electric Power Design and Research Institute Limited Liability Company, Guangzhou 510060, China
*
Authors to whom correspondence should be addressed.
Submission received: 15 November 2021 / Revised: 6 December 2021 / Accepted: 7 December 2021 / Published: 9 December 2021
(This article belongs to the Special Issue Performance-Based Design in Structural Fire Engineering)

Abstract

:
The panel performance of a prefabricated cabin-type substation under the impact of fires plays a vital role in the normal operation of the substation. However, current evaluations of the panel performance of substations under fire still focus on fire resistance tests, which seldom consider the relationship between fire behavior and the mechanical load of the panel under the impact of fires. Aiming at the complex and uncertain relationship between the thermal and mechanical performance of the substation panel under impact of fires, this paper proposes a machine learning method based on a BP neural network. First, the fire resistance test and the stress test of the panel is carried out, then a machine learning model is established based on the BP neural network. According to the collected data, the model parameters are obtained through a series of training and verification processes. Meanwhile, the correlation between the panel performance and fire resistance was obtained. Finally, related parameters are input into the thermal–mechanical coupling evaluation model for the substation panel performance to evaluate the fire resistance performance of the substation panel. To verify the correctness of the established model, numerical simulation of the fire test and stress test of the panel is conducted, and numerical simulation samples are predicted by the trained model. The results show that the prediction curve of neural network is closer to the real results compared with the numerical simulation, and the established model can accurately evaluate the thermal–mechanical coupling performance of the substation panel under fire.

1. Introduction

With the development of the national economy, the demand for electricity, from all walks of life, has increased. After a period of rapid development, large-scale centralized new energy power generation has gradually extended in the direction of decentralization and miniaturization. The requirements of new energy construction cannot be met by conventional transmission substations. Technological development and the improvement of prefabricated substations have become increasingly prominent. As a new type of prefabricated substation [1,2,3], the prefabricated cabin-type substation is becoming an important development direction benefiting from its high degree of integration and high level of intensiveness. Fire has an important effect on the safety of buildings and structures [4,5], thus the performance of the prefabricated substation panel under impact of fires is a guarantee of safety and plays a vital role in the normal operation of the substation. As a structural stress component of the substation panel, at the beginning of the design, the fire safety of the panel needs to be considered to ensure the safety of the overall structure of the substation. A high temperature causes the deterioration of the mechanical properties of the substation panel material, which will bring about different degrees of damage to the substation panel. Therefore, before the construction of the substation, it is necessary to carry out a fire resistance performance test under fire on the panel to ensure the fire resistance safety of the entire project in the event of a fire. Therefore, accurately describing the fire performance of substation panels has become an important issue for the stability of current substations.
Since the substation panels are mainly reinforced concrete structures, the fire performance of the substation panels can refer to the fire resistance test [6,7,8,9,10] and numerical simulation method to analyze fire behavior. Naser and Kodur [11] conducted an experimental study on the fire behavior of composite steel girders subjected to high shear loading. Hawileh et al. [12,13,14] predicted the performance of concrete beams using a finite element model. Aguado et al. [15] used a 3D finite element model for predicting the fire behavior of hollow-core slabs. However, the current research on the performance of substation panels rarely considers correlations, with little consideration of the nonlinear relationship between stress performance and fire resistance under impact of fire.
The neural network, a method of machine learning, is widely used in various fields [16,17,18,19,20,21,22,23]. Abuodeh et al. [24,25] used machine learning techniques to predict behavior of RC beams and compressive strength of ultra-high-performance concrete. Liu et al. [26] established machine-learning-based models to predict shear transfer strength of concrete joints. The neural network also has a precedent in the application of substation [27,28,29,30,31]. Da Silva et al. [32] proposed the use of artificial neural networks to solve the problem of fault location in substations; Wang et al. [33] used deep learning methods to identify the switch status of substations; Jiang Hongyu et al. [34] proposed an adaptive suppression method of transformer noise in substations based on genetic wavelet neural networks for the problem of transformer noise control; Oliveira et al. [35] carried out automatic monitoring on the construction site of substations based on deep learning. Neural networks [36,37,38] with self-learning, self-organization, and extremely strong linear fidelity capabilities can accurately reflect the nonlinear relationship between input and output variables to maintain high accuracy in short-term prediction. Therefore, machine learning is used to establish a non-linear relationship between panel stress and fire resistance from the perspective of thermal–mechanical coupling, which is a worthwhile means for evaluating the performance of substation panels under impact of fire.
To solve the above problem, this paper proposes a machine learning method based on the principle of BP (back propagation) neural networks to analyze the thermal–mechanical coupling performance of substation panels under fire. The evaluation factors are selected, such as the substation panel geometric data, mechanical performance parameters, and fire resistance performance data. After the model training ends, the relationship between panel mechanical performance and fire resistance is established. Finally, predictive samples are input into the model to evaluate the fire resistance performance of the panel. Then, fire resistance test and the stress test of the panel is carried out. A BP neural network model is trained and built through a series of training the samples. Then, numerical simulation of the fire test and stress test of the panel is conducted, and numerical simulation samples is predicted by the trained model and compared with the real results. The results show that predicted samples fit well with the actual output values and better than the result of numerical simulation. Thus, the established model can accurately evaluate the thermal–mechanical coupling performance of the panel under fire.

2. Research Methods and Contents

2.1. The Research Process for Thermal–Mechanical Coupling Evaluation of Prefabricated Cabin-Type Substation Panel Performance

The key to the thermal–mechanical coupling evaluation process of a prefabricated substation panel is to establish an evaluation model based on BP neural networks. By inputting the stress state data of the substation panel into the evaluation model, the corresponding fire resistance parameters can be obtained. The thermal–mechanical coupling performance of the prefabricated substation panel can then be evaluated. The research process of the thermal–mechanical coupling evaluation of prefabricated substation panel performance is shown in Figure 1.

2.2. Thermal–Mechanical Coupling Evaluation Model of the Panel Performance Based on BP Neural Networks

2.2.1. Establishment of Evaluation Factors

In theory, the performance state of the prefabricated substation panel can be better described by the more comprehensive evaluation indexes. However, in practical engineering, on the one hand, it is very difficult to collect data. On the other hand, the more indexes there are, the more complex the nonlinear relationship of the thermal–mechanical coupling evaluation of the prefabricated substation panel performance is. Therefore, the determination of evaluation indexes cannot be simply generalized but should be analyzed in specific cases. As a complex system, the thermal–mechanical coupling evaluation of panel performance is affected by many factors. This study, adhering to the principles of representativeness, integrity, and desirability, takes the geometric parameters, mechanical performance, and fire resistance performance of the panel as evaluation factors of the thermal–mechanical coupling evaluation of the panel’s performance.
  • The geometric parameters of the panel include length, width, and height.
  • The fire resistance performance parameters of the panel include the heating time, average furnace temperature, average temperature of the backfire surface, and pressure parameters.
  • The mechanical performance parameters of the panel include time and bending load.

2.2.2. Construction of BP Neural Network

The BP neural network as a method of machine learning is suitable for addressing complex nonlinear problems, such as the nonlinear relationship between the mechanical performance and the fire resistance performance of substation panels. The research process of the BP neural network model for the thermal–mechanical coupling evaluation of substation panel performance is shown in Figure 2. Firstly, the data parameters are input into the BP neural network for training. Secondly, the thermal–mechanical coupling evaluation results of the panel performance can be obtained through the model after model training. After that, we carried out numerical simulation of fire resistance test and stress test on the panel. We used the curve data of numerical simulation as sample data to predict the sample of numerical simulation. Finally, the correctness of the model is verified by comparing the real results with the numerical simulation results and the neural network prediction results.
As shown in Figure 3, the BP neural network used for the thermal–mechanical coupling evaluation training of the prefabricated cabin-type substation panel performance is composed of three layers, representing the input layer, hidden layer, and output layer, respectively.
The input layer has seven impact indicators corresponding to the identification indicators, which are the length, width, height, heating time, average furnace temperature, average temperature, and pressure of the backfire surface. The output layer represents time and bending load. Therefore, there are seven input layer nodes in this model, six hidden layer nodes, and two output nodes. Each node is a specific output function, and each connection between two nodes represents a weighted value (weight) for the signal passing through the connection. The learning rate determines the amount of weight change generated in each cycle. The fixed learning rate in this research is 0.1, the training target is 0.00001, and the maximum number of learning iterations is 100. Through repeated iterative calculations, the correlation coefficient and threshold are determined. After that, the learning and training process ends, which means the model is successfully established. After the BP neural network model training, the actual value is compared with the predicted value. In order to solve the problem of inconsistency in the units and magnitudes of the input variables in the BP neural network, normalization is used to control the sample data to 0–1.
The normalization formula is as follows:
Y i = X i X m i n X i X m a x α + β
In the formula, X i and Y i represent the variables before and after normalization, respectively; X m i n and X m a x are the minimum and maximum values of X i , respectively; α is a parameter with a value between 0–1, and β = 1 α 2 .

3. Case Application Analysis

3.1. Substation Panel

3.1.1. Fire Resistance Test of Panel

The fire resistance test of panel refer to the requirements of GB/T 9978.1-2008 “Fire resistance Test Methods for Building Components part 1: General Requirements [39]” and GB/T 9978.8-2008 “Fire resistance Test Methods for Building Components Part 8: Characteristics of non-load-bearing vertical dividers [40]”, as shown in Table 1. The test conditions and test plan were formulated according to the requirements of GB/T 9978.1-2008 [39] and GB/T 9978.8-2008 [40].
The length (m) width (m) × height (m) of the special panel for a box-type substation is 2.0 × 1.0 × 0.12. Ten temperature measurement points are set on the backfire surface of the panel with the vertical side on a free side, as shown in Figure 4.
According to the test requirements, the test uses vertical component fire test furnace device in Beijing Gequ fire test laboratory. The device can meet the requirements of the furnace temperature and pressure in Table 1. This device also can measure the temperature and pressure change value of the panel specimen. The data changes during the test can be visually displayed on the display screen of the equipment.
The experiment was terminated at 181 min. The test process was observed and recorded. The test phenomena are shown in Table 2.
The fire resistance data of the panels are shown in Figure 5 and Figure 6.

3.1.2. The Stress Test of the Panel

The same panel specimen as Section 3.1.1 was used in this experiment. Static loading is carried out by force control. A hydraulic jack was used for loading. During the test, the load is acted on the mid-span position of the panel through the actuating head. Once the specimen was destroyed, the test was over. The data of the stress test of the panel are shown in Figure 7.

3.2. Thermal–Mechanical Coupling Evaluation of Panel Performance

The values were recorded every minute from the origin of the coordinates. Figure 5 and Figure 6 show that the test specimen was damaged when heated to the 183rd minute. Figure 7 shows that the test specimen was damaged under stress at 329.052 s. The time from loading to failure was divided into 183 segments for the values recorded every 1.798 s. The fire resistance and stress performance data of the panel are shown in Appendix A. It should be emphasized that the temperature measured in Table A1 has subtract the ambient temperature. The data of columns 1 represent the number of samples; the data of columns 2 represent the heating time of panel; the data of columns 6 represent the load time of the panel.
According to the BP neural network structure constructed in Section 2.2, the thermal–mechanical coupling evaluation model of the panel performance was learned and trained:
  • Initialize the BP neural network. We randomly selected 100 sets of data from Table A1 as the input node data of the training sample, and the remaining 84 sets of data in Table A1 were used as prediction samples. Then, the weights and offsets of the neural network were initialized. Finally, the sample data were normalized.
  • Train the BP neural network. The BP neural network was used to train 100 sets of training sample data until the calculations at the end of the network training. The thermal–mechanical coupling evaluation model of the panel performance based on the BP neural network was obtained when the BP neural network converged after learning and training.
  • Predict the BP neural network. The randomly selected 84 sets of test sample data were predicted through the trained BP neural network to finally obtain the prediction result output. The graph is drawn as shown in Figure 8.
It can be seen from Figure 8 that the predicted output values of the 84 groups of predicted samples fit well with the actual output values for the trend of the sample points showing basically the same, which indicates that the thermal–mechanical coupling evaluation model of panel performance based on a BP neural network is reasonable and accurate.
The mechanical performance data of the panel corresponding to the heating time of the 162nd minute to the 183rd minute were collected, as shown in Figure 9.
It can be seen from Figure 9 that, when the test specimen reaches the maximum bending load of 21.443 KN, the corresponding stress time of the substation plate is 294.888 s. When the time is 325.456 s, the bending load drops sharply from 18.664 KN, which means the material is damaged at this time. The prediction sample data of the fire resistance performance of the substation are input into the thermal–mechanical coupling evaluation model of the panel performance. The corresponding panel performance parameters can then be obtained. The test specimen reaches the maximum bending load of 21.128 KN when the predicted value of the neural network is displayed for 297.147 s. The bending load drops sharply from 18.683 KN for the material being damaged at the time of 323.658 s. By comparing the predicted value and actual value of the time and bending load, it is found that the maximum bending load and the corresponding stress time from the thermal–mechanical coupling evaluation model and actual test is very close, and the two values essentially satisfy the error requirements. This further demonstrates the accuracy and reliability of the thermal–mechanical coupling evaluation model of the panel performance.

3.3. Numerical Simulation

In order to verify the results of neural network calculation, we carried out numerical simulation on the specimen. The length (m) × width (m) × height (m) of the special panel for numerical simulation is 2.0 × 1.0 × 0.12, as shown in Figure 10. The fire resistance test and pressure test of numerical simulation model are consistent with the actual situation in Section 3.1.
The numerical simulation results are shown in Figure 11 and Figure 12.
The curve of the fire resistance test and pressure test parameters for the panel is shown in Figure 13 and Figure 14. Each step in the diagram represents a unit of time.
The failure time step of numerical simulation corresponds to the failure time of fire resistance test and pressure test in real time, and the simulated result curve is also divided into 183 sections. Corresponding values are recorded in each section and 184 sample data of numerical simulation can be obtained.
According to the BP neural network structure trained in Section 3.2, we conduct neural network learning, training and prediction using the sample data of numerical simulation. According to the sample data of numerical simulation, the prediction results of numerical simulation are obtained. By converting the failure time of the real stress curve into the corresponding time step, we plotted the prediction curve of the neural network, the prediction curve of the numerical simulation and the real stress test curve in the same figure, as shown in Figure 15.
It can be seen from Figure 15 that the curve of prediction result of neural network and array simulation is basically consistent with the curve of real pressure test. The force increases gradually and decreases rapidly after reaching the peak value. Numerical simulation results show that when the time step is 15,850, the maximum bending load is 18.11064 kN. The neural network prediction results show that when the time step is 14,687, the bending load reaches the maximum value of 19.963 KN. The actual test results show that when the time step is 15,889, the bending load reaches the maximum value of 21.443 kN. Compared with the results of numerical simulation, the prediction curve of neural network is closer to the real pressure curve. The percentage error of the maximum bending load calculated by numerical simulation is 15.5%, the percentage error of the maximum bending load calculated by neural network prediction is 6.9%, and the error of neural network prediction is about half smaller than that of numerical simulation. The prediction result of neural network is better than that of numerical simulation. Thus, the accuracy and rationality of the neural network prediction model can be proved.

3.4. The Functional Relationship between Fire Resistance and Stress Resistance

The relationship between the parameters of fire resistance and stress resistance can be obtained by deriving the training parameters of the neural network, as shown in Equations (2)–(5):
α h = i = 1 M v i h x i + r h
b h = f ( α h )
y j = h = 1 q w h j b h + θ j
f ( x ) = 1 1 + e x
M refers to the number of nodes in the input layer, M = 7;   x i (i = 1, 2, ……, M) refers to length (m), width (m), height (m), heating time (min), average furnace temperature (°C), average temperature of backfire surface (°C), and pressure parameter (Pa); h refers to the number of hidden layer nodes, h = 6; q is the number of nodes in the output layer, q = 2; y j   ( j = 1 , 2 ) refers to the values of the time (s) and bending load (KPa), respectively. v refers to weight parameters from input layer to hidden layer of neural network; rh refers to threshold parameters from input layer to hidden layer of neural network; W refers to weight parameters from hidden layer to output layer of neural network; θj refers to threshold parameters from hidden layer to output layer of neural network.
v = [ 0 0 0 0 0 0 0 0 0 0 0 0 4.4136 1.6295 0.0460 0.1070 0.8386 0.7978 2.0962 0.1305 1.4522 0.1633 1.0707 0.3356 0.3682 0.0326 0.6002 0.0347 0 0 0 0 0 0 1.0436 1.1642   0.0877 0.2854 0.4381 0.3663 0.0416   0.1397 ]
r h = [ 3.3526 1.0328 0.2724 0.6572 0.1576 0.9250 ] T
θ j = [ 0.2891 1.0308 ] T
w = [ 0.0798 0.1080 0.0044 0.7329 0.9450 0.2738 1.4042 0.0712 0.6059 0.4502 0.7257 0.0036 ]

4. Conclusions

Based on the evaluation factors such as the geometric data of the substation panel, the stress performance, the fire resistance performance data, etc., a BP neural network, a method of machine learning, was used to establish the nonlinear relationship between panel performance stress and fire resistance under impact of fire. This model can quickly predict the performance of the substation panel under fire. The prediction of the thermal–mechanical coupling evaluation model is very close to the actual test, and satisfy the error requirements. Additionally, the specimen was verified by numerical simulation. Comparing the neural network with numerical simulation, the result indicates the error of neural network prediction is about half smaller than that of numerical simulation, the prediction result of neural network is better than that of numerical simulation. The correctness and reliability of the thermal–mechanical coupling performance evaluation model is verified. If meeting the requirements of the test itself and the amount of data required by the structure of the neural network, the thermal–mechanical coupling evaluation model constructed in this study can be directly used for similar models. It does not need to conduct additional tests. As the types and quantities of data for training become richer, the models we build will become more and more refined. Therefore, this can provide a reference for exploring more thermal coupling evaluation models and complex functional relationships of materials based on neural networks under different loading modes in the future.

Author Contributions

Conceptualization, methodology, data curation, formal analysis; writing—review and editing, project administration, funding acquisition, Z.L.; conceptualization, methodology, supervision, project administration, funding acquisition, C.Z.; data curation, formal analysis, writing—original draft, preparation and editing, J.O.; data Curation, X.L.; data curation, Y.W.; data curation, X.W.; data curation, X.Z.; data curation, F.C.; data Curation, C.X., J.O. and X.L. contributed equally to this work and they are co-first authors of this article. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by Science and Technology Project of Guangdong Power Grid Co., Ltd (037700KK52190022), the National Natural Science Foundation of China (NSFC) (Grant No.41977230), the National Key Research and Development Project (Grant No. 2017YFC1501203, No. 2017YFC1501201), the Special Fund Key Project of Applied Science and Technology Research and Development in Guangdong (Grant No. 2015B090925016, No. 2016B010124007).

Institutional Review Board Statement

The study did not require ethical approval.

Informed Consent Statement

Not applicable.

Data Availability Statement

Not applicable.

Acknowledgments

The authors would like to thank the anonymous reviewers for their very constructive and helpful comments.

Conflicts of Interest

The authors declare that they have no conflict of interest.

Appendix A

Table A1. Fire resistance and stress performance data of the panel.
Table A1. Fire resistance and stress performance data of the panel.
SampleHeating Time (min)Average Furnace Temperature (°C)Average Temperature of Backfire Surface
(°C)
Pressure (Pa)Time (s)Bending Load (KN)
100000.0000
2179.843.293.05771.7980.0152
32173.43.064.3493.5960.0619
43267.913.125.40325.3940.0949
54360.64.95.81177.1920.1097
65433.882.365.13428.9910.1888
76468.52311.85810.7890.287
87504.382.56.731112.5870.3562
98535.651.098.49914.3850.4535
109577.353.457.991316.1830.5464
11106252.9410.13317.9810.6664
1211634.442.0411.22119.7790.8527
1312643.431.949.659521.5770.9194
1413664.891.2811.823.3751.0552
1514690.581.99.01725.1731.1762
1615705.343.2711.8726.9721.2974
1716721.013.779.698228.7701.4346
1817734.373.248.137630.5681.5541
1918742.722.8811.70432.3661.7039
2019750.872.6512.75934.1641.8248
2120757.593.610.14535.9621.9602
2221763.693.1312.01437.7602.114
23227732.2713.13639.5582.3759
2423780.882.3315.75141.3562.5796
2524794.493.5212.08643.1542.7826
2625802.422.6714.15944.9532.9971
2726809.72.9616.67246.7513.2553
2827817.292.3213.24448.5493.4676
2928828.681.6811.98950.3473.659
3029840.374.2611.95852.1453.8044
3130844.533.4514.87853.9434.0017
3231851.163.7317.73255.7414.3081
3332857.394.3415.32357.5394.5063
3433854.145.1813.93259.3374.6692
3534861.786.8413.2261.1354.9182
3635867.053.712.9562.9345.1096
3736871.166.414.1464.7325.3131
3837876.675.218.89466.5305.4892
3938881.125.2415.39868.3285.7004
4039884.834.8314.10870.1265.8432
4140887.874.2113.90771.9245.9585
4241889.972.8113.6773.7226.0633
43428931.7312.99275.5206.4429
4443898.571.212.62177.3186.692
4544900.571.5513.43779.1166.9692
4645905.351.6414.01580.9157.1593
4746908.143.6113.94882.7137.3998
4847910.722.7913.94984.5117.6056
4948914.222.1213.95186.3097.9198
5049914.514.9213.95488.1078.1175
5150919.43.9413.98989.9058.3583
5251922.696.1914.36491.7038.634
5352925.516.3814.8493.5018.8009
5453929.226.4414.39995.2999.002
5554933.647.5114.02797.0979.2243
5655938.979.2414.84398.8969.3998
5756942.0311.2416.543100.6949.5576
5857944.3712.4518.004102.4929.7716
5958947.8814.5416.851104.2909.937
6059944.2717.7913.966106.08810.116
6160947.2515.613.934107.88610.237
6261949.4813.4914.342109.68410.439
6362952.0813.3215.022111.48210.679
6463954.1516.9116.586113.28010.9
6564955.8319.3717.402115.07811.103
6665958.8418.0618.863116.87711.379
6766962.1821.6815.877118.67511.6
6867966.2420.3914.384120.47311.794
6968969.5820.6613.91122.27112.07
7069967.6521.8614.387124.06912.231
7170970.9819.6716.629125.86712.374
7271973.6123.6316.834127.66512.742
7372976.4124.616.088129.46312.884
7473978.4826.514.731131.26113.016
7574978.6928.3814.053133.05913.21
7675981.3429.5616.432134.85813.381
7776982.5531.6418.029136.65613.548
7877983.3331.8218.777138.45413.693
797898632.3317.387140.25213.835
8079985.6737.1813.891142.05013.985
8180986.1934.7715.217143.84814.102
8281987.8337.3416.949145.64614.218
8382989.3738.3117.731147.44414.448
8483992.7640.2114.677149.24214.555
8584997.8243.9113.965151.04014.666
8685999.4242.3715.563152.83914.86
87861001.847.4716.311154.63715.123
8887100447.8617.738156.43515.303
89881005.850.8616.551158.23315.469
90891009.252.0614.923160.03115.585
91901006.145.6716.52161.82915.71
92911007.646.317.301163.62715.968
93921008.848.1716.862165.42516.002
94931011.350.7916.047167.22316.039
9594101355.7715.029169.02116.167
96951014.355.8514.896170.82016.339
97961016.157.0115.814172.61816.428
98971017.858.1714.865174.41616.512
99981020.656.4114.865176.21416.596
100991020.356.4814.935178.01216.681
1011001022.558.4216.329179.81016.74
1021011025.153.5517.756181.60816.865
1031021026.356.3515.992183.40616.986
1041031028.858.4715.757185.20417.071
1051041029.361.6117.863187.00217.123
1061051031.758.8315.589188.80117.229
1071061031.560.2514.911190.59917.369
1081071034.360.8816.814192.39717.487
1091081036.560.9314.881194.19517.605
1101091035.164.0714.881195.99317.577
1111101037.665.4216.92197.79117.662
1121111037.86015.529199.58917.877
1131121038.958.9214.885201.38717.853
1141131040.158.8217.807203.18517.876
1151141042.258.5815.499204.98317.968
1161151042.960.2314.923206.78218.081
1171161043.359.8616.86208.58018.181
1181171045.159.0614.858210.37818.182
1191181046.558.5416.83212.17618.27
1201191045.960.7416.83213.97418.366
1211201047.364.1316.83215.77218.392
1221211048.958.2414.897217.57018.488
123122105158.6317.819219.36818.58
1241231052.859.4915.749221.16618.639
1251241054.659.214.052222.96418.655
1261251055.760.5314.97224.76318.733
1271261057.560.2815.006226.56118.916
1281271058.958.2819.863228.35918.961
1291281060.657.7615.857230.15718.998
1301291060.75716.98231.95519.101
1311301063.257.9516.98233.75319.139
1321311064.858.3115.963235.55119.271
1331321066.157.5617.866237.34919.345
1341331068.758.5315.933239.14719.469
1351341066.858.2618.854240.94519.456
1361351069.358.7116.037242.74419.544
1371361070.358.6816.853244.54219.677
1381371070.758.4917.941246.34019.674
139138107258.1115.973248.13819.672
1401391071.658.516.891249.93619.815
1411401073.359.7915.738251.73419.827
142141107460.4214.925253.53219.85
1431421077.259.7515.367255.33019.961
1441431075.358.7816.999257.12820.132
1451441077.259.5616.185258.92620.111
1461451077.159.3415.405260.72520.281
1471461077.660.2614.863262.52320.4
148147107858.8916.935264.32120.36
1491481077.260.5115.918266.11920.547
1501491079.263.7317.923267.91720.555
1511501081.261.4915.921269.71520.593
1521511083.158.7117.925271.51320.787
153152108658.9915.991273.31120.758
1541531085.559.3717.962275.10920.805
1551541087.859.3115.892276.90720.912
156155109059.9615.895278.70621.004
1571561090.959.2315.895280.50421.061
1581571092.758.7115.895282.30221.038
1591581089.658.4215.93284.10021.135
1601591093.558.0315.933285.89821.264
1611601095.260.6415.933287.69621.275
1621611096.758.6616.818289.49421.257
1631621098.258.4815.868291.29221.402
164163109657.9615.868293.09021.405
1651641098.558.8215.868294.88821.443
1661651099.260.6215.868296.68720.986
1671661099.959.0815.943298.48520.603
1681671101.358.5515.946300.28320.502
169168109959.6515.946302.08120.263
170169110158.5615.946303.87919.9
1711701101.861.2717.916305.67719.858
1721711102.257.0915.982307.47519.903
1731721102.456.9219.039309.27319.868
1741731102.457.5517.783311.07119.913
1751741104.558.7515.985312.86919.827
176175110558.9315.037314.66819.665
1771761106.260.0116.871316.46619.445
178177110760.7616.975318.26419.301
1791781107.560.3518.946320.06219.131
1801791107.961.0916.06321.86018.953
1811801108.261.3919.016323.65818.834
1821811089.361.316.809325.45618.664
1831821089.261.4410.868327.25416.174
1841831010.761.6910.19329.05212.114

References

  1. Hazel, T.; Norris, A.; Barbizet, M.; Et, A. Designing prefabricated substation buildings according to GOST standards; Record of Conference Papers; Industry Applications Society; Forty-Ninth Annual Conference. In Proceedings of the 2002 Petroleum and Chemical Industry Technical Conference, New Orleans, LA, USA, 23–25 September 2002; pp. 251–259. [Google Scholar]
  2. Zhengmao, F.; Xiuhua, S.; Hongzhi, C.; Et, A. Optimization design of box structure for prefabricated substation. Int. J. Res. Eng. Technol. 2018, 7, 85–90. [Google Scholar]
  3. Zou, P.L. Comparative analysis of traditional civil construction new energy substation and modular prefabricated cabin substation. Mech. Electr. Inf. 2020, 38, 9. [Google Scholar]
  4. Gerges, M.; Demian, P.; Adamu, Z. Customising Evacuation Instructions for High-Rise Residential Occupants to Expedite Fire Egress: Results from Agent-Based Simulation. Fire 2021, 4, 21. [Google Scholar] [CrossRef]
  5. Ghodrat, M.; Shakeriaski, F.; Nelson, D.J.; Simeoni, A. Existing Improvements in Simulation of Fire–Wind Interaction and Its Effects on Structures. Fire 2021, 4, 27. [Google Scholar] [CrossRef]
  6. Ali, F.; Nadjai, A.; Silcock, G.; Et, A. Outcomes of a major research on fire resistance of concrete columns. Fire Saf. J. 2004, 39, 433–445. [Google Scholar] [CrossRef]
  7. Kodur, V.K.R.; Dwaikat, M.M.S.; Dwaikat, M.B. High-temperature properties of concrete for fire resistance modeling of structures. ACI Mater. J. 2008, 105, 517–527. [Google Scholar]
  8. Ran, L.; Zhao, H.; Huang, W.; Li, X.; Wang, Y.; Hu, Y. Fire resistance analysis of door and wall composite components. Fire Sci. Technol. 2014, 33, 1031–1033. [Google Scholar]
  9. Serrano, R.; Cobo, A.; Prieto, M.I.; Et, A. Analysis of fire resistance of concrete with polypropylene or steel fibers. Constr. Build. Mater. 2016, 122, 302–309. [Google Scholar] [CrossRef] [Green Version]
  10. Tian, J.; Zhu, P.; Qu, W. Study on fire resistance time of hybrid reinforced concrete beams. Struct. Concr. 2019, 20, 1941–1954. [Google Scholar] [CrossRef]
  11. Naser, M.Z.; Kodur, V.K.R. Comparative fire behavior of composite girders under flexural and shear loading. Thin-Walled Struct. 2017, 116, 82–90. [Google Scholar] [CrossRef] [Green Version]
  12. Hawileh, R.A.; Naser, M.Z. Thermal-stress analysis of RC beams reinforced with GFRP bars. Compos. Part B Eng. 2012, 43, 2135–2142. [Google Scholar] [CrossRef]
  13. Hawileh, R.A.; Naser, M.; Zaidan, W.; Al, E. Modeling of insulated CFRP-strengthened reinforced concrete T-beam exposed to fire. Eng. Struct. 2009, 31, 3072–3079. [Google Scholar] [CrossRef]
  14. Hawileh, R.A.; Naser, M.; Zaidan, W.; Al, E. Transient Thermal-Stress Finite Element Analysis of CFRP Strengthened RC beams Exposed to different Fire Scenarios. Mech. Adv. Mater. Struc. 2011, 18, 172–180. [Google Scholar] [CrossRef]
  15. Aguado, J.V.; Albero, V.; Espinos, A.; Al, E. A 3D finite element model for predicting the fire behavior of hollow-core slabs. Eng. Struct. 2016, 108, 12–27. [Google Scholar] [CrossRef]
  16. Faridmehr, I.; Nikoo, M.; Baghban, M.H.; Pucinotti, R. Hybrid Krill Herd-ANN Model for Prediction Strength and Stiffness of Bolted Connections. Buildings 2021, 11, 229. [Google Scholar] [CrossRef]
  17. Avossa, A.M.; Picozzi, V.; Ricciardelli, F. Load-Carrying Capacity of Compressed Wall-Like RC Columns Strengthened with FRP. Buildings 2021, 11, 285. [Google Scholar] [CrossRef]
  18. Abd-Elhamed, A.; Shaban, Y.; Mahmoud, S. Predicting Dynamic Response of Structures under Earthquake Loads Using Logical Analysis of Data. Buildings 2018, 8, 61. [Google Scholar] [CrossRef] [Green Version]
  19. Mishra, P.; Samui, P.; Mahmoudi, E. Probabilistic Design of Retaining Wall Using Machine Learning Methods. Appl. Sci. 2021, 11, 5411. [Google Scholar] [CrossRef]
  20. Jain, N.; Bansal, V.; Virmani, D.; Gupta, V.; Salas-Morera, L.; Garcia-Hernandez, L. An Enhanced Deep Convolutional Neural Network for Classifying Indian Classical Dance Forms. Appl. Sci. 2021, 11, 6253. [Google Scholar] [CrossRef]
  21. Wu, M.; Wang, J. Estimating Contact Force Chains Using Artificial Neural Network. Appl. Sci. 2021, 11, 6278. [Google Scholar] [CrossRef]
  22. Jiao, Z.; Hu, P.; Xu, H.; Al, E. Machine learning and deep learning in chemical health and safety: A systematic review of techniques and applications. ACS Chem. Health Saf. 2020, 27, 316–334. [Google Scholar] [CrossRef]
  23. Wang, W.; Kiik, M.; Peek, N.; Al, E. A systematic review of machine learning models for predicting outcomes of stroke with structured data. PLoS ONE 2020, 15, e234722. [Google Scholar]
  24. Abuodeh, O.R.; Abdalla, J.A.; Hawileh, R.A. Prediction of shear strength and behavior of RC beams strengthened with externally bonded FRP sheets using machine learning techniques. Compos. Struct. 2020, 234, 111698. [Google Scholar] [CrossRef]
  25. Abuodeh, O.; Abdalla, J.A.; Hawileh, R.A. Prediction of compressive strength of ultra-high performance concrete using SFS and ANN. In Proceedings of the 2019 8th International Conference on Modeling Simulation and Applied Optimization (ICMSAO), Sanya, China, 9–10 November 2019; pp. 1–5. [Google Scholar]
  26. Liu, T.; Wang, Z.; Zeng, J.; Al, E. Machine-learning-based models to predict shear transfer strength of concrete joints. Eng. Struct. 2021, 249, 113253. [Google Scholar] [CrossRef]
  27. Chen, C.S.; Tzeng, Y.M.; Hwang, J.C. The application of artificial neural networks to substation load forecasting. Electr. Power Syst. Res. 1996, 38, 153–160. [Google Scholar] [CrossRef]
  28. Hsu, Y.Y.; Lu, F.C. A combined artificial neural network-fuzzy dynamic programming approach to reactive power/voltage control in a distribution substation. IEEE Trans. Power Syst. 1998, 13, 1265–1271. [Google Scholar]
  29. Borkowski, D.; Wetula, A.; Bień, A. Contactless measurement of substation busbars voltages and waveforms reconstruction using electric field sensors and artificial neural network. IEEE Trans. Smart Grid 2014, 6, 1560–1569. [Google Scholar] [CrossRef]
  30. Nguyen, B.N.; Quyen, A.H.; Nguyen, P.H.; Al, E. Wavelet-based Neural Network for recognition of faults at NHABE power substation of the Vietnam power system. In Proceedings of the 2017 International Conference on System Science and Engineering (ICSSE), Ho Chi Minh City, Vietnam, 21–23 July 2017; pp. 165–168. [Google Scholar]
  31. Dudzik, M.; Jagiello, A.; Drapik, S.; Et, P.J. The selected real tramway substation overload analysis using the optimal structure of an artificial neural network. In Proceedings of the 2018 International Symposium on Power Electronics, Electrical Drives, Automation and Motion (SPEEDAM), Amalfi, Italy, 20–22 June 2018; pp. 413–417. [Google Scholar]
  32. Da Silva, A.P.A.; Insfran, A.H.F.; Da Silveira, P.M.; Et, A. Neural networks for fault location in substations. IEEE Trans. Power Deliv. 1996, 11, 234–239. [Google Scholar] [CrossRef]
  33. Wang, J.; You, Z.; Xiao, J.; Tan, Z. Deep learning based state recognition of substation switches. In Proceedings of the AIP Conference Proceedings, Kuala Lumpur, Malaysia, 24–26 July 2018; p. 1971. [Google Scholar]
  34. Jiang, H.; Liu, S.; Zhou, J.; Zhu, G.; Wang, K.; Shi, Z. Adaptive Noise Reduction of Transformer in Substation Based on Genetic Wavelet Neural Network. Electr. Power Sci. Eng. 2020, 36, 25–31. [Google Scholar]
  35. Oliveira, B.A.S.; Neto, A.P.D.F.; Fernandino, R.M.A.; Et, A. Automated Monitoring of Construction Sites of Electric Power Substations Using Deep Learning. IEEE Access 2021, 9, 19195–19207. [Google Scholar] [CrossRef]
  36. Wang, L.; Zeng, Y.; Chen, T. Back propagation neural network with adaptive differential evolution algorithm for time series forecasting. Expert Syst. Appl. 2015, 42, 855–863. [Google Scholar] [CrossRef]
  37. Li, J.; Cheng, J.; Shi, J.; Al, E. Brief introduction of back propagation (BP) neural network algorithm and its improvement. In Advances in Computer Science and Information Engineering; Springer: Berlin/Heidelberg, Germany, 2012. [Google Scholar]
  38. Singh, A.K.; Kumar, B.; Singh, S.K.; Al, E. Multiple watermarking technique for securing online social network contents using back propagation neural network. Future Gener. Comput. Syst. 2018, 86, 926–939. [Google Scholar] [CrossRef]
  39. Fire-Resistance Tests—Elements of Building Construction—Part 1: General Requirements (GB/T 9978.1-2008). Available online: https://gf.1190119.com/list-704.htm (accessed on 15 November 2021).
  40. Fire-Resistance Tests—Elements of Building Construction—Part 8: Specific Requirements for Non-Loadbearing Vertical Separating Elements (GB/T 9978.8-2008). Available online: https://www.doc88.com/p-7798292250942.html (accessed on 15 November 2021).
Figure 1. Research process of thermal–mechanical coupling evaluation of panel performance.
Figure 1. Research process of thermal–mechanical coupling evaluation of panel performance.
Fire 04 00093 g001
Figure 2. Research process of the BP neural network model in the thermal–mechanical coupling evaluation of prefabricated substation panel performance. x1, x2, …, x5, respectively, represents input layer parameters of neural network; u1, u2, …, uk represent hidden layer parameters of the neural network, respectively; yj represents output layer parameters of neural network; Ni represents output results of neural network; ω represents weights of neural network and θ represents thresholds of neural network.
Figure 2. Research process of the BP neural network model in the thermal–mechanical coupling evaluation of prefabricated substation panel performance. x1, x2, …, x5, respectively, represents input layer parameters of neural network; u1, u2, …, uk represent hidden layer parameters of the neural network, respectively; yj represents output layer parameters of neural network; Ni represents output results of neural network; ω represents weights of neural network and θ represents thresholds of neural network.
Fire 04 00093 g002
Figure 3. Application of the BP neural network in the thermal–mechanical coupling evaluation of substation panel performance.
Figure 3. Application of the BP neural network in the thermal–mechanical coupling evaluation of substation panel performance.
Fire 04 00093 g003
Figure 4. Schematic diagram of the measuring point layout on the backfire surface of the test specimen.
Figure 4. Schematic diagram of the measuring point layout on the backfire surface of the test specimen.
Fire 04 00093 g004
Figure 5. Temperature rise curve.
Figure 5. Temperature rise curve.
Fire 04 00093 g005
Figure 6. Pressure curve at 500 mm below the furnace roof.
Figure 6. Pressure curve at 500 mm below the furnace roof.
Fire 04 00093 g006
Figure 7. Stress curve of the panel strength test. Bending load refer to a load that causes bending deformation of a panel during a fixed strength test.
Figure 7. Stress curve of the panel strength test. Bending load refer to a load that causes bending deformation of a panel during a fixed strength test.
Fire 04 00093 g007
Figure 8. Comparison of sample predicted output and actual output.
Figure 8. Comparison of sample predicted output and actual output.
Fire 04 00093 g008
Figure 9. Sample result output of panel performance prediction.
Figure 9. Sample result output of panel performance prediction.
Fire 04 00093 g009
Figure 10. Numerical simulation model.
Figure 10. Numerical simulation model.
Fire 04 00093 g010
Figure 11. Numerical simulation of fire resistance test.
Figure 11. Numerical simulation of fire resistance test.
Fire 04 00093 g011
Figure 12. Numerical simulation of stress test.
Figure 12. Numerical simulation of stress test.
Fire 04 00093 g012
Figure 13. The curve of the fire resistance test.
Figure 13. The curve of the fire resistance test.
Fire 04 00093 g013
Figure 14. Stress curve of the panel samples. Bending load refer to a load that causes bending deformation of a panel during a fixed strength test.
Figure 14. Stress curve of the panel samples. Bending load refer to a load that causes bending deformation of a panel during a fixed strength test.
Fire 04 00093 g014
Figure 15. The stress test curve.
Figure 15. The stress test curve.
Fire 04 00093 g015
Table 1. Reference standards for fire resistance.
Table 1. Reference standards for fire resistance.
Test ItemsStandard ClauseJudgment Criteria
Fire resistanceCompletenessGB/T 9978.8-2008
Article 10
GB/T 9978.1-2008
Article 10.2.2 Article 8.4
The duration of the test specimen’s continuous fire resistance performance in the fire test. Any one of the following limited conditions of the test specimen shall be considered as a loss of integrity:
(a) A cotton pad test is conducted, and the cotton pad is ignited.
(b) A gap probe of 6 mm penetrates the specimen into the furnace and moves 150 mm along the length of the crack; a gap probe of 25 mm penetrates the specimen into the furnace.
(c) A flame appears on the backfire surface and lasts for more than 10 s.
Thermal insulationGB/T 9978.8-2008
Article 10
GB/T 9978.1-2008
Article 10.2.3
If the duration of the fire resistance and heat insulation performance of the test specimen in the fire test as well as the temperature rise of the backfire surface of the test specimen exceeds any of the following limits, it is considered to have lost the heat insulation:
(a) The average temperature rise exceeds the initial average temperature of 140 °C.
(b) The temperature rise at any point exceeds the initial temperature (including the moving thermocouple) by 180 °C (the initial temperature should be the initial average temperature of the back surface at the beginning of the test).
GB/T 9978.1-2008
Article 12.2.2
If the “integrity” of the test specimen does not meet the requirements, it is considered that the “heat insulation” of the test specimen does not meet the requirements.
Table 2. Test phenomena.
Table 2. Test phenomena.
TimeObservation Record
0Test start.
30No significant change from the previous stage.
60No significant change from the previous stage.
90No significant change from the previous stage.
120Concave deformation.
150No significant change from the previous stage.
181Integrity and thermal insulation are undamaged; test is stopped.
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Lei, X.; Ouyang, J.; Wang, Y.; Wang, X.; Zhang, X.; Chen, F.; Xia, C.; Liu, Z.; Zhou, C. Thermal–Mechanical Coupling Evaluation of the Panel Performance of a Prefabricated Cabin-Type Substation Based on Machine Learning. Fire 2021, 4, 93. https://0-doi-org.brum.beds.ac.uk/10.3390/fire4040093

AMA Style

Lei X, Ouyang J, Wang Y, Wang X, Zhang X, Chen F, Xia C, Liu Z, Zhou C. Thermal–Mechanical Coupling Evaluation of the Panel Performance of a Prefabricated Cabin-Type Substation Based on Machine Learning. Fire. 2021; 4(4):93. https://0-doi-org.brum.beds.ac.uk/10.3390/fire4040093

Chicago/Turabian Style

Lei, Xiangsheng, Jinwu Ouyang, Yanfeng Wang, Xinghua Wang, Xiaofeng Zhang, Feng Chen, Chang Xia, Zhen Liu, and Cuiying Zhou. 2021. "Thermal–Mechanical Coupling Evaluation of the Panel Performance of a Prefabricated Cabin-Type Substation Based on Machine Learning" Fire 4, no. 4: 93. https://0-doi-org.brum.beds.ac.uk/10.3390/fire4040093

Article Metrics

Back to TopTop