Flow Calculation Model for Trapezoidal Gate Based on Artificial Neural Network

doi:10.11988/ckyyb.20230491

Abstract

Abstract:

This study introduces an approach for calculating the free flow and submerged flow through flat-trapezoidal gate. The flow calculation model was established based on BP (Back Propagation) neural network and RBF (Radial Basis Function) neural network, with multiple variables and combinations as well as single output. The input variables for the model included slope coefficient, gate opening, total head in front of the gate, hydraulic radius, contraction depth behind the gate, and downstream channel water depth. The output variable was measured flow rate. The model was trained and tested using experimental data, and extensive validation confirmed that both BP and RBF artificial neural network models demonstrated strong predictive performance. These models exhibited excellent adaptability and high accuracy in predicting flow rates for trapezoidal gates in canal systems in irrigation areas, thereby enabling precise flow control.

Key words: trapezoidal gate, free discharge, submerged discharge, BP neural network, RBF neural network, flow prediction

CLC Number:

TV131水力理论、计算、实验

Cite this article

MENG Wan-shang, ZHAO Shuai-jie, LI Lin. Flow Calculation Model for Trapezoidal Gate Based on Artificial Neural Network[J]. Journal of Yangtze River Scientific Research Institute, 2024, 41(9): 86-92.

share this article

0
/ / Recommend

Add to citation manager EndNote|Ris|BibTeX

URL: http://ckyyb.crsri.cn/EN/10.11988/ckyyb.20230491

http://ckyyb.crsri.cn/EN/Y2024/V41/I9/86

Figures/Tables 8

Fig.1 Trapezoidal gate diagram

Table 1 Input variable combinations for neural network

工况	2个变量	3个变量	4个变量	5个变量
自由出流	1-2,1-3,1-4, 2-3,2-4,3-4	1-2-3,1-2-4, 1-3-4,2-3-4	1-2-3-4
淹没出流	1-2,1-3,1-5, 1-6,2-3,2-5, 2-6,3-5,3-6, 5-6	1-2-3,1-2-5, 1-2-6,1-3-5, 1-3-6,1-5-6, 2-3-5,2-3-6, 2-5-6,3-5-6	1-2-3-5, 1-2-3-6, 1-2-5-6, 1-3-5-6, 2-3-5-6	1-2-3-5-6

Table 1 Input variable combinations for neural network

工况	2个变量	3个变量	4个变量	5个变量
自由出流	1-2,1-3,1-4, 2-3,2-4,3-4	1-2-3,1-2-4, 1-3-4,2-3-4	1-2-3-4
淹没出流	1-2,1-3,1-5, 1-6,2-3,2-5, 2-6,3-5,3-6, 5-6	1-2-3,1-2-5, 1-2-6,1-3-5, 1-3-6,1-5-6, 2-3-5,2-3-6, 2-5-6,3-5-6	1-2-3-5, 1-2-3-6, 1-2-5-6, 1-3-5-6, 2-3-5-6	1-2-3-5-6

Table 2 Predicted results for different combinations of variables in free flow state

神经网络模型	组合方式	组合数据	训练集误差/%	测试集误差/%
BP神经网络	2个变量组合	1-2	27.64	28.58
		1-3	30.66	39.34
		1-4	23.57	25.35
		2-3	7.28	8.06
		2-4	24.34	20.88
		3-4	3.35	3.68
	3个变量组合	1-2-3	4.71	5.68
		1-2-4	8.06	12.85
		1-3-4	2.21	2.34
		2-3-4	2.14	2.28
RBF神经网络	2个变量组合	1-2	27.16	28.14
		1-3	52.93	36.52
		1-4	23.64	25.48
		2-3	6.87	8.09
		2-4	24.14	20.95
		3-4	2.78	4.34
	3个变量组合	1-2-3	2.00	4.02
		1-2-4	10.85	12.91
		1-3-4	2.47	4.25
		2-3-4	2.10	3.18

Table 2 Predicted results for different combinations of variables in free flow state

神经网络模型	组合方式	组合数据	训练集误差/%	测试集误差/%
BP神经网络	2个变量组合	1-2	27.64	28.58
		1-3	30.66	39.34
		1-4	23.57	25.35
		2-3	7.28	8.06
		2-4	24.34	20.88
		3-4	3.35	3.68
	3个变量组合	1-2-3	4.71	5.68
		1-2-4	8.06	12.85
		1-3-4	2.21	2.34
		2-3-4	2.14	2.28
RBF神经网络	2个变量组合	1-2	27.16	28.14
		1-3	52.93	36.52
		1-4	23.64	25.48
		2-3	6.87	8.09
		2-4	24.14	20.95
		3-4	2.78	4.34
	3个变量组合	1-2-3	2.00	4.02
		1-2-4	10.85	12.91
		1-3-4	2.47	4.25
		2-3-4	2.10	3.18

Fig.2 Comparison between optimal predicted and measured free flow rates for different combinations of variables

Table 3 Predicted results for different combinationsof variables in submerged flow condition

神经网络模型	组合方式	组合数据	训练集误差/%	测试集误差/%
BP神经网络	2个变量组合	1-2	14.15	15.28
		1-3	26.44	23.28
		1-5	27.49	23.61
		1-6	23.90	22.25
		2-3	9.70	12.52
		2-5	14.25	16.10
		2-6	10.59	14.20
		3-5	22.33	20.54
		3-6	22.81	21.25
		5-6	21.86	19.56
	3个变量组合	1-2-3	9.82	20.46
		1-2-5	14.78	16.26
		1-2-6	10.48	13.59
		1-3-5	21.27	22.82
		1-3-6	22.66	33.88
		1-5-6	21.22	18.59
		2-3-5	4.58	5.30
		2-3-6	7.86	8.56
		2-5-6	2.89	3.93
		3-5-6	17.69	17.49
	4个变量组合	1-2-3-5	2.28	2.26
		1-2-3-6	7.59	25.50
		1-2-5-6	3.59	4.14
		1-3-5-6	20.05	31.33
		2-3-5-6	1.71	2.04
RBF神经网络	2个变量组合	1-2	14.10	15.60
		1-3	35.60	24.12
		1-5	27.72	22.12
		1-6	29.33	21.40
		2-3	10.65	11.19
		2-5	14.07	16.12
		2-6	10.89	13.47
		3-5	23.11	23.53
		3-6	38.61	21.19
		5-6	27.61	20.48
	3个变量组合	1-2-3	10.37	11.85
		1-2-5	13.78	16.20
		1-2-6	10.76	13.44
		1-3-5	22.47	25.49
		1-3-6	29.90	26.33
		1-5-6	29.84	20.28
		2-3-5	5.41	6.68
		2-3-6	8.46	8.89
		2-5-6	3.94	4.49
		3-5-6	33.51	21.37
	4个变量组合	1-2-3-5	2.35	3.38
		1-2-3-6	7.52	8.51
		1-2-5-6	3.83	4.77
		1-3-5-6	22.04	46.55
		2-3-5-6	1.56	2.06

Table 3 Predicted results for different combinationsof variables in submerged flow condition

神经网络模型	组合方式	组合数据	训练集误差/%	测试集误差/%
BP神经网络	2个变量组合	1-2	14.15	15.28
		1-3	26.44	23.28
		1-5	27.49	23.61
		1-6	23.90	22.25
		2-3	9.70	12.52
		2-5	14.25	16.10
		2-6	10.59	14.20
		3-5	22.33	20.54
		3-6	22.81	21.25
		5-6	21.86	19.56
	3个变量组合	1-2-3	9.82	20.46
		1-2-5	14.78	16.26
		1-2-6	10.48	13.59
		1-3-5	21.27	22.82
		1-3-6	22.66	33.88
		1-5-6	21.22	18.59
		2-3-5	4.58	5.30
		2-3-6	7.86	8.56
		2-5-6	2.89	3.93
		3-5-6	17.69	17.49
	4个变量组合	1-2-3-5	2.28	2.26
		1-2-3-6	7.59	25.50
		1-2-5-6	3.59	4.14
		1-3-5-6	20.05	31.33
		2-3-5-6	1.71	2.04
RBF神经网络	2个变量组合	1-2	14.10	15.60
		1-3	35.60	24.12
		1-5	27.72	22.12
		1-6	29.33	21.40
		2-3	10.65	11.19
		2-5	14.07	16.12
		2-6	10.89	13.47
		3-5	23.11	23.53
		3-6	38.61	21.19
		5-6	27.61	20.48
	3个变量组合	1-2-3	10.37	11.85
		1-2-5	13.78	16.20
		1-2-6	10.76	13.44
		1-3-5	22.47	25.49
		1-3-6	29.90	26.33
		1-5-6	29.84	20.28
		2-3-5	5.41	6.68
		2-3-6	8.46	8.89
		2-5-6	3.94	4.49
		3-5-6	33.51	21.37
	4个变量组合	1-2-3-5	2.35	3.38
		1-2-3-6	7.52	8.51
		1-2-5-6	3.83	4.77
		1-3-5-6	22.04	46.55
		2-3-5-6	1.56	2.06

Fig.3 Comparison between optimal predicted and measured submerged flow rates for different combinations of variables

Table 4 Calculated results for each evaluation indicator

出流方式	神经网络模型	R	RMSE/(m³·s^-1)	MAPE/%
自由出流	BP	0.999 1	1.60×10^-4	1.76
自由出流	RBF	0.998 6	2.03×10^-4	3.23
淹没出流	BP	0.996 1	2.43×10^-4	2.15
淹没出流	RBF	0.994 7	2.83×10^-4	2.19

Table 4 Calculated results for each evaluation indicator

出流方式	神经网络模型	R	RMSE/(m³·s^-1)	MAPE/%
自由出流	BP	0.999 1	1.60×10^-4	1.76
自由出流	RBF	0.998 6	2.03×10^-4	3.23
淹没出流	BP	0.996 1	2.43×10^-4	2.15
淹没出流	RBF	0.994 7	2.83×10^-4	2.19

Fig.4 Comparison of trapezoidal gate flow rate between prediction and measurement

Related Articles 15

[1]	ZHANG Yan, LIU Jia-hao, LÜ Yuan, LI Zhi-hao, ZHANG Yan-nan, LI Sheng, CHEN Zhuan-wen. Influence of Stone Powder Parameters on Mechanical Properties of Concrete [J]. Journal of Yangtze River Scientific Research Institute, 2023, 40(3): 166-173.
[2]	LI Qing-wen, CAO Hang, YANG Lu, WANG Xue-zhi, LIU Hua-xin. Effects of Steel-PVA Hybrid Fiber on Shear Performance of Self-compacting Concrete Beams [J]. Journal of Yangtze River Scientific Research Institute, 2022, 39(10): 121-127.
[3]	GUO Zhen-tian, HUANG Feng, GUO Li-dan, WU Yao. Causes and Countermeasures of Hydrological Regime Evolution in Poyang Lake [J]. Journal of Yangtze River Scientific Research Institute, 2021, 38(6): 27-31.
[4]	XING Yue, TIAN Zheng-hong, DU Hui. Intelligent Evaluation of Interlayer Bonding Quality of RCC Dam [J]. JOURNAL OF YANGTZE RIVER SCIENTIFIC RESEARCH INSTI, 2020, 37(8): 142-149.
[5]	ZHAO Li-xue, HUANG Jie-jun, CHENG Xue-jun, SHEN Shao-hong, YUAN Yan-bin. A Method of River Flow Prediction Based on VMD-BP Model [J]. JOURNAL OF YANGTZE RIVER SCIENTIFIC RESEARCH INSTI, 2020, 37(7): 47-52.
[6]	XIAO Xiao, XU Jian, ZHAO Deng-zhong, CHENG Xue-jun, LI Guo-zhong, ZHAO Bao-cheng, XU Jian. Remote Sensing Retrieval of Total Suspended Solids Concentration for Typical Reach of Hanjiang River Using Hyperspectral Data [J]. Journal of Yangtze River Scientific Research Institute, 2020, 37(11): 141-148.
[7]	YAN Hong-bo, ZHOU Bin, LU Xian-jian, LIU Hai-feng. Prediction of Dam Deformation Monitoring Data Based on EEMD-GA-BP Model [J]. JOURNAL OF YANGTZE RIVER SCIENTIFIC RESEARCH INSTI, 2019, 36(9): 58-63.
[8]	DONG Dan-dan, ZU An-jun, SUN Xue-lian. Model of Dam Deformation Monitoring Based on Genetic Ant Colony Optimization and Back Propagation Improved by Markov Chain [J]. JOURNAL OF YANGTZE RIVER SCIENTIFIC RESEARCH INSTI, 2019, 36(7): 48-54.
[9]	RAO Yun-kang, DING Yu, XU Wen-nian, ZHANG Liang, ZHANG Heng, PAN Bo. Estimating Maximum Dry Density of Gravel Soil by Back Propagation Neural Network Optimized by Genetic Algorithm [J]. JOURNAL OF YANGTZE RIVER SCIENTIFIC RESEARCH INSTI, 2019, 36(4): 88-92.
[10]	GUO Yi, WU Xin-miao, QIE Zhi-hong, RAN Yan-li. Parameter Calibration of MIKE SHE Model Based on BP Neural Network [J]. JOURNAL OF YANGTZE RIVER SCIENTIFIC RESEARCH INSTI, 2019, 36(3): 26-30.
[11]	HE Hua-gang. Analysis of Deformation Mechanism and Risk Assessment of Tunnel in Operation [J]. JOURNAL OF YANGTZE RIVER SCIENTIFIC RESEARCH INSTI, 2019, 36(12): 36-42.
[12]	WANG Fei. Factors Influencing Deformation of Subway’s Deep Foundation Pit Based on Quantification Theory III [J]. JOURNAL OF YANGTZE RIVER SCIENTIFIC RESEARCH INSTI, 2019, 36(11): 110-114.
[13]	WANG Xiao-bing, XIA Xiao-zhou, ZHANG Qing. Reliability Analysis on Anti-sliding Stability of Levee SlopeBased on Orthogonal Test and Neural Network [J]. JOURNAL OF YANGTZE RIVER SCIENTIFIC RESEARCH INSTI, 2019, 36(10): 89-93.
[14]	SHEN Jing-xin, FANG Bin, ZHENG Dong-jian, GUO Zhi-yun, LI Dan. Dam Deformation Monitoring by Radial Basis Function Model Optimized by Particle Swarm Optimization with Inertia Weight and AdaBoost [J]. JOURNAL OF YANGTZE RIVER SCIENTIFIC RESEARCH INSTI, 2018, 35(5): 57-62.
[15]	ZHOU Yong-sheng. Application of Multi-stage Progressive Model to Predicting Foundation Pit Deformation [J]. JOURNAL OF YANGTZE RIVER SCIENTIFIC RESEARCH INSTI, 2017, 34(8): 47-51.

Metrics

Viewed

Full text

Abstract

Comments

Abstract
Figures/Tables
Related Articles
Recommended Articles
Metrics
Comments

TOP