Research on the Standardized Management System and Operational Indicators of Water Control Dikes Based on GA-BP Artificial Neural Network Model

Zhiwei Zhou, Shibiao Fang, Qing Wang, Wenrong TuWater 2023 · DOI

TL;DR

This paper introduces a novel GA-BP (Genetic Algorithm-Backpropagation) artificial neural network model to standardize the management and evaluation of water control dikes. By combining the strengths of AHP for weighting indicators and GA-BP for robust prediction, the model offers a more accurate and comprehensive assessment of dike management, crucial for ensuring the safety and longevity of vital infrastructure.

What Problem Does This Paper Address?

Water control dikes are critical infrastructure for national development, supporting flood control, irrigation, and power generation. However, many existing dikes, particularly in regions like Jiangxi Province, China, suffer from structural aging, increasing risks, and reduced effectiveness. This necessitates efficient and standardized management. Current management practices often face challenges like inadequate responsibilities, funding shortages, and a lack of qualified personnel. Existing evaluation methods, while improving, can be subjective, struggle with correlated indicators, and don't always capture the dynamic nature of engineering management. This paper aims to address these shortcomings by developing a more objective, comprehensive, and data-driven system for evaluating the standardized management of water control dikes.

Methods

The study employs a multi-faceted approach:

  • Analytic Hierarchy Process (AHP): AHP is used to establish a hierarchical structure for evaluation indicators and to determine their relative weights. This involves pairwise comparisons of factors by experts to reduce subjectivity and quantify their importance.
  • BP Artificial Neural Network (BP-ANN): A standard BP-ANN is employed for its ability to model complex, nonlinear relationships. The network is trained on sample data to learn the mapping between dike management indicators and their overall evaluation score.
  • Genetic Algorithm (GA) Enhanced BP-ANN (GA-BP): To overcome limitations of standard BP-ANNs, such as sensitivity to initial weights and potential convergence to local minima, the researchers integrate a Genetic Algorithm. The GA is used to optimize the initial weights of the BP network, effectively searching for a superior solution space and improving the network's ability to find the global optimum. This hybrid approach incorporates momentum factors and adaptive learning rates for enhanced performance.

Datasets

The study utilizes dike data from Jiangxi Province, China. Specifically, they generated simulated training data based on 44 sub-indicators (three-level indicators) derived from a hierarchical index system. This data was created using a random function in MATLAB, with 196 sets of simulated data and an additional 4 data sets representing critical values. For validation, 20 test data points were used.

Key Insights and Ideas

  • Hierarchical Structure is Key: Breaking down dike management into distinct levels (Management Foundation, Safety Management, Operation Management, Maintenance Management, Management Assurance) and then further into secondary and tertiary indicators provides a structured and comprehensive evaluation framework.
  • AHP for Objective Weighting: While expert opinion is inherently subjective, AHP provides a systematic way to aggregate these opinions and derive quantifiable weights for indicators, making the evaluation more objective than purely qualitative assessments.
  • GA-BP for Robustness: The combination of GA and BP is crucial. The GA helps "seed" the BP network with better initial weights, preventing it from getting stuck in suboptimal solutions and leading to more reliable and accurate evaluations. This is particularly important for complex systems like dike management where many factors interact non-linearly.
  • Data Normalization is Essential: The paper emphasizes the critical role of data normalization (using the maximum-minimum method) in ensuring that all input data falls within a consistent range (0 to 1). This prevents features with larger magnitudes from dominating the learning process and improves the overall accuracy of the neural network.

Key Results

  • Improved Accuracy: The GA-BP model achieved an average error rate of 1.2% on test data, with most samples falling within 1% error. This is a notable improvement compared to a standard BP-ANN model, which showed an average error of around 2% in their comparison.
  • Faster Convergence: The GA-BP model demonstrated faster computation speed, achieving desired accuracy in approximately 100 iterations, compared to the standard BP-ANN.
  • Optimal Network Topology: Through analysis of error rates with varying hidden nodes, the study identified that a network with 8 hidden nodes provided the lowest error rate (0.5%).
  • Validation of the Model: The model's performance was validated through a case study in Jiangxi Province, demonstrating its accuracy and comprehensiveness in reflecting the actual situation of dike management.

Limitations and Future Work

  • Subjectivity in AHP: While AHP aims to reduce subjectivity, the initial pairwise comparisons still rely on expert judgment, which can introduce some bias.
  • Data Availability: The study generated simulated data for training. While this is a common practice, real-world, comprehensive datasets could further enhance the model's applicability.
  • Generalizability: The model was validated in Jiangxi Province. Further testing and adaptation might be needed for dikes in different geographical or geological contexts.
  • Future Work: The authors suggest expanding the research to include more in-depth studies on standardized management systems, exploring other evaluation methods (like TOPSIS, SIMUS, SAW, PROMETHEE), and integrating the developed model into existing hydraulic engineering management information systems. They also emphasize the need for continuous improvement and practical application of the model.

Summary updated

Abstract

Water control dikes, as an important infrastructure for national economic and social development, play an important supporting and guaranteeing role in flood control, irrigation, power generation, water supply, tourism, and other aspects. Jiangxi is a major province in water conservancy, with dense rivers and lakes, and it owns tens of thousands of water control dikes of various types. Most of the water control dikes exhibit structural aging, continuous medical risks, and reduced benefits, which urgently require efficient maintenance and standardized management. Management is a complex task, and the level of management directly affects the functional efficiency and service life of dikes. In view of these issues, this study takes dikes as essential and typical water conservancy engineering objects and analyzes the evaluation criteria of safe production and the demands of engineering management. It establishes an evaluation index system suitable for normalized management. The Analytic Hierarchy Process (AHP) model is utilized to determine indicator weights, and a neural network water conservancy engineering evaluation algorithm is constructed to match the evaluation model. Finally, an improved algorithm for the GA (genetic algorithm)-BP (backpropagation) neural network is proposed, incorporating additional momentum factors and considering adaptive learning rates. The developed model is validated through a case study in Jiangxi, China, and the results demonstrate its accuracy and comprehensiveness in reflecting the actual situation. This research is relevant to designers, contractors, and governments seeking solutions to achieve standardized management in water control dikes.
BibTeX
@inproceedings{zhou2023research,
  author = {Zhiwei Zhou and Shibiao Fang and Qing Wang and Wenrong Tu},
  title = {{Research on the Standardized Management System and Operational Indicators of Water Control Dikes Based on GA-BP Artificial Neural Network Model}},
  booktitle = {Water},
  year = {2023},
  doi = {10.3390/w15213713},
}
Figure