Structural Condition Assessment of Urban Drainage Pipes Based on Particle Swarm Optimization-Extreme Learning Machine

Abstract:

Structural condition assessment of drainage pipes has been a major concern for asset managers in maintaining the required performance of urban drainage systems. This paper proposed a neural network model combing extreme learning machine (ELM) and particle swarm optimization (PSO) to classify the structural condition status of urban drainage pipes. Besides, in an attempt to look for better classification performance, it used the PSO algorithm to optimize the input weight matrix and the hidden layer offset of ELM. Moreover, it validated the PSO-ELM model by using the dataset supplied from the Yangshan Bonded Port Area in Shanghai. Furthermore, it compared the predictive performance of PSO-ELM with ELM on the same dataset. The result shows that the PSO-ELM can achieve a higher classification accuracy by utilizing less neuron nodes in the hidden layer, and improve the fitting capability of ELM. The method proposed has feasibility and effectiveness for structural condition assessment of urban drainage pipes.