Research On Multi-objective Optimization Of The Construction Process Of A Data-driven Trailing Suction Dredger

Production,energy consumption,and overflow loss,as the three core factors that affect dredging efficiency and energy consumption in the construction of trailing suction dredgers,have long been the subject of attention by operators and scientific and technological workers.With the continuous development of dredging engineering,the construction data of trailing suction dredgers has shown exponential growth.The multi-objective optimization research of the construction process of trailing suction dredgers based on data-driven has likewise become a hot and difficult topic in the current dredging field research.This article focus on the problems of low production,high energy consumption,and large overflow loss during the construction process of trailing suction dredgers.Firstly,the deep stack anomaly detection algorithm and temporal correlation statistical method are used to preprocess the construction data;Secondly,feature selection and fuzzy clustering algorithms were used to optimize the structure of production and energy consumption prediction models and enhance the quality of construction sample data.Production prediction models,energy consumption prediction models,and overflow loss prediction models based on stacked fuzzy neural networks were constructed;Finally,a multi-strategy artificial bee colony algorithm based on dynamic weights is used for multi-objective optimization of production,energy consumption,and overflow losses,achieving the goal of improving dredging efficiency and reducing energy consumption.Research ideas and specific work of this article are as following:(1)In response to the problems of multi-source heterogeneity,noise interference,and monitoring anomalies in the historical construction data of suction dredgers,this thesis adopts deep stacking strategy and time-series related anomaly detection methods to preprocess the construction data,and verifies the accuracy of the detection results through experiments.(2)In response to the problems of high model complexity and uneven distribution of data samples in the process of predicting the production and energy consumption of trailing suction dredgers,this thesis adopts feature reduction dimension fuzzy C-means clustering resampling method to optimize the model structure and enhance the quality of data samples.At the same time,this thesis also uses stacked fuzzy neural networks as feature extractors to carry out the production prediction model The construction and evaluation of energy consumption prediction models and overflow loss prediction models provide a reliable model structure for subsequent multi-objective optimization research.(3)In response to the multi-objective optimization problem of drag suction dredgers,a multi dynamic weight neighborhood search strategy is adopted to optimize the traditional artificial bee colony algorithm,and the entropy TOPSIS method is synchronously used for multi-objective optimization analysis of the given Pareto solution to obtain construction guidance opinions.