Optimization of diesel engine vibration monitoring point layout based on deep learning

Abstract: With the continuous progress of industrial technology, deep learning, as an important branch of artificial intelligence, has brought new possibilities to various fields. In this paper, the optimal location of diesel engine vibration monitoring is studied, focusing on the problem of data overlap and mutual interference in vibration monitoring. By discussing the best position and number of sensors on diesel engine, a method of sensor layout optimization based on graph pool neural network is proposed. In this method, the sensor points are regarded as the nodes of the graph, and the adjacency matrix is processed by the convolution layer of the graph, and the eigenvector of each node is obtained. The feature vectors are screened and sorted by information entropy and independence methods, and the representative feature vectors are selected as important nodes. Finally, classification and screening are carried out in the pooling layer, taking into account factors such as coverage and cost, so as to determine the optimal sensor placement and number.
The experimental results show that the proposed graph pool network model can effectively optimize the layout of sensor measuring points, and has high accuracy and stability. This method is not only suitable for diesel engine vibration monitoring, but also can be extended to other problems requiring sensor layout optimization.