The Application Of Neural Network Technique In The Research On Wear Of The Cylinder

The cylinder-piston-ring friction pair of the marine main engine works under adverse environments, such as high temperature, high pressure, strong vibration, corrosion, variable operation condition etc. So there are various possible causes that result in cylinder's wear and tear, and it is very difficult to real-time measure the cylinder's bore because of the particular operation and navigation environment. Great efforts have been made to solve this problem by the author.Firstly, the cylinder-piston-ring tribology system is analyzed on the basis of forerunners' researches in the paper, reveals the structure of the system, and then carries out further research on the structure. After consulting a great number of experts, the weight of the first and second layers are worked out by the method of Analytical Hierarchy Process, which shows the quantitative effect of the factors. The above results will throw light on latter research.Secondly, for the sake of deciding the excessive wear of the cylinder, the paper studies the common symptoms of the failure during the consultation, and complements the symptoms. The author takes advantage of DELPHI expert questionnaire in figuring out the weights, which will become the criteria of estimating the excessive wear.Finally, the author develops a Simulation System of the Wear of the Cylinder (SSWC) based on above research and the information of main engine of the ship such as engine log and telephone note. The author researches accumulation effects of the ordinary parameters used by engineers frequently, and obtains an application model which includes multi-input and single output, then finds mapping law between input and output by Artificial Neural Network, which makes the model applicable. Given inputs, the network will give output梩he wear value of the cylinder. The SSWC system has been tested by data of a SULZER 5RND68 diesel used as main engine of a ship. The author computes the samples, most of them were trained by Radial Basis Function (RBF) Neural Network, and checks the model by the rest of samples. The result is satisfactory.The conclusions drew in the paper will provide useful reference when the marine engineers make decision, also build up a strong basis for the research on wear life of the cylinder.