The Applied Research Of Integrated Neural Networks On The Estimating Of Diesel Abrasion Coefficient

On the foundation of diesel fault diagnosis system, this task further researched the estimating of diesel abrasion coefficient. During the research period, the author combined with the project developed by Northeastern University Equipment Diagnosis Center and Certain Military Area.This article mainly considered the collectivity abrasion, and forecast abrasion, to make sure direct or indirect factor can be found, make the abrasion can be prevent. By using the instantaneous speed, the temperature of lubricating oil, and the pressure of lubricating oil, abrasion quantity build the nonlinear function. Use the ANN to build the relation. Offer the new way of abrasion forecast. This method is suitable to the practice, and provides all right condition for diesel engine.In order to complete the training of the ANN, there must be a great deal of date swatch. But it is impossible to obtain all the dates from the experimentation because of the limited condition. So we should to obtain others dates using math method base on the limited date. This article investigated the date and the relationship between them. Base on the relationship between the date, finally spread the date successfully.Because there are lots of date and two many sorts, if only a single ANN being used, the configuration of the ANN must be very large. The time of the ANN training is too long, and may be it is difficult to train. Considering this instance, the strategies and principles is established with individual neural network and information integration. The strategies and principles for the realization and form of integrated neural network are analyzed and a case is given to prove the efficiency and feasible of the integrated neural network.