Agricultural Diesel Engine Lubricating Oil, Abrasive And Concentration Trend Forecast

Much more mechanical accident happens in department of agricultural machinery every year. For the machinery using in farm, its work surrounding is very abominable, and its work condition is complicated and changeful, so mechanical wear is very serious. The wear particles in the engine lubricating oil include a large amount of information about the health condition of the diesel engine, which has important significances for researching the wear particles to predict the break out accident, and it can die out the accident at germination state to realize the defendable maintenance of mechanism equipment.The ferrography is a wear diagnosis technology based on the analysis of wear particle, and the practice has proved that it is one of the most effective methods for wear condition monitoring and wear fault diagnosis. It is more accurate for the analysis about wear particle and the extraction about concentration data. Usual fault diagnosis methods often based on experts'experience or simple math models are hard to deal with nonlinear problems, and it can not to meet the demand of defendable maintenance. Support Vector Machines (SVM) derived from Statistical Learning Theory is a new machine learning technology, it's based on Structural Risk Minimization (SRM) principle, and has shown many advantages compared with ANN. Least Squares Support Vector Machines (LS-SVM), a branch of SVM, can avoid solving the Quadratic Programming (QP), and improve the solving speed, which is characterized with high generalization ability, and can solve many practical problems exist in most traditional learning methods, such as small samples, over learning, high dimension and local minima.At present, most of engines used in agriculture are diesel engine, so wear particles in lubricating oil collected from a forklift diesel engine in different wear condition and typical work condition are researched in the paper. The wear particles are extracted from diesel engine lubricating oil using ferrography analysis method, then the wear condition and wear particle characteristic are analysed. Meanwhile, the Statistical Learning Theory which is the basis of SVM is introduced firstly, and the detailed procedures of using SVM for classification problem and for regressive problem are given. After that, on the basis of analyzing the parameter performance of SVM, a Least Squares Support Vector Machines (LS-SVM) method for wear concentration prediction is presented in this paper.For the first time, a concentration prediction system based on LS-SVM is proposed which can auto-select the parameters in the model. Meanwhile, its methods are testified through data experiments in practice. Results from the LS-SVM modeling are compared with predictions obtained from Artificial Neural Network (ANN) models, which shows that LS-SVM models performed better for wear concentration forecasting than Artificial Neural Network (ANN) models, so it is an effective method for being used in forecasting of wear particle in lubricating oil.