Research On Intelligent Fault Prediction Of Marine Diesel Engine

Plan Recognition is an active research focus on artificial intelligence now. It has been widely used in knowledge inference and intrusion detection etc. And it has achieved good results. Its main idea is the process that inferring to the reasonable, complete and comprehensive planning describe according to the observed fragmental Agent and trivial phenomenon. A plan recognizer plans can supplement the actually but not have be observed phenomenon, and reasonably infer the possible actions of Agent which could be applied in the future.The diesel fault prediction is the technology to identify the real-time situations of the mechanical devices or its components by processing and analysis the measured signals, combining with the historical diagnostic situation, and to predict the relevant exceptions, faults and future working conditions. The paper introduces Planning Recognition into the diesel fault prediction field on the basis of the others'research results.This paper focus on plan recognition and on the background of the fault prediction for marine diesel engine, involving the following aspects:(1) Improve Plan knowledge graph, the original knowledge graph consider that abstract node only exists in solution graph, and the support is 1, but the branch of abstract node have possibility. Experiments show that the improved model can effectively improve the prediction accuracy.(2) The working process of large-scale low-speed two-stroke diesel engines 5S50MC is performed by the AVL Boost, prove the availability of the data model by comparing the test data and simulation data, and prepare the conditions for plan fault prediction.(3) The paper realized the fault prediction system of marine diesel engine based on plan recognition theory, and the experimental results proved the effectiveness of fault prediction, further confirms the feasibility of plan recognition theory in fault prediction. The paper research on the fault prediction based on planning recognition through analyzing the plan recognition prediction ability, and carry out the availability of the method. On the one hand the paper expended the application space of plan recognition theory; on the other hand it enriched the field of fault prediction.