5. FAILURE DIAGNOSIS
With the diesel engine simulation method reference values and symptomatic parameters under different performance failures can be obtained without much experimental modeling. With relative deviation analysis of thermodynamic variables one failure's modeling can be utilized at any probable operating condition encountered during ship navigation. By this way artificial neural network can be used in marine diesel engine.
Taking averaged relative deviations of different performance failures at the calculated rated condition and partial condition as input vector X and simulated 18 failure codes as output vector Y, the network conjunction weight set Wl (l=1,2…L) was then achieved with an integrate BP algorithm. Then taking the relative deviations of different failures at partial condition as input vector X and the calculated Wl (l=1,2,…L) as network conjunction weight set the diagnosis results are shown in Table IV where values in row i and column j indicate the confidence of relative deviation sample i to failure j. As can be seen from table IV the diagnosed results are accurate enough by using relative deviation at different running conditions with artificial neural network method.
[ACKNOWLEDGMENTS]
This work was founded by Shanghai Educational Development Foundation and supported by Shanghai Education Committee. The authors would also like to thank professor Tianhao Tang in Shanghai Maritime University for the help in modeling and diagnosis with artificial neural network metbod, and professor Yulong Zhan in Shanghai Maritime University; professor Eiichi Nishikawa and Miss Rong Ma in Kobe University of Mercantile Marine, for kindly offering their advice and help in this study.
[REFERENCE]
[1] Gu Hongthong; 'Marine diesel engine handbook (3)' (in Chinese), National defense industry press. 1979
[2] Hu Yihuai, Wan Biyu, Zhan Yulong, 'Performance failures simulation and informational characters analysis for diesel engine, 'Transactions of Chinese society for Internal Combustion Engines, Vol 17, No.4, 1999
[3] Tang Tianhao etc, 'A study of integrate intelligent marine monitoring system, 'Pro. Of IMECE'97. Shanghai, P. R. China May, 1997
