Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/1727
標題: 以類神經網路模式來進行引擎故障診斷
An Artificial Neural Netwok model for the Diagnosis of Automotive Engine
作者: 陳永耀
Ten, Yiin-Yew
關鍵字: Exhaust back pressure;排氣背壓;Temperature of Engine Emission;Exhaust Emission;Engine vibrations;Neural Network;排氣溫渡;廢氣排放;引擎振動;類神經網路
出版社: 機械工程學系所
引用: 【1】張斐章、張麗秋,“類神經網路”,東華書局,2005。 【2】陳耀茂、殷純淵,“類神經網路PCNeuron”,鼎茂圖書公司,2004。 【3】葉怡成,“類神經網路模式應用與實作”,儒林圖書公司,2001。 【4】汪國禎,“汽車學—汽油引擎篇”,復文書局,台南,1998。 【5】葉懷仁,“類神經網路應用於引擎聲訊之辨識”, 碩士論文,國立中山大學/物理研究所,2003。 【6】張志瑋,“類神經網路於引擎故障診斷之研究”, 碩士論文,國立台北科技大學/車輛工程系,2003。 【7】張嘉仁,“振動與噪音信號在引擎診斷上的應用”, 碩士論文,國立台北科技大學/車輛工程系,2003。 【8】王泗華、彭錦樵,“排氣背壓與廢氣排放診斷引擎故障之研究”, 行政院國家科學委員會專題研究計劃成果報告,計劃編號:NSC90-2313-B-005-133,2001 【9】CEFIRO 2.0(A32)引擎修護手冊,裕隆汽車製造股份有限公司,1996。 【10】Wang, S. H. and J. Peng. Development of Engine Trouble Diagnosis Technology Based on Exhaust Back Pressure and Emission. ISMAB 2002 International Symposium on Automation and Mechatronics of Agricultural and Bioproduction Systems: 127~134. Chiayi, Taiwan. 2002. 【11】K.A.et al Marko, K. A.‚ James, J., Dosdall, J. and Mirphy, J., “Automotive control system diagnosis using Neural Nets for rapid pattern classification of large data set”,IJCNN-89-San.,II,pp.13-16.1989. 【12】Marko‚ K.A.‚ Feldkamp, L. A. and Puskorius, G. V., "Automotive Diagnostics Using Trainable Classifiers:Statistical Testing and Paradigm Selection" Proc. IJCNN‚Ⅰ‚ 1990, pp.33-38. 【13】K.et al. Arai,“Application of Neural Computation to Sound Analysis for Valve Diagnosis”, IJCNN-91,I,pp.177-182.1991. 【14】Jesion, G., Gierczak, C. A., Puskorius G. V., Feldkamp L. A., and Butler J. W., “The Application of Dynamic Neutral Networks to the Estimation of Feedgas Vehicle Emissions,” IEEE paper 0-7803-4859-1/98, pp. 69-73, 1998. 【15】D. Yuanwang, z.Meilin,X.Dong,C.Xiaobei,An analysis for effert of cetane number on exhaust emissions from engine with the neural network,Fuel 81,2003,pp.1963-1970 【16】Adnan, P.,Yasar, I.,Halit, Y.and Aysun,E., “Application of artificial neural network to predit specific fuel consumption and exhaust temperature for a Diesel engine” Applied Thermal Engineering 26, 2006, pp. 824-828 【17】Wojtek J.BOCK,Magdalena S.Nawrocka,Waclaw Urbanczyk,Jacek Rostkowski,Tadeusz Maryynkien., “Fiber-optic sensor for automotive applications”,Environmental Technology entre,National Research Council,Ottawa,Canada. 【18】Crouse,W.H. And D.L. Anglin.1997. Automotive mechanics. 10th edi. Macmillan / Mc Graw-Hill Co. New York,USA.
摘要: 
引擎運轉時的排氣背壓值,向來被表示為排氣的順暢與否,且作為討論引擎性能的參考,因此排氣背壓與引擎性能有相當密切的關係;燃油噴射量改變與燃料是否完全燃燒直接影響廢氣排放成份濃度,且其廢氣排放也難維持理想的排放值;若燃燒室內燃料未被點燃而排放到排氣歧管中自燃,將提高引擎排氣溫度;再者,當引擎運轉條件產生變化時,也將導致振動量的增加。本研究即利用引擎排氣背壓值、廢氣排放成份濃度、排氣溫度變化及引擎振動量與類神經網路建構出倒傳遞類神經網路模式應用於引擎故障診斷上。
本研究是以裕隆公司所生產的CEFIRO 2.0(A32)V-6型四行程之電子點火電腦控制噴射引擎,進行惰速定速運轉實驗,作測試故障的項目包括引擎各缸噴油嘴故障,各缸點火失效,各缸火星塞間隙過大或過小,含氧感知器失效等,其中利用訊號擷取儀器紀錄引擎的排氣背壓信號與振動信號,同時利用廢氣分析儀紀錄廢氣成份濃度與溫度感知器量測排氣溫度變化。引擎振動與其轉速有關,排氣背壓亦與其轉速有關。以振動頻譜中轉速頻率的倍頻振幅來表現故障的特徵,而排氣背壓頻譜中轉速頻率的倍頻振幅亦可用來表現故障的特徵。根據實驗取得引擎故障分類資料,建立引擎故障類型,作為倒傳遞診斷與類神經網路學習的依據。
引擎是一個複雜且包含非線性因素的系統,很難以理論數學解析進行引擎故障的診斷。而類神經網路標榜著其具有類似生物神經元的自我組織與平行分散處理的能力。可依照所需的輸出與輸入目標,進行網路學習,學習後的網路具有預測能力,能夠準確辨識未學習的相似引擎故障。這種故障診斷的方式可以提供引擎使用者與維修人員作為參考,以協助其診斷引擎故障,予以即時的排除,期能發揮引擎效能,以減少燃料的耗費。

Exhaust back pressure of an operating engine is represented as the smoothness of emissions and is referred to as discussions on engine efficiency; that is, exhaust back pressure is closely related to engine efficiency. That the amount of fuel injection changes and if fuel combusts completely or not directly influence the thickness of exhaust; the amount of exhaust emissions is difficult to be maintained in a perfect level. If the indoor fuel is not ignited but emitted to exhaust manifold for spontaneous combustion, the temperature of engine emissions will rise. In addition, that the condition of an operating engine changes will increase vibration. This research attempts to employ neural networks to construct anti-communicative neural network model and applies it in the diagnosis of engine malfunction by means of exhaust back pressure, the thickness of exhaust, the temperature of engine emissions, and engine vibrations.
This research performs idle and steady speed experiments with Nissan Cefiro 2.0 (A32) V6 equipped with electronic ignition and engine-control computer. The items of malfunction testing include fuel nozzle malfunction, ignition malfunction, the improper size of the gaps between each spark plug, and oxygen sensor malfunction. In the meantime, a signal collector is utilized to record exhaust back pressure and vibration signal of an engine; an exhaust analyzer to record the thickness of exhaust, and an air temperature sensor to measure changes of temperature. Engine vibrations are relevant to the rotation rate, and so is exhaust back pressure. The amplitude of vibration both in a vibration spectrum and exhaust back pressure spectrum displays features of malfunction. According to the experiments above, we acquire the classified data of and establish types of engine malfunction, both of which are served as basis for anti-communicative diagnosis and neural network emulation.
An engine is a complex system comprising non-linear factors. Engine malfunction is difficult to be diagnosed by means of theoretical mathematics analysis. Neural networks are marked by abilities of self-organization and parallel-distributed management that are similar to a biological neuron. In light of the required input and output goals, we can proceed with network emulation. After the emulation, the network may predict and identify exactly non-emulated similar engine malfunction. The method of malfunction diagnosis may provide references for engine users and maintenance staff, assisting them in diagnosing and obviating engine malfunction, in order to elaborate engine efficiency and decrease fuel consumption.
URI: http://hdl.handle.net/11455/1727
其他識別: U0005-2707200602161800
Appears in Collections:機械工程學系所

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.