Research On Diesel Engine Fault Diagnosis Based On Improved ELM

As the source of power for mechanical equipment,diesel engine’s operation condition plays a leading role in the safety of the whole machine.Efficient and accurate fault diagnosis of diesel engines can not only reduce unnecessary maintenance costs,but more importantly,prevent and control the occurrence of dangerous accidents.The cause of the fault based on the working principle of the diesel engine is analyzed in this article,and a comprehensive and systematic analysis of the fault of the diesel engine according to the fault diagnosis process is conducted.The main research contents are as follows:(1)Aiming at the problem that the diesel engine vibration signal contains strong background noise,the SVD algorithm is introduced to process the noisy signal.It can be seen from the SVD principle that what restricts its performance is the selection of the effective order of singular values.After analyzing the shortcomings of the singular value difference spectrum method,a strategy with the smallest error is proposed to determine its effective order.Experimental results show that this method not only effectively weakens the influence of noise on the original signal,but also retains the effective information of the original signal to the greatest extent,and achieves the purpose of signal-to-noise separation.(2)Aiming at the problem that diesel engine fault characteristics are weak and difficult to extract,VMD algorithm is introduced to decompose the denoised fault signal.Fully considering the influence of the main parameters in the VMD algorithm on its decomposition performance,an improved VMD method based on correlation coefficients is proposed.The optimal value of the VMD method is determined by calculating the Pearson correlation coefficient before and after decomposition.The experimental results show that a suitable value can improve the performance of the VMD.In order to select the features with high sensitivity to the fault type,the Fisher score method is introduced to score the feature parameters,and the feature parameters that best characterize the fault information are selected to form a feature vector to improve the model’s ability to identify diesel engine faults.(3)Aiming at the problem of small diesel engine fault samples and various fault types,ELM was introduced to identify the diesel engine fault types.The effect of different activation functions and different hidden layer nodes on the performance of ELM is analyzed through experiments to determine the selection of these two parameters.Considering that ELM’s weights and bias random assignment will cause the problem of unstable performance,a bat algorithm is introduced to optimize the parameter combination.In addition,corresponding optimization strategies are proposed for the initialization method of bat algorithm,individual update formula,and population diversity.Experiments prove that the improved bat algorithm has better convergence effect than the standard bat algorithm and PSO,and its convergence speed is extremely fast and the improvement effect is obvious.Finally,a diesel engine fault diagnosis model based on IBA-ELM is proposed.The experimental results show that the IBAELM model can quickly and accurately identify the diesel engine fault type compared to ELM,SVM and BP.