Research On Fault Diagnosis And Performance Assessment Method Of Rotating Machinery Based On Data

Most of the mechanical equipment is in a harsh working environment,which are not suitable for operators to operate on site.Therefore,it is very meaningful to collect the operational data of mechanical equipment in real time and remote diagnose equipment's failure and assess equipment's performance.As an important part of rotating machinery,rolling bearings are more prone to failure.According to the data,30% of mechanical equipment failures are caused by damage of rolling bearings.Therefore,analyzing the mechanisms of rolling bearings,and then diagnosing the faults of bearings,final assessing the performances of bearings are key issues in the research of rotating machinery.Aiming at the problem that the traditional fault diagnosis methods used a small number of features in the feature extraction process and cannot reach the optimal fault diagnosis accuracy,a fault diagnosis method based on ensemble empirical mode decomposition and multi-features fusion is proposed.The method can accurately classify the faults of the inner rings and outer rings of the rolling bearings under different severity of damage,and the effect is better than the result of extracting a small number of features.However,this method has many features that are extracted,resulting in a slower calculation speed.Aiming at the problem that multi-features are complicated and the parameters of traditional support vector machine classification are easy to fall into local optimum,a fault diagnosis method based on ensemble empirical mode decomposition and cuckoo search-support vector machine is proposed.This method only extracts the root mean square as the feature of the intrinsic mode function,and uses the cuckoo search to adaptively optimize the parameter of support vector machine.The method can not only accurately diagnose the faults of the inner rings and outer rings of the bearings under different degrees of damage,but also accurately diagnose the different fault positions of the bearing rolling elements.Based on variational mode decomposition has two problems: the center frequency is easy to fall into local optimum and the number of decomposition layers needs artificial selection.In addition,the traditional performance assessment methods lack the consideration of time sequence information,so the fitting error to the signal data is large.Aiming at the above problems,a method for assessing the performance of rolling bearings based on improved variational mode decomposition and long short-term memory network is proposed.The method uses the cuckoo search to optimize the iteration of center frequency,and adaptively selects the signal decomposition layers by the instantaneous frequency theory to completely solve the problems of variational mode decomposition.It is proposed to use long short-term memory network to fit the extracted time sequence of variance features,which greatly reduces the fitting error and analyses the fitted signal to find out the variation law of rolling bearings performance.This method introduces long short-term memory network into the performance assessment method of rolling bearings,and provides an effective solution for assessing the performance of rolling bearings.