Fault Diagnosis Method Based On Big Data Analysis

With the increasing complexity of mechanical equipment and the increasing level of automation,mechanical fault diagnosis technology has received more and more attention.Mechanical fault diagnosis technology is very important in current mechanical engineering and related fields.With the increase of the amount of data monitored by the equipment and the improvement of computer computing power,equipment fault diagnosis has entered the era of "big data",using big data analysis methods for fault diagnosis,allowing the fault diagnosis model to learn characteristics and complete fault identification.Make fault diagnosis more intelligent and automated based on higher accuracy.This paper studies the application of big data analysis methods in fault diagnosis.Firstly,the integrated learning algorithm is studied.It is found that the weak learning device in the GBDT algorithm in the integrated learning algorithm adopts the regression tree algorithm,which is very good for judging whether the device is faulty under the condition of multi-data fusion.The multi-sensor data is used to verify the operation of the fan.The results show that the GBDT algorithm can get a good diagnosis..In the process of fault diagnosis,it is important to determine the fault of the equipment while accurately locating the fault with accurate maintenance of the equipment.Therefore,this paper studies the convolutional neural network algorithm and proposes a one-dimensional convolutional neural network model that directly acts on the vibration signal.The data is enhanced by overlapping sampling to generate samples from the monitored vibration signals.The learning method and visualization technology train and monitor the one-dimensional convolutional neural network model,and use the bearing experimental data to identify and verify,and the verification effect is good.For the direct use of one-dimensional vibration data,there may be a single data feature.The two-dimensional convolutional neural network is studied,and a two-dimensional convolutional neural network fault diagnosis model for vibration signals is proposed.Firstly,the samples obtained by the enhanced sampling are obtained by continuous wavelet transform to obtain the time-frequency map,and then the time-frequency map is compressed and used as the input of the two-dimensional neural network model,and the model is trained in a supervised learning manner.The experimental data of the bearing shows that the accuracy of the model for bearing fault type is 99.98%.For the common 2-D CNN model,the problem of structural optimization is needed to identify the type and degree of fault at the same time.This paper firstly proposes a shallow MSCNN model for fault diagnosis,and uses bearing experimental data to verify.It has been verified and compared with the recognition results of the traditional convolutional neural network,and it has achieved good results.