Research And Implementation Of Vibration Fault Identification Technology For Rotating Machinery Based On Machine Learning

Rotating machinery is an indispensable type of machinery and equipment in modern society and is widely used in industrial fields.In industrial production,in order to ensure the safe and high-load operation of equipment systems and avoid accidents,it is necessary to monitor the operating status of rotating machinery at any time,and timely and accurately identify and handle the failures of rotating machinery.With the continuous development of industrial technology,the operation of rotating machinery is becoming more and more precise and complicated,so higher requirements for the fault identification of rotating machinery are also put forward.As an important technology in the development of information technology,machine learning can be data-centric and integrate multi-disciplinary advantages.It is an effective tool to improve efficiency and ensure benefits.Based on this,the machine learning theory is integrated into the field of rotating machinery fault recognition,which meets the current development trend and the requirements of the times,and can improve the accuracy,timeliness and reliability of rotary fault recognition.Traditional methods and emerging methods for fault recognition of rotating machinery have their own limitations,so appropriate algorithms need to be selected for model construction in actual work.This paper uses machine learning theory to carry out targeted research on vibration fault identification of rotating machinery.First of all,this paper takes the rotor as the research object,and introduces the classification of vibration failures in rotating machinery.It mainly includes three types of forced vibration and oil film oscillation:rotor unbalance,friction between moving and stationary parts,and loose connection of rotor support parts;two types of self-excited vibration:oil film oscillation and oil film vortex.These provides a theoretical basis for vibration fault identification of rotating machinery in the following.Second,a fault recognition model based on fault discrimination is designed.If the fault identification module determines that the input data belongs to abnormal data,then send the data to the fault recognition module for further fault classification.The fault recognition module uses an improved long-short-term memory recurrent neural network,this network integrates long and short-term memory recurrent neural network and deep confidence network,followed by softmax classifier.Finally,in the experimental part,the development language and experimental environment used by the model,as well as the data preprocessing method,are introduced.Then,the experimental bench failure data and field failure sample data are used to perform experimental simulation.And compared with other classification algorithms.The inferior classic three types of evaluation indicators,recall rate,accuracy rate and accuracy rate,are used to confirm the superiority and efficiency of the vibration fault identification technology of rotating machinery proposed in this paper.