Due to the impact of the environment in the actual operation of rolling bearings, various failures will occur when they are impacted and worn.In the industry, sensors are often used to collect vibration data to monitor it, but these data often have the characteristics of large samples, many types, and unobvious labels. If judgment and diagnosis are made manually, the time period for data marking is long, and failure to do so will consume human resources and cause economic losses.
Article Directory
foreword
滚动轴承是核心部件的组成部分之一，往往被称为“机器工业关节的主要构成者”。它常常在风力发电机，航空电机中担任着传动部件的主要支撑者。
I. Introduction
The diagram below shows the application of bearings in these systems. These industrial systems are often operated in the open air, coupled with the influence of various uncertain factors during operation, resulting in great variability in their working conditions. Under the condition of variable load and variable working conditions, various components greatly increase the probability of failure. Rolling bearings are responsible for transmitting loads and supporting loads in these systems, and their failure rate is as high as 40% compared to other components. A series of accidents will be caused after the failure of the rolling bearing. If the failure is relatively minor, all parts of the transmission components will be damaged and shut down due to impact, resulting in economic losses. The size of the economic loss is judged according to the importance of the task, often The economic loss caused by each task is immeasurable. When the fault is serious, if the parts are not replaced immediately, the entire machine equipment will be affected. In serious cases, it can cause the transmission parts to break and cause a great safety accident. Therefore, how to monitor and diagnose bearing faults is very necessary.
2. Abnormal detection and fault diagnosis
Anomaly detection is mainly to monitor the state of the rolling bearing to determine whether there is an abnormal working condition. At this stage, it is not necessary to accurately diagnose the type of his fault, but to detect and distinguish the abnormality. Fault diagnosis requires accurate fault location of the data to determine which type of fault has occurred in the bearing.
In the traditional artificial judgment stage, the most primitive observation method is often used for judgment. Through artificial listening, to judge the different sounds of rolling bearings when they are operating in different working conditions, experienced workers can often distinguish and diagnose. In the future, the emergence of some electronic diagnostic tools has slightly improved this traditional method. In the signal processing stage, with the continuous innovation of technology, the detection method has been improved, and temperature analysis method, oil sample analysis method, and vibration method have emerged, among which the detection method of vibration method is the most prominent. The vibration detection method is further divided into simple diagnosis and precise diagnosis. Both diagnoses require various methods of signal analysis. The simple diagnosis mainly uses the time-domain analysis method to make a rough judgment by judging its various characteristic indicators. The precision diagnosis method is mainly the spectrum analysis method. The diagnosis is completed by converting the signal into the frequency domain and drawing a spectrogram to compare its characteristic frequency. After that, various improved methods for noise reduction and feature extraction have emerged, such as EMD decomposition, Greek Halbert-Huang transform to draw envelope spectrum, wavelet packet decomposition.
3. Artificial intelligence method
At this stage, due to the large increase in collected data, electromechanical equipment has entered the era of big data. At the same time, the data of bearings has also increased significantly, and the traditional method is obviously not suitable for this "big data" background era. Therefore, the fault detection and fault diagnosis of rolling bearings has also entered the era of artificial intelligence based on big data.