Home>Bearing knowledge>Impact pulse method to evaluate the coefficient of rolling bearing faults

/*728*90 created on 2018/5/16*/ var cpro_id = "u3440131";

Impact pulse method to evaluate the coefficient of rolling bearing faults

---

/*250*250 was created on 2017/12/25*/ var cpro_id = 'u3171089';

Foreword Rolling Bearing is an important part of the machinery industry with large volume; its working condition directly affects the working quality of machinery and equipment; the machine equipment caused by rolling bearing problems also has frequent attacks; thus; The evaluation of the condition of the bearing led to the attention of many mechanical depiction and condition monitoring technicians. At that time, the evaluation of the working condition of the rolling bearing; one of the most commonly used and most useful methods in the industrial field was the shock pulse method (SPM method) [1 However, the method for the identification of the bearing condition is the curve method of the experience; this curve is obtained under very ambitious conditions; that is: the bearing load and the speed are within the scale of the drawing; the device of the SPM sensor needs to fully conform to the specification. The detailed working conditions of the industrial site are very messy; the machine equipment in the operation is a multi-dimensional nonlinear system with many uncertain factors and difficult to describe; the accurate model is difficult to mention. The SPM method evaluates the ambition of the bearing operation status. Conditions cannot be completed at the industrial site; for this purpose, the self-learning target should be monitored together in the machine. Detailed condition of the device; adaptively assessing the correction coefficient of the condition of the rolling bearing; making the monitoring system after a period of operation; gradually adapting to the detailed messy situation of the target machine. According to this thinking; we are in China Yangzi Petrochemical When the company's plastics factory granulates multiple rolling bearings to establish an online monitoring system; the SPM monitoring method for aligning rolling bearings; a self-approval coefficient method based on artificial neural network is proposed. The system is now in use at the site; the status of the application is indicated The system can be used to adaptively attack the parameters used for correction; the accuracy of the monitoring system for the evaluation of the working condition of the rolling bearing is improved.
1 According to the principle of BP network self-approval module 1. 1SPM method rolling bearing in the operation; if the roll body touches the short-term area of ​​the raceway surface; the low-frequency impact effect will occur; the impact pulse signal will occur; The resonance of the SPM sensor; the resonance waveform is usually 20 kHz to 60 kHz; the amplitude modulated wave including the low frequency impact and the random disturbance; the circuit is formed by the narrow band filter and the pulse; the pulse sequence including the high frequency and the low frequency is obtained. The SPM method is According to the pulse sequence reflecting the impact force, the bearing condition is judged; and the new standard of the shock pulse value (SV), dBc and dBm, is selected; the decibel value is used in practice.
dBc (Carpet Value) value: also known as carpet value or high frequency value. It is caused by the random excitation of the rolling surface of the raceway; it is related to the smooth condition of the bearing and the thickness of the oil film. It is presented at a high frequency (above 1 kHz) and The value is small.
dBm (Maximum value) value: also known as strong pulse value or low frequency value; indicates the amplitude of the strong pulse sequence in the pulse sequence. It is related to the damage of the rolling bearing and the impurities in the smooth oil; the frequency of presentation is low (below 1 kHz) And the value is larger.
The difference between dBm and dBc is directly related to the bearing fault; the connection between the two is related to the detailed characteristics of the single bearing; it is also related to the messy situation in the field; its accurate mathematical model can not be extracted.
1.2 The principle of the self-revision module is after the bearing is faulty and replaced; the on-site operator fills in the bearing replacement system procedure; on the one hand, the procedure completes the function of bearing replacement; on the other hand, the high and low frequencies when replacing the bearing The values ​​(dBc and dBm) are passed to the BP network; together, the operator's assessment of the suitability and timeliness of this bearing replacement is used to correct the traditional SPM method identification curve.
Artificial neural network is a new science developed in these years [2]; BP neural network is the most widely used artificial neural network; it is a multi-layer network including hidden layers.
The BP model consists of three neural network element levels; the input layer, the hidden layer, and the output layer. The neurons in each layer satisfy the interconnection; the neurons in each layer have no connection. The input and output characteristics are Nonlinear differentiable non-decreasing function; usually taken as Sigmoid function; indicating the fullness of neurons. Due to the introduction of hidden nodes; BP network with three layers of Sigmoid neurons can impede any function with arbitrary precision [3].
The BP network used in practice is the layout of 1-48-1; the input is the low frequency value of the SPM method; the output is the high frequency value; the messy relationship between the high and low frequencies of the SPM method is imposed. Because the neural network is mainly used for accounting It is not classification discrimination; therefore, the precision requirement is high. As long as a single sample reaches an error of less than 8×10-5, it is not lower than the measurement accuracy of the system hardware; that is, the error of the BP network is not the first item of the system error. In order to reach such a high The precision; we choose a large number of hidden layer nodes: 48. The value of this is the speed of the operation is reduced. But this is only a long time in the pre-learning time; in the system practice time; BP network can not learn; Even learning; only a single sample of learning; the number of elections will never exceed dozens of times; so it will not affect the real-time nature of the system.
In the construction of the BP network; we first pre-train the BP network based on the detailed data of the site and the experience data of the field operators; to generate a set of initial weights. The so-called detailed data of the site refers to offline SPM for field operators. Each historical data collected by the instrument. When there is no detailed data on site; use the experience data supplied by the field operator. Select some representative data from the data as a practice sample. After the pre-work is completed; select the above data. The data that is not used as a sample is used as the verification data of the BP network; if the accuracy requirement is not reached, the pre-practice process is repeated.
The time of the system operation; the BP network first uses the initial weight to calculate the input data; to determine whether the accuracy of the reservation can be reached. Since the input data represents the detailed characteristics of the bearing; if the BP network arrives at the reservation accuracy, the BP network is clarified. The detailed condition of the bearing is common; the weight is not corrected, if it is not reached; then the BP network and the bearing detailed condition are not very common; it is necessary to use the BP algorithm to repeat the learning of the weight; until the detailed situation with the scene It tends to be common; it represents the detailed status of the machine. The weights after correction are kept as the basis of the next BP network operation.
Correct the SPM method identification curve; according to the BP network operation results and the replacement evaluation passed by the bearing replacement procedure; that is, the detailed characteristics of the field bearing and the timely degree of the previous bearing replacement. The identification curve after the correction can be used for the measurement monitoring system. To correct and compensate for the impact of industrial conditions on the chaotic conditions of the industrial site on the monitoring method.
2The industrial site uses the SPM method to evaluate the self-batch modification coefficient module of the rolling bearing operation status; it is installed in the analysis software of the online monitoring system of PP granulation of Nanjing Yangzi Petrochemical Company Plastics Factory; more than one year in the industrial field operation; The measured data and the experience of the field operators for many years of accumulation; BP neural network has been used to correct the evaluation curve of the original SPM bearing fault; make it basically fit the practical situation of this PP granulation bearing operation. The bearing is a single-row radial ball bearing with an outer diameter of 360 mm; an inner diameter of 200 mm; a working speed of 1480 r/min. The curve according to the SPM method; when the low-frequency value is 86 dB; clarify that the bearing has been damaged; The practice site personnel were found several times during the inspection; when the above values ​​were reached; the first bearing was now severely damaged; the operation could not be resumed; this fully clarified that there is a certain difference between the experience curve of the SPM method and the characteristics of the detailed machine target. According to this; we use the BP neural network to self-approve the coefficient software; the experience curve of the SPM method has been revised; pre-practice BP network The initial data uses the experience data of the field operator; that is, the demand of the first bearing in the low-frequency impact value dBm is 46dB; the high-frequency value dBc at this moment is calculated by the BP network; thus the modified high and low frequency values ​​are used as The evaluation specification of the No. 1 bearing. Pre-learning and correcting the different bearings one by one; each bearing has the corresponding specification for evaluating the movement condition. The field has been used for more than a year; compared with the previous one; the bearing fault caused the stop event Cuts; ensure long-term safe operation of granulation. Practice proves; the evaluation data after the evaluation curve of SPM method is in line with the practice status of PP granulation; it is very suitable.
3 Conclusion This paper points to the impact of the extremely messy working conditions of the industrial site on the online monitoring and diagnosis system of the machine; detailed monitoring of the rolling bearing by the SPM method; using the new artificial neural network science developed in these years; the test is constructed based on the BP neural network. The self-approval coefficient module. After a period of on-site learning, the system will be able to represent the on-site conditions and the detailed characteristics of the on-site target machine; the adaptive attack represents the detailed correction coefficient of the on-site target machine; the SPM method is used to identify the bearing condition. The experience curve; to compensate for the impact of on-site cluttered working conditions on the SPM monitoring method. The practice of this system in China Yangzi Petrochemical Company; the construction of the certification system is successful and reliable.

---

Recommend to friends comments close window

Bearing related knowledge

Oscillation motor imported bearing device The door bearing is selected from the analysis of the smooth grease. The accuracy grade of the bearing is self-smoothed. The four typical dynamic pressure bearings and their application environment

This article links to http://
Please indicate the bearing network http://

Previous: The reason for the abnormal phenomenon of NTN rolling bearing attack: The high quality tapered roller bearing 32009X

Simn Furnace Lining Carbon Paste

Simn Furnace Lining Carbon Paste,Simn Electric Furnace Lining Carbon Paste,Electric Furnace Lining Carbon Paste,Simn Lining Carbon Paste

Pingluo Zhongxing Carbon Co.,Ltd , http://www.ztecarbon.com