Method for identifying steel biting impact in vibration signal

Abstract

The invention discloses a method for recognizing steel biting impact in vibration signals, and relates to the technical field of rolling mill equipment monitoring. The method comprises the steps that the vibration signals of a target rolling mill within a certain time period are collected; searching steel biting impact signals from the collected vibration signals, and screening a plurality of steel biting impact signals to form a steel biting impact signal database; vibration signals of a target rolling mill are collected in real time, cross-correlation operation is carried out on the vibration signals and each steel biting impact signal in the steel biting impact signal database through a cross-correlation function, and steel biting impact signals in the vibration signals of the target rolling mill are determined; and judging the effectiveness of the steel biting impact signal. According to the method, steel biting signal recognition can be carried out without depending on the working condition of the rolling mill, parameter setting is simple, the mechanism of the steel biting process in the vibration signal is considered, the engineering application effect is better, and in addition, compared with deep learning, the model parameter training amount is small, and the needed sample amount is small.

Description

Technical Field

[0001] The invention relates to the technical field of rolling mill equipment monitoring, and in particular to a method for identifying steel biting impact in a vibration signal. Background Art

[0002] Rolling mills are key equipment for metallurgical production enterprises. They can press heated billets or ingots into plates, bars, pipes and other products of the required shape and size. Metallurgical enterprises suffer huge economic losses every year due to sudden major equipment accidents and "excessive maintenance". Therefore, online monitoring and diagnosis of the status of rolling mill equipment to reduce or even avoid losses has become an industry trend. Currently, the status monitoring of rolling mill equipment is generally considered from the perspective of vibration signals, but its vibration signals are easily disturbed by complex working conditions, especially the strong impact signals generated when steel enters and leaves the rollers.

[0003] Vibration sensors are often used for the condition monitoring of rolling mills to monitor whether there are mechanical failures in the rolling mills. However, during the operation of the rolling mill, when the steel enters the rolls and is clamped by the rolls and begins to deform (biting the steel), a strong impact signal will be generated. This impact signal will seriously affect the data analysis, and it is necessary to reasonably eliminate it. The current commonly used method is usually to locate the impact time point in the vibration signal according to the time point of the rolling pass signal and the rolling force signal in the process signal, but its disadvantage is that in actual implementation, the process signal may not be obtained, or the timestamp is not easy to align. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a method for identifying steel biting impact in a vibration signal.

[0005] In order to solve the above technical problems, the technical solution of the present invention is as follows: A method for identifying steel biting impact in a vibration signal, comprising: Collect vibration signals of the target rolling mill within a certain period of time; Find the steel biting impact signal from the collected vibration signals, and select several steel biting impact signals to form a steel biting impact signal database; The vibration signal of the target rolling mill is collected in real time, and a cross-correlation operation is performed on the vibration signal of the target rolling mill with each steel bite impact signal in the steel bite impact signal database through a cross-correlation function to determine the steel bite impact signal in the vibration signal of the target rolling mill; Determine the validity of the steel biting impact signal.

[0006] As a preferred solution of the method for identifying steel biting impact in the vibration signal of the present invention, the acquisition of the vibration signal of the target rolling mill within a certain time period includes: An acceleration vibration sensor is arranged on the target rolling mill, and the vibration signal of the target rolling mill within a certain time period is collected by the acceleration vibration sensor.

[0007] As a preferred solution of the method for identifying steel biting impact in the vibration signal of the present invention, the step of searching for the steel biting impact signal from the collected vibration signal includes: Differentiating the rolling force of the rolling mill; The moment when the difference value is greater than the threshold is selected as the steel biting moment; The steel biting impact signal is taken as the center of the steel biting impact signal, and the moments at which half the length of the steel biting impact signal is taken on both sides are taken as the boundary points of the steel biting impact signal to screen out the steel biting impact signal.

[0008] As a preferred solution of the method for identifying steel biting impact in the vibration signal of the present invention, the screening of a plurality of steel biting impact signals to form a steel biting impact signal database comprises: Intercept several steel biting impact signals and analyze their power spectrum differences; Classifying the steel biting impact signal based on power spectrum difference; The steel-biting impact signal with the largest kurtosis index in each type of steel-biting impact signal is taken as a sample representative and added to the steel-biting impact signal database.

[0009] As a preferred solution of the method for identifying steel bite impact in the vibration signal of the present invention, the real-time acquisition of the vibration signal of the target rolling mill and the cross-correlation operation of the vibration signal with each steel bite impact signal in the steel bite impact signal database through the cross-correlation function include: The vibration signal collected in real time is cross-correlated with the steel biting impact signal in the steel biting impact signal database in turn, and the cross-correlation function calculation formula is: ,in, represents the cross-correlation function sequence between the steel biting impact signal and the actual vibration signal, x represents the steel biting impact signal, y represents the vibration signal collected in real time, m represents the length of the steel biting impact signal, n represents the length of the vibration signal collected in real time, and k represents the length of the cross-correlation function sequence.

[0010] As a preferred solution of the method for identifying steel bite impact in the vibration signal of the present invention, the method of determining the steel bite impact signal in the vibration signal of the target rolling mill includes: The absolute value of the cross-correlation function sequence is greater than the judgment threshold The index position is used as the center index of the steel biting impact signal , ; Indexed by the center of the steel impact signal Left and right sides The signal within the point range is determined as the steel biting impact signal , .

[0011] As a preferred solution of the method for identifying steel biting impact in the vibration signal of the present invention, the validity of the steel biting impact signal is judged as follows: Determine the number of steel bite impact signals; If there is one steel-biting impact signal, the validity of the steel-biting impact signal is judged based on the front and back amplitudes of the steel-biting impact signal; if there is more than one steel-biting impact signal, the validity of the steel-biting impact signal is judged based on the front and back amplitudes of the steel-biting impact signal and the time difference between the centers of adjacent steel-biting impact signals.

[0012] As a preferred solution of the method for identifying steel biting impact in the vibration signal of the present invention, if there is only one steel biting impact signal, judging the validity of the steel biting impact signal based on the amplitudes before and after the steel biting impact signal includes: Calculate the left boundary of the steel biting impact signal to The effective value of the point data ; Calculate the right boundary of the steel biting impact signal to The effective value of the point data ; like , then the steel-biting impact signal is valid. , then the steel-biting impact signal is invalid, where j represents the value index number from the boundary of the steel-biting impact signal to both sides, which is used to judge whether the two sides of the steel-biting impact signal are no-load signals and rolling signals, and k is the minimum proportional coefficient of the no-load signal amplitude to the rolling signal amplitude.

[0013] As a preferred solution of the method for identifying steel biting impact in the vibration signal of the present invention, if the steel biting impact signal is greater than one, judging the validity of the steel biting impact signal based on the amplitudes before and after the steel biting impact signal and the time difference between the centers of adjacent steel biting impact signals includes: Calculate the left boundary of the steel biting impact signal to The effective value of the point data ; Calculate the right boundary of the steel biting impact signal to The effective value of the point data ; Calculate the time series difference corresponding to the signal center of adjacent steel biting impact signals , and determine whether it is greater than the minimum time interval ; like and , then all steel-biting impact signals are valid; like and , then all steel biting impact signals are invalid; like And only one steel bite impact signal satisfies , then the steel biting impact signal is valid. If there are multiple steel biting impact signals that meet , then the entire section of actual mill signal is filtered out; like and , then all steel biting impact signals are invalid.

[0014] As a preferred solution of the method for identifying steel biting impact in the vibration signal of the present invention, if the number of value indexes on both sides of the steel biting impact signal boundary is less than j, all values ​​on this side are taken.

[0015] The beneficial effects of the present invention are: The present invention can identify steel biting signals independently of the rolling mill operating conditions. Moreover, compared with other machine learning methods such as clustering and classification, in a large number of actual signal tests, the parameter setting of the present invention is simple, and the mechanism of the steel biting process in the vibration signal is taken into consideration, and its engineering application effect is better. In addition, compared with deep learning, the present invention has less model parameter training and requires less sample size. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0017] Figure 1 A schematic flow chart of a method for identifying steel biting impact in a vibration signal provided in an embodiment; Figure 2 It is a schematic diagram of the interception of the rolling mill vibration signal and the steel biting impact signal in the embodiment; Figure 3 A schematic diagram showing the trend of rolling force and rolling passes of a typical rolling mill provided in the embodiment; Figure 4 Schematic diagram of a typical rolling force differential sequence of a rolling mill in the embodiment; Figure 5 A schematic diagram of power spectra of multiple different steel biting impact signals provided in the embodiment; Figure 6 Schematic diagram of the cross-correlation spectrum and threshold value of the steel biting impact signal and the actual vibration signal in the embodiment; Figure 7 Schematic diagram for confirming the validity of the steel biting impact signal in the embodiment. DETAILED DESCRIPTION

[0018] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific implementation modes and in combination with the accompanying drawings.

[0019] Figure 1 The flowchart of the method for identifying steel biting impact in the vibration signal provided in the embodiment of the present application is as follows. The method includes steps S101 to S104. The specific steps are described as follows: Step S101: collecting vibration signals of a target rolling mill within a certain time period.

[0020] Specifically, an acceleration vibration sensor is installed on the target rolling mill. When the target rolling mill is running, the acceleration vibration sensor can collect vibration signals during the operation of the rolling mill in real time. The acceleration vibration sensor collects vibration signal data of the target rolling mill over a period of time.

[0021] Step S102: searching for steel-biting impact signals from the collected vibration signals, and screening a number of steel-biting impact signals to form a steel-biting impact signal database.

[0022] Specifically, the impact signal caused by steel biting is searched from the collected vibration signal. Typical impact signals caused by steel biting are as follows: Figure 2 This process requires communication with experienced engineers on site or confirmation with other process signals whether the signal is a steel bite impact signal caused by steel bite.

[0023] The method used in this embodiment is to filter the impact signal through the rolling force signal of the rolling mill. Usually, the vibration signal acquisition system of the rolling mill is independent of the working condition signal acquisition system of the rolling mill. The purpose of this embodiment is to identify the steel biting impact signal of the rolling mill without relying on the working condition signal, and then filter out the impact to obtain the vibration data of stable operation. However, when constructing the steel biting impact signal database in the early stage of the method, the working condition signal can be relied on for screening. The specific method is as follows Figure 3 , Figure 4 As shown, Figure 3 is the rolling force signal and rolling pass signal of the rolling mill. When rolling in each pass, the rolling force of the steel is low when it is not in the rolling mill and is idling. When the steel enters the rolling mill and is rolling, the rolling force is high. Figure 4 As shown, the rolling force is differentiated, and the moment when the differential value is greater than the threshold, that is, 2x10'N, is selected as the steel biting moment. The vibration signal corresponding to this moment is the center of the steel biting impact signal. Taking this moment as the center, m / 2 points on both sides are taken as the steel biting impact signal.

[0024] In this embodiment, the sampling frequency of the vibration signal =5120, the length m of the steel biting impact signal is 548, which includes the entire process from the beginning of the impact to the end of the attenuation.

[0025] A number of steel biting impact signals with a length of m are screened from the collected vibration signals to form a steel biting impact signal database, which is convenient for subsequent screening of steel biting impact signals in actual vibration signals. The screening method is as follows: Firstly, several steel-biting impact signals with a length of m are selected from the collected vibration signals, and their power spectrum differences are analyzed. Then, the steel-biting impact signals are classified based on the power spectrum differences to obtain several types of steel-biting impact signals with different power spectra. Finally, the steel-biting impact signal with the largest kurtosis index in each type of steel-biting impact signal is taken as a sample representative and added to the steel-biting impact signal database.

[0026] See also Figure 5 In this embodiment, the power spectrum of the steel biting impact signal mainly appears at 30 Hz, 50 Hz, 200 Hz, 300 Hz and 500 Hz, and the signals with close peak values ​​at the corresponding frequencies are selected as the same type of steel biting impact signals.

[0027] Numerous documents indicate that steel bite impact is the most common external load impact during the operation of a rolling mill. In a very short period of time after the rolled piece is bitten, the system torque will have a sudden torque spike, and the rolling mill will produce strong torsional vibration. The impact response caused by steel bite monitored by the acceleration vibration sensor mainly includes a high-amplitude step signal that decays rapidly with the natural frequency of the rolling mill torsion (the natural frequency may change due to the change in the clearance at each position of the bearing, but should be relatively consistent under normal conditions). At the same time, because the vibration sensor is installed at the roller bearing and gearbox position of the rolling mill, the signal is superimposed with the signal generated by the operation of the equipment such as the meshing frequency of the gear.

[0028] like Figure 5 As shown in the figure, by comparing the power spectra of randomly selected steel rolling vibration impacts from different batches and passes, it can be observed that the main frequency components of multiple steel biting impact signals are similar, but the amplitudes of the frequency components are different.

[0029] To better illustrate this embodiment, according to the spectral component structure of the steel biting impact signal, the steel biting impact signal with certain differences is selected for classification, and the steel biting impact signal with the largest kurtosis index is selected as the representative in each category. and As a representative sample, a steel biting impact signal database is formed.

[0030] Step S103: collecting the vibration signal of the target rolling mill in real time, and performing a cross-correlation operation on the vibration signal of the target rolling mill and each steel bite impact signal in the steel bite impact signal database through a cross-correlation function to determine the steel bite impact signal in the vibration signal of the target rolling mill.

[0031] Specifically, the vibration signal collected in real time and the steel biting impact signal in the steel biting impact signal database are sequentially cross-correlated, and the cross-correlation function calculation formula is: ,in, represents the cross-correlation function sequence between the steel biting impact signal and the actual vibration signal, x represents the steel biting impact signal, y represents the vibration signal collected in real time, m represents the length of the steel biting impact signal, n represents the length of the vibration signal collected in real time, and k represents the length of the cross-correlation function sequence.

[0032] The cross-correlation function sequence between the steel biting impact signal and the actual rolling mill signal is as follows: Figure 6 As shown, the absolute value of the cross-correlation function sequence is greater than the judgment threshold The index position is used as the center index of the steel biting impact signal , the center index of the steel bite impact signal is defined as: . Indexed by the center of the steel impact signal Left and right sides The signal within the point range is determined as the steel biting impact signal , the steel bite impact signal is defined as: In this formula It is the vibration signal data collected in real time.

[0033] Above Indicates the judgment threshold for judging whether the position of the cross-correlation sequence is a steel biting impact signal. In this embodiment, The value is 200. Indicates the single-side window length for selecting the steel biting impact signal. , m is the length of the steel biting impact signal. In order to reduce the impact of the impact signal, the range is appropriately expanded, so a coefficient of 1.5 is added, and the steel biting impact signal is: .

[0034] Step S104: judging the validity of the steel biting impact signal.

[0035] Specifically, after the steel biting impact signal is identified, the following logic is performed to determine whether the identified steel biting impact signal is valid: Step S104a: Determine the number of steel-biting impact signals. When there is only one steel-biting impact signal, the validity of the steel-biting impact signal only needs to be judged based on the amplitude before and after the steel-biting impact signal. When there is more than one steel-biting impact signal, the validity of the steel-biting impact signal needs to be judged based on the amplitude before and after the steel-biting impact signal and the time difference between the centers of adjacent steel-biting impact signals.

[0036] Step S104b: If there is only one steel biting impact signal, the validity of the steel biting impact signal is determined based on the amplitudes before and after the steel biting impact signal. Specifically: Calculate the left boundary of the steel biting impact signal to The effective value of the point data ; Calculate the right boundary of the steel biting impact signal to The effective value of the point data ; like , then the steel-biting impact signal is valid. , then the steel biting impact signal is invalid.

[0037] Among them, j represents the value index number from the boundary of the steel biting impact signal to both sides, which is used to judge whether the two sides of the steel biting impact signal are no-load signals and rolling signals, and k is the minimum proportional coefficient of the no-load signal amplitude to the rolling signal amplitude.

[0038] In this embodiment, j is 10000 and k is 2. It should be noted that if the steel biting impact signal is just at the edge of the rolling mill signal, that is, the number of indexes to the left or right is less than j, all values ​​on that side are taken.

[0039] Step S104c: If there is more than one steel biting impact signal, the validity of the steel biting impact signal needs to be determined based on the amplitudes of the steel biting impact signal before and after and the time difference between the centers of adjacent steel biting impact signals. Specifically: Calculate the left boundary of the steel biting impact signal to The effective value of the point data ; Calculate the right boundary of the steel biting impact signal to The effective value of the point data ; Calculate the time series difference corresponding to the signal center of adjacent steel biting impact signals , and determine whether it is greater than the minimum time interval ; like and , then all steel-biting impact signals are valid; like and , then all steel biting impact signals are invalid; like And only one steel bite impact signal satisfies , then the steel biting impact signal is valid. If there are multiple steel biting impact signals that meet ,Theoretically, this is not in line with the mechanism and its effectiveness cannot be judged, so the entire section of the actual mill signal is filtered out; like and , then all steel biting impact signals are invalid.

[0040] See Table 1 for specific judgment criteria.

[0041]

[0042] Table 1 According to the rolling process, if the steel is repeatedly rolled, the cross-section of the steel will become smaller and smaller, the roller speed will become faster and faster, and the corresponding time for a complete pass through the roller will become shorter and shorter. The time of the last rolling pass is taken as , that is, the normal steel biting interval should be greater than this time. The value is 7 seconds.

[0043] Figure 7 The blue dotted box in the figure shows the single-side window length of the steel biting impact signal.

[0044] In addition to the above embodiments, the present invention may also have other implementation modes; any technical solutions formed by equivalent replacement or equivalent transformation shall fall within the protection scope required by the present invention.

Claims

1. A method for identifying steel biting impact in a vibration signal, characterized in that: include: Collect vibration signals of the target rolling mill within a certain period of time; Find the steel biting impact signal from the collected vibration signals, and select a number of steel biting impact signals to form a steel biting impact signal database; The vibration signal of the target rolling mill is collected in real time, and a cross-correlation operation is performed on the vibration signal of the target rolling mill with each steel bite impact signal in the steel bite impact signal database through a cross-correlation function to determine the steel bite impact signal in the vibration signal of the target rolling mill; Determine the validity of the steel biting impact signal.

2. The method for identifying steel biting impact in vibration signals according to claim 1, characterized in that: The collecting of the vibration signal of the target rolling mill within a certain time period includes: An acceleration vibration sensor is arranged on the target rolling mill, and the vibration signal of the target rolling mill within a certain time period is collected by the acceleration vibration sensor.

3. The method for identifying steel biting impact in vibration signals according to claim 1, characterized in that: The step of searching for the steel biting impact signal from the collected vibration signal comprises: Differentiating the rolling force of the rolling mill; The moment when the difference value is greater than the threshold is selected as the steel biting moment; The steel biting impact signal is taken as the center of the steel biting impact signal, and the moments at which half the length of the steel biting impact signal is taken on both sides are taken as the boundary points of the steel biting impact signal to screen out the steel biting impact signal.

4. The method for identifying steel biting impact in vibration signals according to claim 1, characterized in that: The screening of a plurality of steel biting impact signals to form a steel biting impact signal database comprises: Intercept several steel biting impact signals and analyze their power spectrum differences; Classifying the steel biting impact signal based on power spectrum difference; The steel-biting impact signal with the largest kurtosis index in each type of steel-biting impact signal is taken as a sample representative and added to the steel-biting impact signal database.

5. The method for identifying steel biting impact in vibration signals according to claim 1, characterized in that: The real-time collection of the vibration signal of the target rolling mill and the cross-correlation operation of the vibration signal with each steel bite impact signal in the steel bite impact signal database through a cross-correlation function include: The vibration signal collected in real time is cross-correlated with the steel biting impact signal in the steel biting impact signal database in turn, and the cross-correlation function calculation formula is: ,in, represents the cross-correlation function sequence between the steel biting impact signal and the actual vibration signal, x represents the steel biting impact signal, y represents the vibration signal collected in real time, m represents the length of the steel biting impact signal, n represents the length of the vibration signal collected in real time, and k represents the length of the cross-correlation function sequence.

6. The method for identifying steel biting impact in vibration signals according to claim 5, characterized in that: Determining the steel biting impact signal in the target rolling mill vibration signal comprises: The absolute value of the cross-correlation function sequence is greater than the judgment threshold The index position is used as the center index of the steel biting impact signal , ; Indexed by the center of the steel impact signal Left and right sides The signal within the point range is determined as the steel biting impact signal , .

7. The method for identifying steel biting impact in vibration signals according to claim 6, characterized in that: The validity of judging the steel biting impact signal includes: Determine the number of steel bite impact signals; If there is one steel-biting impact signal, the validity of the steel-biting impact signal is judged based on the front and back amplitudes of the steel-biting impact signal; if there is more than one steel-biting impact signal, the validity of the steel-biting impact signal is judged based on the front and back amplitudes of the steel-biting impact signal and the time difference between the centers of adjacent steel-biting impact signals.

8. The method for identifying steel biting impact in vibration signals according to claim 7, characterized in that: If there is one steel biting impact signal, judging the validity of the steel biting impact signal based on the amplitudes before and after the steel biting impact signal includes: Calculate the left boundary of the steel biting impact signal to The effective value of the point data ; Calculate the right boundary of the steel biting impact signal to The effective value of the point data ; like , then the steel biting impact signal is valid. , then the steel-biting impact signal is invalid, where j represents the value index number from the boundary of the steel-biting impact signal to both sides, which is used to judge whether the two sides of the steel-biting impact signal are no-load signals and rolling signals, and k is the minimum proportional coefficient of the no-load signal amplitude to the rolling signal amplitude.

9. The method for identifying steel bite impact in vibration signals according to claim 7, characterized in that: If the steel biting impact signal is greater than one, judging the validity of the steel biting impact signal based on the front and rear amplitudes of the steel biting impact signal and the time difference between the centers of adjacent steel biting impact signals includes: Calculate the left boundary of the steel biting impact signal to The effective value of the point data ; Calculate the right boundary of the steel biting impact signal to The effective value of the point data ; Calculate the time series difference corresponding to the signal center of adjacent steel biting impact signals , and determine whether it is greater than the minimum time interval ; like and , then all steel-biting impact signals are valid; like and , then all steel biting impact signals are invalid; like And only one steel bite impact signal satisfies , then the steel biting impact signal is valid. If there are multiple steel biting impact signals that meet , then the entire section of actual mill signal is filtered out; like and , then all steel biting impact signals are invalid.

10. The method for identifying steel biting impact in vibration signals according to claim 8 or 9, characterized in that: If the number of value indexes on both sides of the steel biting impact signal boundary is less than j, all values ​​on that side are taken.

Images