Search

CN-121988620-A - Rolling mill strip steel vibration pattern detection method based on vibration recorder data acquisition

CN121988620ACN 121988620 ACN121988620 ACN 121988620ACN-121988620-A

Abstract

The invention relates to a rolling mill strip steel vibration pattern detection method based on vibration recorder data acquisition, and belongs to the technical field of steel rolling equipment detection. The method comprises the steps of obtaining vibration data, process data and visual image data of a mill housing and a working roller, recording and collecting time stamps, integrating the time stamps into an original data set, establishing a time stamp alignment association rule, adjusting a vibration data filtering strategy according to working conditions of the mill, removing interference data by adopting a dynamic interference judgment rule, integrating the interference data to form a standardized detection data set, classifying and extracting vibration pattern related characteristics, associating multidimensional characteristics through an attention mechanism, presetting a detection judgment threshold value by combining strip steel attributes, presetting a vibration pattern and fault cause similarity matching index, calculating characteristic matching degree, triggering fault cause judgment, and outputting a corresponding analysis adjustment scheme. The vibration grain accurate detection and fault tracing integration is realized, the rejection rate of the strip steel and the equipment operation and maintenance cost are reduced, and the steel rolling production continuity and the strip steel surface quality are ensured.

Inventors

  • LI SHUXIAO

Assignees

  • 上海朗尚传感技术有限公司

Dates

Publication Date
20260508
Application Date
20251223

Claims (12)

  1. 1. A rolling mill strip steel vibration pattern detection method based on vibration recorder data acquisition is characterized by comprising the following steps: s1, respectively obtaining vibration data of a mill housing and a working roll through a vibration recorder, synchronously obtaining process data and visual image data in the rolling process, recording acquisition time stamps of the vibration data, the process data and the visual image data, integrating the three types of data into an original data set, setting an automatic triggering association rule for aligning the data time stamps; S2, adjusting a corresponding vibration data filtering strategy according to the current working condition of the rolling mill, presetting interference characteristic thresholds under different working conditions by adopting a dynamic interference judging rule and identifying interference data, eliminating the interference data deviating from an effective vibration signal by time-frequency analysis, integrating the processed vibration data with corresponding process data according to the dimension of vibration parameters, and forming a standardized detection data set; S3, based on the standardized detection data set, classifying and extracting relevant characteristics of strip steel vibration patterns according to vibration time domain characteristics, process time sequence characteristics and visual space texture characteristics, and carrying out correlation processing on the multidimensional characteristics through an attention mechanism; S4, presetting similarity matching indexes of vibration patterns and fault causes, presetting a multidimensional feature model, calculating matching degrees according to preset weights, triggering corresponding fault cause judgment according to the matching degrees, and finally outputting corresponding analysis adjustment schemes.
  2. 2. The method of claim 1, wherein the specific process of respectively obtaining the vibration data of the mill housing and the working roll through the vibration recorder is that the vibration recorder respectively collects vibration signals of the mill housing in the transverse direction, the longitudinal direction, the vertical direction and the working roll in the radial direction and the axial direction, the sampling frequency of the vibration recorder is preset, and the vibration data of the mill housing and the working roll are received and stored in real time.
  3. 3. The method of claim 1, wherein the specific process of synchronously acquiring the process data and the visual image data in the rolling process comprises the steps of establishing a data transmission link, presetting a data acquisition interval, reading rolling speed, strip steel tension value, roller pressure value and roller gap value in real time, temporarily storing the read parameters in a data buffer area according to the acquisition time sequence, simultaneously acquiring and storing continuous visual images of the strip steel surface in real time through a linear array camera, and synchronizing the acquisition time of the process data and the visual image data.
  4. 4. The method according to claim 1, wherein the specific process of recording the acquisition time stamps of the vibration data, the process data and the visual image data is to configure a time synchronization function for the acquisition equipment of the vibration data, the process data and the visual image data, control the acquisition time errors of the three types of data within a preset time range, synchronously generate corresponding time stamps every time the data are acquired, and import the three types of data with the time stamps into the same database table in time sequence.
  5. 5. The method of claim 1, wherein the specific process of setting the automatic triggering association rule for data timestamp alignment is to preset a timestamp contrast threshold, monitor a timestamp difference value of any data, and interpolate or intercept the process data and the visual image data based on the vibration data timestamp.
  6. 6. The method of claim 1, wherein the specific process of adjusting the corresponding vibration data filtering strategy according to the current working condition of the rolling mill comprises the steps of judging the current working condition by analyzing the rolling speed and the rolling rotating speed in the process data, calling the corresponding vibration data filtering modes aiming at different working conditions, pre-storing the corresponding filtering parameters of the different working conditions, and automatically calling the corresponding parameters when the working conditions are switched.
  7. 7. The method according to claim 1, wherein the specific process of presetting the interference characteristic threshold values under different working conditions and identifying the interference data by adopting a dynamic interference judging rule is to preset a motor fundamental frequency interference threshold value for a rolling working condition, identify the interference data when the amplitude of the Hertz frequency component in the vibration signal exceeds the motor fundamental frequency interference threshold value, and identify the interference data when the amplitude exceeds the impact interference preset threshold value and the duration accords with a preset range for a roll changing working condition.
  8. 8. The method of claim 1, wherein the specific process of eliminating interference data deviating from the effective vibration signal through time-frequency analysis comprises the steps of generating a time-frequency diagram by short-time Fourier transform on the identified interference data, determining the distribution range of the interference signal in a time-frequency domain, cutting the original vibration signal according to the distribution range, removing signal components corresponding to the interference region in the time-frequency diagram, and reserving the effective vibration signal components in a set frequency range.
  9. 9. The method according to claim 1, wherein the specific process of forming the standardized detection data set is to associate the vibration data in each dimension with the process data of the same timestamp by taking the time domain amplitude and the frequency domain peak value of the vibration data as classification dimensions, sort and store the vibration data according to a preset format, and form the standardized detection data set.
  10. 10. The method according to claim 1, wherein the specific process of classifying and extracting the related features of the strip steel vibration patterns comprises the steps of calculating amplitude peak values and standard deviations as time domain features of vibration data, calculating rolling speed change rates and tension fluctuation amplitudes as time sequence features of process data, calculating texture gray level differences and stripe periods as space features of visual image data through gray level histogram analysis, and storing the three types of features in different fields of a preset feature database respectively.
  11. 11. The method of claim 1, wherein the specific process of the similarity matching index of the preset vibration pattern and the fault cause comprises the steps of selecting an amplitude fluctuation coefficient in a vibration time domain characteristic, a peak frequency duty ratio in a frequency domain characteristic and a parameter change rate in a process time sequence characteristic as matching dimensions, presetting a weight coefficient for each dimension, calculating a similarity value of the feature to be matched and the fault feature through a weighted sum formula, and presetting calculation accuracy of the similarity value.
  12. 12. The method of claim 1, wherein the specific process of triggering the corresponding fault cause judgment is to compare the calculated matching degree with a characteristic matching degree threshold preset by each fault type one by one, determine the fault type which is matched with the current matching degree and meets the preset matching requirement, and take the fault type as the fault cause corresponding to the current vibration pattern, and record multi-dimensional characteristic data and matching comparison results involved in the judging process.

Description

Rolling mill strip steel vibration pattern detection method based on vibration recorder data acquisition Technical Field The invention belongs to the technical field of steel rolling equipment detection, and particularly relates to a rolling mill strip steel vibration pattern detection method based on vibration recorder data acquisition. Background In steel rolling production in the steel industry, strip steel is a key product widely applied, and the surface quality of the strip steel directly determines the processability and appearance quality of downstream products. The vibration lines are common surface defects in the rolling process of the rolling mill, are expressed as periodic transverse stripes of the surface of the strip steel along the rolling direction, can not only cause the surface roughness of the strip steel to be increased and the adhesive force of the coating to be reduced, but also cause the rejection of the whole strip steel in the subsequent processing and cause significant economic loss to enterprises when serious, and meanwhile, the generation of the vibration lines is often accompanied with abnormal abrasion of core parts of the rolling mill, thereby shortening the service lives of parts such as rollers, bearings and the like and increasing the operation and maintenance cost. The existing strip steel vibration pattern detection technology of the rolling mill is mainly divided into three types, and has obvious limitations, so that the requirements of modern steel rolling production on high-precision and high-efficiency monitoring are difficult to meet: the first category is traditional manual detection techniques. The technology relies on operators to sample and inspect the strip steel at the outlet side of the rolling mill through naked eyes or a simple optical instrument, and has the core defects that the strip steel is fast in moving speed, tiny vibration lines are difficult to capture in real time by manpower, the problem of missing inspection is remarkable, the detection result is greatly influenced by personnel experience and fatigue degree, subjectivity is high, a standardized conclusion cannot be formed, the technology belongs to post detection, vibration risks cannot be early warned in advance, the strip steel with defects is difficult to rework, and only degradation or scrapping can be realized. The second category is detection techniques based on machine vision. The technology collects the surface image of the strip steel through the visual equipment at the outlet of the rolling mill, recognizes the vibration patterns by means of an image processing algorithm, achieves automation, but still has a short plate, is greatly influenced by dust, water mist and oil stains on the site of the rolling mill, is easy to cause image blurring and obvious in misjudgment, can only recognize the formed surface patterns, cannot be related to the vibration sources of rolling mill parts and the like, greatly reduces the detection precision when the vibration patterns are not obvious due to the coverage of iron oxide scale and rolling oil on the surface of the strip steel, has strict requirements on the installation position of the visual equipment, is difficult to find a proper installation space in part of old rolling mill production lines, and is high in transformation difficulty and cost. The third category is detection technology based on vibration signals, which is a current industry research hotspot. The technology is characterized in that vibration sensors are arranged on parts such as a rolling mill stand and a roller system to collect vibration data of equipment, characteristics are extracted by means of a signal processing algorithm to judge abnormal vibration, and the problem that the vibration abnormality can be captured in advance to realize early warning is solved, but key problems still exist that the vibration data collection dimension is single, only single part data of a rolling mill housing or a working roller are collected, the difference of vibration characteristics of the vibration data collection dimension and the vibration data is not considered, vibration sources are difficult to distinguish, the detection precision is insufficient, a filtering strategy is not matched with the working condition of the rolling mill, the prior art adopts a fixed filtering algorithm, parameters cannot be dynamically adjusted according to the working condition, the condition that effective signals are removed or interference signal residues are easy to occur, the feature extraction accuracy is influenced, only whether vibration lines exist can be judged, the fault traceability is lacking, specific causes cannot be positioned after the abnormality is detected, operation and maintenance personnel need to stop to check one by one, the production continuity is influenced, and defect expansion is possibly caused. In addition, in the prior art, a correlation mechanism of vibration data, process d