Rolling mill vibration ripple on-line monitoring system based on non-linear dynamics

Abstract

The invention provides a rolling mill vibration ripple on-line monitoring system based on nonlinear dynamics, the system can synchronously acquire rolling speed, rolling force and roll gap displacement parameters through a vibration sensor array and a process parameter acquisition unit, a nonlinear characteristic analysis module maps vibration signals and process parameters to a multi-dimensional phase space, and the nonlinear characteristic analysis module analyzes the vibration signals and the process parameters. The method comprises the following steps of: calculating parameters such as a recursion rate and a certainty coefficient by a recursion quantitative analysis unit based on a phase space trajectory to identify whether a system is in a stable state, and performing similarity matching with a vibration mode feature library by a dynamic mode matching unit through an improved distance algorithm to obtain a vibration mode feature library; according to the real-time monitoring and accurate diagnosis system, the current vibration mode is accurately recognized, the dynamic threshold value adjusting module dynamically adjusts the alarm threshold value according to the similarity matching result and the real-time working condition change, it is ensured that alarm is accurately triggered, and the fault early warning capacity and the production stability of the rolling mill are greatly improved through real-time monitoring and accurate diagnosis.

Description

Technical Field

[0001] The present invention relates to an on-line monitoring system for rolling mill vibration patterns based on nonlinear dynamics. Background Art

[0002] In modern rolling steel processes, the vibration state of rolling mill equipment directly affects production efficiency, product quality, and equipment safety. Therefore, real-time monitoring of the rolling mill vibration state and timely identification of equipment abnormalities or failures are crucial for improving the reliability and stability of the production process. However, traditional rolling mill vibration monitoring technologies often rely on simple frequency analysis and direct vibration signal acquisition, making it difficult to effectively handle complex nonlinear dynamic behaviors.

[0003] With the development of the intelligence and automation of industrial production equipment, traditional vibration monitoring methods have gradually been unable to meet the requirements of high precision and high efficiency. Especially in complex systems such as rolling mills, the equipment operation process is affected by various factors, generating nonlinear and time-varying vibration patterns. These vibration signals often contain rich nonlinear characteristics, which can provide important early warning information when the equipment state changes. Therefore, simple vibration signal monitoring can no longer comprehensively and accurately reflect the true working state of the equipment. Moreover, in the existing technology, especially in an environment where dynamic working conditions change rapidly, most methods still have problems such as insufficiently refined vibration data analysis, many false alarms and missed alarms.

[0004] Therefore, how to improve the response speed and accuracy of the monitoring system through precise pattern matching and dynamic threshold adjustment is a problem that needs to be solved by those skilled in the art. Summary of the Invention

[0005] The purpose of the present invention is to solve the above-mentioned deficiencies of the prior art, and provide an on-line monitoring system for rolling mill vibration patterns based on nonlinear dynamics. By integrating multi-source signal acquisition, nonlinear feature analysis, recurrence quantification analysis, dynamic pattern matching, and dynamic threshold adjustment technologies, the detection accuracy and dynamic response ability of rolling mill vibration patterns are improved. The purpose of the present invention is achieved as follows:

[0006] The present invention provides an on-line monitoring system for rolling mill vibration patterns based on nonlinear dynamics, which includes a multi-source signal acquisition module, a nonlinear feature analysis module, a recurrence quantification analysis unit, a dynamic pattern matching unit, and a dynamic threshold adjustment module. The multi-source signal acquisition module includes a vibration sensor array and a process parameter acquisition unit. The process parameter acquisition unit collects rolling speed, rolling force, and roll gap displacement parameters in real time through a wired connection method, and synchronously samples the output signals of the vibration sensor array; the nonlinear feature analysis module includes a phase space reconstruction unit. The phase space reconstruction unit maps the vibration signal and the process parameter sequence to a multi-dimensional phase space through a time-delay embedding algorithm to generate a multi-dimensional phase space trajectory that couples the equipment state and process parameters; the recurrence quantification analysis unit extracts the nonlinear attractor structure features based on the multi-dimensional phase space trajectory; the dynamic pattern matching unit is used to perform similarity matching with a pre-stored vibration pattern feature library according to the nonlinear attractor structure features through an improved distance algorithm; the dynamic threshold adjustment module dynamically adjusts the alarm trigger threshold based on the result of the similarity matching through a PID control algorithm and the real-time working condition change rate.

[0007] Furthermore, the process parameter acquisition unit includes: a rolling speed sensor for real-time monitoring and acquisition of the rolling speed of the rolling mill; a rolling force sensor for real-time monitoring and acquisition of the rolling force during the rolling process; a roll gap displacement sensor for real-time monitoring and acquisition of the displacement signal of the roll gap; the process parameter acquisition unit synchronously samples the signals collected by each sensor with the output vibration signal of the vibration sensor array through a wired connection method to form a time-synchronized composite data stream.

[0008] Furthermore, the phase space reconstruction unit includes a time-delay embedding module for generating a time-delay sequence according to the vibration signal and the process parameter sequence, and selecting an appropriate embedding dimension according to the time-delay sequence; a mapping module for mapping the time-delay sequence to a multi-dimensional phase space, generating a multi-dimensional phase space trajectory including equipment state and process characteristics by taking the vibration signal and process parameter data as inputs, and generating a nonlinear coupling relationship between the equipment state and process parameters according to the multi-dimensional phase space trajectory; a trajectory extraction module for extracting feature information from the generated multi-dimensional phase space trajectory, and the feature information includes the distribution pattern, point cloud structure, and periodicity of the trajectory.

[0009] Furthermore, the recursive quantitative analysis unit includes: a recurrence rate calculation module for calculating the recurrence rate based on the multi-dimensional phase space trajectory; a determinism coefficient calculation module for calculating the determinism coefficient according to the phase space trajectory, which is used to measure whether the dynamic behavior of the system is deterministic, and evaluate whether the system presents regularity and non-randomness by analyzing the density of the attractor structure and the stability of the trajectory; a laminarity parameter calculation module for calculating the laminarity parameter to determine whether the system is in a stable state or in a laminar attractor structure; and a non-linear attractor feature extraction module for extracting non-linear attractor features according to the calculation results of the recurrence rate, the determinism coefficient and the laminarity parameter.

[0010] Furthermore, the dynamic pattern matching unit includes: a feature optimization module for optimizing the feature space of the non-linear attractor structure features and the vibration pattern features, reducing the data complexity and improving the matching efficiency; a weighted matching module for dynamically adjusting the weights of the features in the matching process according to the features of the vibration pattern to improve the matching accuracy; a multi-scale matching module for matching the features at different time and frequency scales to adapt to different pattern changes; and a matching evaluation module for outputting a comprehensive matching degree according to the similarity calculation result to judge the matching degree between the current state and the historical pattern.

[0011] Furthermore, the dynamic threshold adjustment module includes a real-time working condition acquisition module for acquiring working condition data in real time; a working condition change rate calculation module for calculating the change rate of the real-time working condition according to the working condition data; a PID control algorithm module for adjusting the alarm threshold according to the change rate to ensure accurate triggering of the alarm when the working condition changes; and an alarm optimization module for changing the alarm strategy according to the alarm threshold.

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention synchronously acquires the vibration signals and process parameter data of the rolling mill in real time through the multi-source signal acquisition module, and combines the phase space reconstruction technology in the non-linear feature analysis module to map these signals into a multi-dimensional phase space. The non-linear attractor structure features are extracted through the recursive quantitative analysis unit, and the subtle changes in the vibration mode of the rolling mill are accurately captured, enabling the present invention to accurately monitor the dynamic behavior of the rolling mill, thereby significantly improving the accuracy of fault detection and vibration mode recognition and reducing the possibility of misdiagnosis and missed diagnosis.

[0013] The present invention adopts real-time synchronous sampling technology, simultaneously obtains vibration signals and process parameter data through a multi-source signal acquisition module, and ensures the real-time transmission of each signal through a wiring connection method, reducing data transmission delay. The dynamic threshold adjustment module combines a PID control algorithm to automatically adjust the alarm trigger threshold according to the real-time working condition change rate, ensuring rapid response to abnormal situations under any working conditions, greatly improving the response speed of the monitoring system, and ensuring timely early warning and effective fault prevention. Brief Description of the Drawings

[0014] Figure 1 It is a schematic structural diagram of an on-line monitoring system for rolling mill vibration streaks based on nonlinear dynamics. Detailed Embodiment

[0015] To deepen the understanding of the present invention, the present invention will be further described in detail below in conjunction with embodiments and drawings. The embodiments are only used to explain the present invention and do not constitute a limitation on the protection scope of the present invention.

[0016] Please refer to Figure 1 , an embodiment of the present invention provides an on-line monitoring system for rolling mill vibration streaks based on nonlinear dynamics, including a multi-source signal acquisition module, a nonlinear feature analysis module, a recurrence quantification analysis unit, a dynamic pattern matching unit, and a dynamic threshold adjustment module. The multi-source signal acquisition module includes a vibration sensor array and a process parameter acquisition unit. The process parameter acquisition unit real-time collects rolling speed, rolling force, and roll gap displacement parameters through a wiring connection method and performs synchronous sampling with the output signal of the vibration sensor array. The nonlinear feature analysis module includes a phase space reconstruction unit. The phase space reconstruction unit maps the vibration signal and the process parameter sequence to a multi-dimensional phase space through a time delay embedding algorithm to generate a multi-dimensional phase space trajectory coupling the equipment state and process parameters. The recurrence quantification analysis unit extracts the characteristics of the nonlinear attractor structure based on the multi-dimensional phase space trajectory. The dynamic pattern matching unit is used to perform similarity matching with a pre-stored vibration pattern feature library according to the characteristics of the nonlinear attractor structure through an improved distance algorithm. The dynamic threshold adjustment module dynamically adjusts the alarm trigger threshold based on the results of the similarity matching through a PID control algorithm and the real-time working condition change rate.

[0017] In one embodiment, the process parameter acquisition unit includes: a rolling speed sensor for real-time monitoring and acquisition of the rolling speed of the rolling mill; a rolling force sensor for real-time monitoring and acquisition of the rolling force during the rolling process; a roll gap displacement sensor for real-time monitoring and acquisition of the displacement signal of the roll gap; the process parameter acquisition unit synchronously samples the signals collected by each sensor and the output vibration signal of the vibration sensor array through a wiring method to form a time-synchronized composite data stream; the phase space reconstruction unit includes a time delay embedding module for generating a delayed time series based on the vibration signal and the process parameter sequence and selecting an appropriate embedding dimension according to the delayed time series; a mapping module for mapping the delayed time series to a multi-dimensional phase space, generating a multi-dimensional phase space trajectory containing the device state and process characteristics by taking the vibration signal and the process parameter data as inputs, and generating a non-linear coupling relationship between the device state and the process parameters according to the multi-dimensional phase space trajectory; a trajectory extraction module for extracting feature information from the generated multi-dimensional phase space trajectory, and the feature information includes the distribution pattern, point cloud structure and periodicity of the trajectory; the recurrence quantification analysis unit includes: a recurrence rate calculation module for calculating the recurrence rate based on the multi-dimensional phase space trajectory; a determinism coefficient calculation module for calculating the determinism coefficient according to the phase space trajectory, and this coefficient is used to measure whether the dynamic behavior of the system is deterministic, and evaluate whether the system presents regularity and non-randomness by analyzing the density of the attractor structure and the stability of the trajectory; a laminarity parameter calculation module for calculating the laminarity parameter to judge whether the system is in a stable state or in a laminar attractor structure; a non-linear attractor feature extraction module for extracting non-linear attractor features according to the calculation results of the recurrence rate, the determinism coefficient and the laminarity parameter; the dynamic pattern matching unit includes: a feature optimization module for optimizing the feature space of the non-linear attractor structure features and the vibration pattern features, reducing the data complexity and improving the matching efficiency; a weighted matching module for dynamically adjusting the weights of each feature in the matching process according to the features of the vibration pattern to improve the matching accuracy; a multi-scale matching module for matching the features at different time and frequency scales to adapt to different pattern changes; a matching evaluation module for outputting a comprehensive matching degree according to the similarity calculation result and judging the matching degree between the current state and the historical pattern; the dynamic threshold adjustment module includes a real-time working condition acquisition module for real-time acquisition of working condition data; a working condition change rate calculation module for calculating the change rate of the real-time working condition according to the working condition data; a PID control algorithm module for adjusting the alarm threshold according to the change rate to ensure accurate triggering of the alarm when the working condition changes; an alarm optimization module for changing the alarm strategy according to the alarm threshold.

[0018] It should be noted that the vibration sensor array consists of 16 high-precision vibration sensors, which are distributed at various key parts of the rolling mill to collect the vibration signals of the rolling mill. The sampling frequency of each vibration sensor is 10 kHz to capture the subtle vibration fluctuations during the operation of the rolling mill; the rolling speed sensor is used to monitor the rolling speed of the rolling mill, with a sampling frequency of 1 Hz and an accuracy of ±0.01 m / s; the rolling force sensor is used to monitor the rolling force of the rolling mill, with a sampling frequency of 1 Hz and an accuracy of ±5 kN; the roll gap displacement sensor is used to monitor the roll gap displacement of the rolling mill, with a sampling frequency of 1 Hz and an accuracy of ±0.01 mm; all sensors synchronously collect vibration signals and process parameter signals through industrial-grade sensor interfaces and wiring methods to form a time-synchronized composite data stream.

[0019] Specifically, the phase space reconstruction unit uses the time-delay embedding algorithm to generate a time-delay time series by selecting the delay time. According to the autocorrelation function of the time series, the embedding dimension is selected as 6 to accurately reflect the non-linear relationship between the vibration signals and process parameters of the rolling mill; the mapping module maps the time-delay time series into a multi-dimensional phase space to obtain an 8-dimensional phase space trajectory. By taking the vibration signals and rolling mill process parameter data as inputs, the generated trajectory shows the non-linear coupling relationship between the equipment state and process parameters; the trajectory extraction module extracts key feature information from the generated multi-dimensional phase space trajectory, such as the distribution pattern, point cloud structure, and periodicity of the trajectory, and uses the principal component analysis algorithm to extract the main components. The extracted feature information provides a basis for subsequent recurrence quantification analysis.

[0020] Furthermore, the recurrence rate calculation module calculates the recurrence rate based on the generated multi-dimensional phase space trajectory. By analyzing the time dependence between trajectory points, the recurrence rate value is 0.75, indicating that the system has a high degree of time dependence and is suitable for further non-linear analysis; the determinism coefficient is calculated according to the trajectory, with a value of 0.85, indicating that the vibration mode of the rolling mill exhibits strong deterministic behavior and the system operates regularly; the laminarity parameter calculation module calculates the laminarity parameter value of 0.92 by analyzing the stability of the multi-dimensional phase space trajectory, indicating that the rolling mill is in a stable state under the current working conditions and the vibration mode shows a laminar attractor; the non-linear attractor feature extraction module extracts non-linear attractor features including the recurrence rate, determinism coefficient, and laminarity parameter based on the calculation results. These features are used to judge the regularity and stability of the rolling mill vibration; the feature optimization module: through spatial optimization of the non-linear attractor structure features and pre-stored vibration mode features, the data dimension is reduced, thereby improving the matching efficiency. The feature dimension is optimized to 5 dimensions to ensure the calculation speed and matching accuracy; the weighted matching module: during the matching process, the weights of the features are dynamically adjusted according to the vibration mode features. By adjusting the weights, the matching accuracy of the vibration mode is improved, and the similarity value with the historical mode feature library is 0.92, indicating that the current rolling mill vibration mode is highly similar to the historical mode.

[0021] Furthermore, the multi-scale matching module matches vibration features at different time and frequency scales, enabling the system to adapt to different pattern changes; it uses wavelet transform and spectral analysis methods to perform feature matching in the high-frequency and low-frequency ranges respectively; the matching evaluation module outputs a comprehensive matching degree based on the similarity calculation result. The comprehensive matching degree is 0.88, indicating that the current rolling mill vibration pattern highly matches the pre-stored vibration pattern, and the system can accurately predict the equipment status; the real-time working condition acquisition module obtains the working state data of the rolling mill in real time through this module, including information such as production load, equipment temperature, and rotational speed; the working condition change rate calculation module: calculates the working condition change rate based on the real-time acquired working condition data. Through the analysis of the change rate, for example, if the working condition change rate is obtained as 0.05 / s, it means that the current working condition change is relatively stable; the PID control algorithm module dynamically adjusts the alarm trigger threshold through the PID control algorithm according to the matching result of the working condition change rate and the historical vibration pattern. By adjusting the PID parameters, the alarm threshold is adjusted to 1.5g according to the real-time change of the working condition, enabling the system to accurately respond to different production conditions; the alarm optimization module, based on the adjustment of the alarm threshold, the system optimizes the alarm strategy according to the adjusted threshold, avoiding premature false alarms and ensuring timely early warning in case of faults.

[0022] In a possible usage scenario, a rolling mill processes metal materials into shapes such as thin plates and thin strips through rolling. However, during operation, the rolling mill will generate certain vibrations. Therefore, timely monitoring and diagnosing abnormal vibrations of the rolling mill is of crucial significance for improving production efficiency, ensuring equipment stability, and reducing production accidents. In a steel rolling production line of a certain steel plant, the factory is using the online monitoring system for rolling mill vibration patterns based on nonlinear dynamics of the present invention. The main equipment of this production line includes multiple high-precision rolling mills, and each rolling mill is equipped with a vibration sensor array and multiple process parameter sensors. The following are the specific steps: Each rolling mill uses the vibration sensor array to monitor the vibration state of the rolling mill in real time, and synchronously collects key process parameters such as rolling speed, rolling force, and roll gap displacement through the process parameter acquisition unit. These data are transmitted to the nonlinear feature analysis module in real time through wiring for subsequent processing. The synchronous acquisition of process parameters ensures the time alignment of all signals and provides accurate input data for subsequent multi-dimensional phase space reconstruction. Step two: The phase space reconstruction unit maps the vibration signal and the process parameter sequence to a multi-dimensional phase space. The system combines the delayed time series of different process parameters and the vibration signal through the time delay embedding algorithm to generate a multi-dimensional phase space trajectory that couples the equipment state and process parameters. The recursive quantitative analysis unit calculates the recurrence rate, determinism coefficient, and laminarity parameter based on the generated multi-dimensional phase space trajectory to extract the characteristics of the nonlinear attractor structure and identify whether the system is in a stable or unstable state. In this way, the system can deeply understand the dynamic behavior and vibration mode of the equipment. Through learning historical data, the dynamic pattern matching unit in the system uses an improved distance algorithm to perform similarity matching between the currently extracted nonlinear attractor features and the pre-stored vibration mode feature library. This matching process can accurately identify whether the current vibration mode matches the historical fault mode. If the system detects that the current vibration mode is similar to a known fault mode, the system will dynamically adjust the alarm trigger threshold within the dynamic threshold adjustment module through the PID control algorithm. In this way, the system can flexibly adjust the alarm threshold according to real-time working conditions, avoiding the blindness and over-alarm problems of the fixed threshold in the traditional system. When the vibration mode highly matches the fault mode, the system will trigger the alarm mechanism and automatically take corresponding measures according to the set alarm strategy. For example, it will notify the operator to check through an alarm or directly start the emergency shutdown procedure. The alarm optimization module will further optimize the alarm strategy based on the dynamic threshold to ensure the accuracy and timeliness of the alarm.

[0023] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. An on-line monitoring system for rolling mill vibration patterns based on nonlinear dynamics, characterized in that, It includes a multi-source signal acquisition module, a non-linear feature analysis module, a recurrence quantification analysis unit, a dynamic pattern matching unit, and a dynamic threshold adjustment module. The multi-source signal acquisition module includes a vibration sensor array and a process parameter acquisition unit. The process parameter acquisition unit collects rolling speed, rolling force, and roll gap displacement parameters in real time through a wired connection method, and synchronously samples with the output signals of the vibration sensor array. The non-linear feature analysis module includes a phase space reconstruction unit. The phase space reconstruction unit maps the vibration signal and the process parameter sequence to a multi-dimensional phase space through a time-delay embedding algorithm to generate a multi-dimensional phase space trajectory coupling the equipment state and the process parameters. The recurrence quantification analysis unit extracts non-linear attractor structure features based on the multi-dimensional phase space trajectory. The dynamic pattern matching unit is used to perform similarity matching with a pre-stored vibration pattern feature library according to the non-linear attractor structure features through an improved distance algorithm. The dynamic threshold adjustment module dynamically adjusts the alarm trigger threshold based on the results of the similarity matching through a PID control algorithm and the real-time working condition change rate.

2. The on-line monitoring system for rolling mill vibration patterns based on nonlinear dynamics according to claim 1, characterized in that, The process parameter acquisition unit includes: a rolling speed sensor for real-time monitoring and acquisition of the rolling speed of the rolling mill; a rolling force sensor for real-time monitoring and acquisition of the rolling force during the rolling process; a roll gap displacement sensor for real-time monitoring and acquisition of the displacement signal of the roll gap. The process parameter acquisition unit synchronously samples the signals collected by each sensor with the output vibration signals of the vibration sensor array through a wired connection method to form a time-synchronized composite data stream.

3. An on-line monitoring system for rolling mill vibration patterns based on nonlinear dynamics according to claim 1, characterized in that, The phase space reconstruction unit includes a time-delay embedding module for generating a time-delay time series according to the vibration signal and the process parameter sequence, and selecting an appropriate embedding dimension according to the time-delay time series. A mapping module for mapping the time-delay time series to a multi-dimensional phase space, generating a multi-dimensional phase space trajectory containing equipment state and process characteristics by taking the vibration signal and process parameter data as inputs, and generating a non-linear coupling relationship between the equipment state and the process parameters according to the multi-dimensional phase space trajectory. A trajectory extraction module for extracting feature information from the generated multi-dimensional phase space trajectory, where the feature information includes the distribution pattern, point cloud structure, and periodicity of the trajectory.

4. An on-line monitoring system for rolling mill vibration patterns based on nonlinear dynamics according to claim 1, characterized in that, The recurrence quantification analysis unit includes: a recurrence rate calculation module for calculating the recurrence rate based on the multi-dimensional phase space trajectory; a determinism coefficient calculation module for calculating the determinism coefficient according to the phase space trajectory, which is used to measure whether the dynamic behavior of the system is deterministic, and evaluating whether the system shows regularity and non-randomness by analyzing the density of the attractor structure and the stability of the trajectory; a laminarity parameter calculation module for calculating the laminarity parameter to judge whether the system is in a stable state or in a laminar attractor structure; a non-linear attractor feature extraction module for extracting non-linear attractor features according to the calculation results of the recurrence rate, the determinism coefficient, and the laminarity parameter.

5. The on-line monitoring system for rolling mill vibration patterns based on non-linear dynamics according to claim 1, wherein The dynamic pattern matching unit includes: a feature optimization module for optimizing the feature space of the non-linear attractor structure features and vibration mode features, reducing the complexity of data, and improving the matching efficiency; a weighted matching module for dynamically adjusting the weights of each feature in the matching process according to the features of the vibration mode to improve the matching accuracy; a multi-scale matching module for matching features at different time and frequency scales so as to adapt to different mode changes; a matching evaluation module for outputting a comprehensive matching degree according to the similarity calculation result and judging the matching degree between the current state and the historical mode.

6. The on-line monitoring system for rolling mill vibration patterns based on non-linear dynamics according to claim 1, characterized in that, The dynamic threshold adjustment module includes a real-time working condition acquisition module for acquiring working condition data in real time; a working condition change rate calculation module for calculating the change rate of the real-time working condition according to the working condition data; a PID control algorithm module for adjusting the alarm threshold according to the change rate to ensure accurate triggering of the alarm when the working condition changes; an alarm optimization module for changing the alarm strategy according to the alarm threshold.

Images