A device condition monitoring and maintenance system

Abstract

The present application relates to the technical field of equipment state monitoring, and especially relates to a kind of equipment state monitoring and maintenance system, the present application proposes the following scheme, by collecting the three-dimensional stress energy density distribution data of roll bearing seat, the extension trend of roll surface microcrack is predicted, dynamic residual life index is generated, and the rolling force compensation is carried out in combination with pipe section ovality data;At the same time, by synchronously collecting Barkhausen noise signal and cooling medium flow data, cooling intensity is adjusted in real time;To ensure the health of roll, based on dynamic residual life index, roll health weight is generated, rolling force compensation model is constructed, rolling force distribution is optimized by particle swarm optimization algorithm, and pipe wall thickness uniformity is ensured;The method effectively prevents roll damage and pipe wall thickness mutation, improves production efficiency and product quality, and prolongs the service life of equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment status monitoring, and in particular to an equipment status monitoring and maintenance system. Background Art

[0002] During the continuous rolling of steel pipes, the dynamic contact between the rolls and the tube billet leads to energy efficiency losses. This is particularly true in high-frequency welded pipe mills, where the coupled failure of hidden roll surface damage and sudden changes in pipe wall thickness during high-speed rolling of thin-walled pipes is becoming increasingly prominent. Traditional rolling force control methods typically rely on simple empirical adjustments or preset rules, which are difficult to handle under complex rolling conditions and cannot monitor and compensate for roll damage and pipe geometry changes in real time. This lack of precise monitoring and adjustment often leads to uneven pipe wall thickness, fluctuating product quality, and easily causes roll damage, shortening equipment life. Furthermore, existing cooling control technologies fail to effectively address the problems of roll surface overheating or uneven cooling, and lack dynamic adjustment, further exacerbating equipment wear and product quality issues during the rolling process. Therefore, accurately predicting and compensating for roll damage to ensure quality and equipment health during the rolling process is crucial to improving steel pipe production efficiency and reducing costs.

[0003] The problem raised by this background technology is that during the continuous rolling of steel pipes, there is an irreversible accumulation problem of energy efficiency loss due to dynamic contact between the rollers and the tube blanks. To solve the above problem, this application designs an equipment status monitoring and maintenance system. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to address the deficiencies of the existing technology and provide an equipment status monitoring and maintenance system. By collecting the three-dimensional stress energy density distribution data of the roller bearing seat, the micro-crack expansion trend on the roller surface is predicted, a dynamic remaining life index is generated, and rolling force compensation is performed in combination with the pipe cross-section ovality data. At the same time, the cooling intensity is adjusted in real time by synchronously collecting the Barkhausen noise signal and the cooling medium flow data. To ensure the health of the roller, the roller health weight is generated based on the dynamic remaining life index, and a rolling force compensation model is constructed. The rolling force distribution is optimized by the particle swarm optimization algorithm to ensure the uniformity of the pipe wall thickness. This method effectively prevents roller damage and sudden changes in pipe wall thickness, improves production efficiency and product quality, and extends the service life of the equipment.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An equipment status monitoring and maintenance system is applied to a steel pipe continuous rolling production line. The system includes a data acquisition module, a data analysis module, and an equipment maintenance module, wherein:

[0007] The data acquisition module is used to collect three-dimensional stress energy density distribution data, Barkhausen noise signal and cooling medium flow data of the roller bearing seat;

[0008] The data analysis module establishes an energy dissipation rate model based on the three-dimensional stress energy density distribution data, predicts the growth trend of microcracks on the roll surface and generates a dynamic remaining life index. At the same time, based on the amplitude and frequency domain characteristics of the Barkhausen noise signal, it establishes a mapping relationship between the local magnetic domain changes of the roll and the abnormal phase change characteristics. It also determines the effectiveness of the current cooling mode based on the cooling medium flow data and identifies the cooling abnormality area.

[0009] The equipment maintenance module compensates for the roll reduction amount according to the output of the data analysis module and adjusts the opening strategy of the cooling nozzle combination.

[0010] The data analysis module includes:

[0011] The crack prediction unit builds an energy dissipation rate model based on three-dimensional stress energy density distribution data, dynamically corrects time-varying parameters through a residual-driven algorithm, and generates a thermal map of the crack initiation location and a time-varying growth rate curve;

[0012] The remaining life analysis unit is used to couple crack prediction results with production condition data, generate the remaining life probability distribution interval through Monte Carlo simulation, and generate a dynamic remaining life index based on order requirements;

[0013] The cooling intensity analysis unit is used to analyze the magnetic domain-phase change mapping relationship of the Barkhausen noise signal and identify abnormal phase change areas and cooling failure areas based on the cooling medium flow distribution characteristics.

[0014] The crack prediction unit is configured with prediction logic, which includes:

[0015] Calculating local energy density distribution based on the three-dimensional stress energy density distribution data;

[0016] Extracting the energy dissipation rate peak and the spatial distribution entropy value based on the local energy density distribution, and establishing a mapping relationship between the energy threshold and the crack initiation position in combination with the fatigue crack growth rate test data of the roll material;

[0017] Comparing the mapping relationship with the predicted relationship of the offline calibrated finite element simulation model, and correcting the time-varying parameters of the energy dissipation rate model through a residual-driven algorithm, wherein the time-varying parameters characterize the crack growth coefficient, exponential factor, and crack growth activation energy of the damage accumulation characteristics of the roll material;

[0018] A crack growth trend prediction is generated based on the modified energy dissipation rate model, wherein the crack growth trend prediction includes a probability distribution heat map of the crack initiation position and a time-varying growth rate curve of the crack length.

[0019] The method of correcting the time-varying parameters of the energy dissipation rate model by using a residual driving algorithm includes:

[0020] Perform frequency domain decomposition on the residual signal between the mapping relationship and the prediction relationship to extract the energy dominant mode of the characteristic frequency band;

[0021] performing sensitivity analysis on the time-varying parameters in the energy dissipation rate model according to the energy dominant mode, calculating the sensitivity weights of the time-varying parameters through the constitutive equation of the roll material, and generating a priority sequence;

[0022] Performing a staged Bayesian correction on the time-varying parameters according to the priority sequence, wherein the staged Bayesian correction includes material constraints, energy constraints, and set constraints;

[0023] The time-varying parameters are iteratively modified until a predetermined number of iterations is met.

[0024] The remaining life analysis unit is configured with life analysis logic, and the life analysis logic includes:

[0025] Couple the probability distribution heat map and the time-varying expansion rate curve to extract the credibility weight of the crack initiation position and the probability density function of the expansion rate;

[0026] Generating a probability distribution interval of the remaining life through Monte Carlo simulation according to the credibility weight and the probability density function;

[0027] The probability distribution interval is matched with the order rolling mileage requirements in the production plan database, the remaining life confidence level that meets the current production task requirements is calculated, and a dynamic remaining life index is generated.

[0028] The cooling intensity analysis unit is configured with cooling identification logic, which includes:

[0029] The amplitude and frequency domain characteristics of the Barkhausen noise signal are used to identify the local magnetic domain changes of the roller, and a mapping relationship is established between the abnormal phase change characteristics of the roller surface.

[0030] When the abnormal phase change characteristics indicated by the mapping relationship exceed the process allowable range, the position and extent of the abnormal roll area are determined by extracting the multimodal features of the Barkhausen noise signal.

[0031] The cooling identification logic further includes:

[0032] determining the effectiveness of the current cooling mode according to the cooling medium flow data;

[0033] When the cooling medium flow rate data locally deviates from the set value, the local change in the cooling medium flow rate is correlated to determine the spatial region with uneven cooling distribution.

[0034] The equipment maintenance module includes:

[0035] A rolling force compensation unit, configured to receive ovality detection data and a dynamic remaining life index and generate a rolling force compensation parameter set;

[0036] The cooling intensity adjustment unit adjusts the nozzle opening and closing combination and flow distribution according to the abnormal phase change area and cooling failure area.

[0037] The rolling force compensation unit is provided with compensation logic, which includes:

[0038] The ovality data of the pipe section is divided into N detection areas according to the circumferential angle, the wall thickness deviation rate of each area is calculated, and the ovality feature matrix is ​​generated;

[0039] converting the dynamic remaining life index into a roll health weight;

[0040] A rolling force compensation model based on the health status of the equipment is constructed, the ellipticity characteristic matrix and the roll health weight are used as inputs of the rolling force compensation model, optimization is performed through the rolling force compensation model, and a rolling force compensation parameter set is calculated.

[0041] Optimizing the rolling force compensation model to calculate a rolling force compensation parameter set includes:

[0042] generating an initial compensated pressure distribution using the ellipticity characteristic matrix as input;

[0043] Using the health weight of the roll as a health weight constraint condition;

[0044] The Pareto optimal solution with the minimum total rolling force increment is calculated by a particle swarm optimization algorithm, and the Pareto optimal solution is used as a rolling force compensation parameter set.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] This method utilizes a variety of information sources, including three-dimensional stress energy density distribution data, dynamic remaining life index, tube ovality data, and Barkhausen noise signals, to accurately predict roll health and crack growth trends. Based on this information, rolling force and cooling intensity are dynamically adjusted. The particle swarm optimization algorithm optimizes the rolling force distribution, ensuring uniformity of tube wall thickness, reducing roll damage and tube quality defects. This allows for precise prediction and compensation of roll damage, ensuring quality and equipment health during the rolling process, and improving steel pipe production efficiency and reducing costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0048] Figure 1 This is a module diagram of an equipment status monitoring and maintenance system according to embodiment 1 of the present invention;

[0049] Figure 2 This is a flow chart of a device status monitoring and maintenance method according to embodiment 2 of the present invention. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0051] Example 1

[0052] See also Figure 1 The present invention provides an embodiment of an equipment status monitoring and maintenance system, which is applied to a steel pipe continuous rolling production line. The system includes a data acquisition module, a data analysis module, and an equipment maintenance module, wherein:

[0053] The data acquisition module is used to collect three-dimensional stress energy density distribution data, Barkhausen noise signal and cooling medium flow data of the roller bearing seat;

[0054] The data analysis module establishes an energy dissipation rate model based on the three-dimensional stress energy density distribution data, predicts the growth trend of microcracks on the roll surface and generates a dynamic remaining life index. At the same time, based on the amplitude and frequency domain characteristics of the Barkhausen noise signal, it establishes a mapping relationship between the local magnetic domain changes of the roll and the abnormal phase change characteristics. It also determines the effectiveness of the current cooling mode based on the cooling medium flow data and identifies the cooling abnormality area.

[0055] The equipment maintenance module compensates for the roll reduction amount according to the output of the data analysis module and adjusts the opening strategy of the cooling nozzle combination.

[0056] The data analysis module comprises:

[0057] A crack prediction unit constructs an energy dissipation rate model based on three-dimensional stress energy density distribution data, dynamically corrects time-varying parameters through a residual driving algorithm, and generates a crack initiation location thermal map and a time-varying expansion rate curve;

[0058] A residual life analysis unit is used to couple the crack prediction results and production condition data, generate a residual life probability distribution interval through Monte Carlo simulation, and match the order demand to generate a dynamic residual life index;

[0059] A cooling intensity analysis unit is used to analyze the magnetic domain-phase transition mapping relationship of the Barkhausen noise signal, identify abnormal phase transition regions and cooling failure regions in combination with the cooling medium flow distribution characteristics.

[0060] The crack prediction unit is configured with a prediction logic, and the prediction logic comprises:

[0061] The local energy density distribution is calculated according to the three-dimensional stress energy density distribution data;

[0062] The energy dissipation rate peak value and the spatial distribution entropy value are extracted according to the local energy density distribution, and the mapping relationship between the energy threshold value and the crack initiation location is established in combination with the fatigue crack propagation rate test data of the roll material;

[0063] The mapping relationship is compared with the prediction relationship of the finite element simulation model calibrated offline, and the time-varying parameters of the energy dissipation rate model are corrected through a residual driving algorithm, wherein the time-varying parameters represent the crack propagation coefficient, the exponential factor and the crack propagation activation energy of the roll material damage accumulation characteristic;

[0064] A crack propagation trend prediction is generated according to the corrected energy dissipation rate model, wherein the crack propagation trend prediction includes a probability distribution thermal map of the crack initiation location and a time-varying expansion rate curve of the crack length.

[0065] The time-varying parameters of the energy dissipation rate model are corrected through a residual driving algorithm, comprising:

[0066] The residual signals between the mapping relationship and the prediction relationship are frequency domain decomposed to extract the energy dominant mode of the characteristic frequency band;

[0067] The time-varying parameters in the energy dissipation rate model are analyzed for sensitivity according to the energy dominant mode, the sensitivity weight of the time-varying parameters is calculated through the roll material constitutive equation, and a priority sequence is generated;

[0068] The time-varying parameters are corrected in stages through a Bayesian method according to the priority sequence, wherein the Bayesian correction in stages includes material constraints, energy constraints and set constraints.

[0069] iteratively correcting the time-varying parameters until a number of iterations is satisfied.

[0070] The remaining life analysis unit is configured with a life analysis logic, which includes:

[0071] The probability distribution thermogram and the time-varying propagation rate curve are coupled for analysis to extract a credibility weight of a crack initiation position and a probability density function of a propagation rate;

[0072] According to the credibility weight and the probability density function, a probability distribution interval of the remaining life is generated through Monte Carlo simulation;

[0073] The probability distribution interval is matched with an order rolling mileage requirement in a production plan database to calculate a remaining life confidence satisfying a current production task requirement and generate a dynamic remaining life index.

[0074] The cooling intensity analysis unit is configured with a cooling identification logic, which includes:

[0075] The local magnetic domain variation of the roll is identified through amplitude and frequency domain characteristics of the Barkhausen noise signal to establish a mapping relationship with abnormal phase change characteristics of the roll surface;

[0076] When the abnormal phase change characteristics indicated by the mapping relationship exceed a process allowable range, the position and degree of the abnormal roll region are determined through multi-modal feature extraction of the Barkhausen noise signal.

[0077] The cooling identification logic further includes:

[0078] The effectiveness of the current cooling mode is determined according to the cooling medium flow data;

[0079] When the cooling medium flow data locally deviates from the set value, the local variation of the cooling medium flow is associated to determine a spatial region of unevenly distributed cooling.

[0080] The equipment maintenance module includes:

[0081] The rolling force compensation unit receives the ovality detection data and the dynamic remaining life index to generate a rolling force compensation parameter set;

[0082] The cooling intensity adjustment unit adjusts the nozzle opening and closing combination and flow distribution according to the abnormal phase change region and the cooling failure region.

[0083] The rolling force compensation unit is configured with a compensation logic, which includes:

[0084] The ovality data of the pipe section is divided into N detection areas according to the circumferential angle, the wall thickness deviation rate of each area is calculated, and the ovality feature matrix is ​​generated;

[0085] converting the dynamic remaining life index into a roll health weight;

[0086] A rolling force compensation model based on the health status of the equipment is constructed, the ellipticity characteristic matrix and the roll health weight are used as inputs of the rolling force compensation model, optimization is performed through the rolling force compensation model, and a rolling force compensation parameter set is calculated.

[0087] Optimizing the rolling force compensation model to calculate a rolling force compensation parameter set includes:

[0088] generating an initial compensated pressure distribution using the ellipticity characteristic matrix as input;

[0089] Using the health weight of the roll as a health weight constraint condition;

[0090] The Pareto optimal solution with the minimum total rolling force increment is calculated by a particle swarm optimization algorithm, and the Pareto optimal solution is used as a rolling force compensation parameter set.

[0091] Example 2

[0092] See also Figure 2 The present invention provides an embodiment: a method for monitoring and maintaining equipment status, which is applied to a steel pipe continuous rolling production line. The specific steps of the method are as follows:

[0093] S1: Collect the three-dimensional stress energy density distribution data of the roller bearing seat and predict the micro-crack growth trend of the roller surface based on the energy dissipation rate model;

[0094] In this embodiment, a deformation energy sensor array is used to collect real-time three-dimensional stress energy density distribution data on the roller bearing seat. Using an energy dissipation rate model and combining it with the roller's stress concentration areas to predict microcrack growth trends, the potential crack initiation locations on the roller surface can be accurately identified. Real-time monitoring of energy loss provides early warning of early fatigue cracks in the roller, enabling cracks to be addressed before they reach a certain size, thereby extending equipment life and reducing production interruptions caused by roller damage. This solves the problem of failure to promptly identify roller surface damage during traditional rolling processes and improves equipment maintenance efficiency.

[0095] S2: Generate dynamic remaining life index;

[0096] In this embodiment, a dynamic remaining life index (RMLI) is generated based on the crack growth trend data predicted in S1 and the actual operating conditions of the rolls. This index not only reflects the damage state of the rolls but also incorporates the properties of the roll material to accurately estimate their remaining service life under the current load. This allows real-time monitoring of the health of the rolls, providing timely data support for maintenance and avoiding emergency shutdowns and unnecessary repairs caused by unclear equipment status. This provides predictable equipment maintenance cycles for steel pipe production lines, effectively reducing unplanned downtime and improving production continuity and stability.

[0097] S3: When the dynamic remaining life index is lower than the preset threshold, rolling force compensation is performed based on the tube cross-section ovality data;

[0098] In this embodiment, real-time monitoring of pipe cross-sectional ovality data identifies uneven wall thickness that may occur during the rolling process. When the ovality deviation exceeds the process tolerance, the hydraulic AGC system automatically activates to dynamically adjust the rolling force to ensure uniform force distribution. This effectively reduces sudden changes in pipe wall thickness caused by roll damage and avoids structural instability caused by uneven pipe wall thickness. This improves product quality during production, reduces production costs, and ensures the reliability of the high-frequency welded pipe unit during high-speed rolling of thin-walled pipes.

[0099] S4: Synchronously collects Barkhausen noise signals and cooling medium flow data, and adjusts the cooling intensity based on a real-time feedback loop;

[0100] In this embodiment, the amplitude and frequency domain characteristics of the Barkhausen noise signal are used to identify areas of abnormal phase change on the roll surface. This is then combined with coolant flow rate data to determine the effectiveness of the current cooling mode. For identified areas of uneven cooling, the coolant flow rate and injection angle are adjusted in real time to optimize the cooling intensity distribution, ensuring uniform cooling of the roll surface and avoiding material phase change issues caused by uneven cooling. This effectively suppresses material deformation on the roll surface caused by overheating or uneven cooling, improving rolling quality and preventing thermal cracking caused by insufficient local cooling.

[0101] During the continuous rolling of steel pipes, especially when thin-walled pipes are rolled at high speeds, conventional technologies still face challenges in addressing the energy efficiency loss caused by dynamic contact between the rolls and the tube blanks. In high-frequency pipe welding units in particular, the coupled failure of hidden roll surface damage and sudden changes in pipe wall thickness not only affects product quality but also increases equipment maintenance costs. Existing technologies primarily address this issue through experience or simple feedback adjustments, but they do not comprehensively analyze and optimize complex factors such as the propagation of microcracks on the roll surface and energy efficiency losses. Therefore, the key to solving this problem is how to reduce roll damage and achieve precise control of pipe wall thickness while ensuring production efficiency.

[0102] In this embodiment, the dynamic state of the rolls is precisely monitored by systematically combining multiple data sources (such as three-dimensional stress energy density distribution, ovality data, Barkhausen noise signals, and coolant flow rate). Intelligent adjustment is achieved through the use of advanced energy dissipation rate models, dynamic remaining life index generation, and rolling force compensation. By providing real-time feedback on the health of the rolls and the pipe, the system promptly identifies potential damage risks and adjusts rolling force and cooling intensity based on actual operating conditions, thus avoiding equipment failure and pipe quality issues caused by long-term neglect of microcrack growth and localized temperature unevenness.

[0103] Specifically, the system first collects the three-dimensional stress energy density distribution data of the roller bearing seat in real time through a multi-directional deformation energy sensor array, and predicts the growth trend of microcracks on the roller surface based on the energy dissipation rate model to generate a dynamic remaining life index. Through this method, the system can accurately grasp the health status of the roller and determine its remaining life, providing a scientific basis for subsequent compensation measures. At the same time, combined with the cross-sectional ovality data of the pipe, when the dynamic remaining life index falls below the preset threshold, the system automatically performs rolling force compensation to ensure that the roller pressure is evenly distributed and avoid local wall thickness mutations. This compensation not only targets the damage to the roller, but also comprehensively considers the geometric changes of the pipe, making the compensation effect more accurate and improving the rolling quality.

[0104] Furthermore, by synchronously collecting Barkhausen noise signals and cooling medium flow data, the system can dynamically adjust the cooling intensity under the action of a real-time feedback loop. This technology can accurately identify abnormal phase changes that may occur on the surface of the roller and judge the effectiveness of the cooling mode based on the flow data of the cooling medium. Through this technology, the unevenness in the cooling process is effectively suppressed, avoiding roller wear or material organization problems caused by local overheating or overcooling. Compared with traditional cooling control methods, the feedback loop of this application optimizes the distribution of cooling intensity, which not only improves the service life of the roller, but also improves the quality of rolled products.

[0105] The specific steps of S1 are as follows:

[0106] S1.1: Calculating local energy density distribution based on the three-dimensional stress energy density distribution data;

[0107] In this embodiment, a multi-directional deformation energy sensor array is first used to collect three-dimensional stress energy density distribution data of the roller bearing seat in real time. This data reflects the stress state and energy distribution in different areas of the roller surface during the rolling process.

[0108] Specifically, the three-dimensional data is mapped to local regions to identify stress concentration at different locations on the roll surface. Therefore, based on the three-dimensional stress energy density data, the entire roll surface is divided into multiple small regions using a spatial grid division technique, and the local energy density distribution is obtained by solving the local stress energy density in each region. The stress distribution at different locations on the roll surface is modeled in detail, providing basic data for subsequent crack propagation trend prediction. The high-risk areas of stress concentration are identified, and accurate spatial information is provided for subsequent fatigue crack prediction, so that the subsequent maintenance work can be more targeted.

[0109] S1.2: Extracting energy dissipation rate peak value and spatial distribution entropy value from the local energy density distribution, combining with the fatigue crack propagation rate test data of the roll material, establishing the mapping relationship between the energy threshold and the crack initiation position;

[0110] In this embodiment, the two indicators respectively reflect the concentration degree and spatial non-uniformity of energy consumption of the local region on the roll surface during rolling. The energy dissipation rate peak value can reveal the risk of crack propagation in the stress concentration area, and the spatial distribution entropy value can reflect the distribution characteristics of local energy and provide an important clue for the possibility of crack propagation.

[0111] Specifically, by conducting experiments on the fatigue crack propagation rate of different roll materials under different stress conditions, and combining with the energy data under actual working conditions, the threshold value of crack propagation and the corresponding crack initiation position are calculated. The experimental data used can accurately map different characteristics of the material and actual operating conditions, thereby ensuring the reliability of crack initiation prediction and providing more accurate input for the prediction model.

[0112] S1.3: Comparing the mapping relationship with the prediction relationship of the offline calibrated finite element simulation model, and correcting the time-varying parameters of the energy dissipation rate model through a residual driving algorithm, wherein the time-varying parameters represent the crack propagation coefficient, exponential factor, and crack propagation activation energy of the crack propagation cumulative damage characteristics of the roll material;

[0113] In this embodiment, the finite element simulation model provides a theoretical prediction of the stress on the roll surface and crack propagation, and combined with the constitutive relationship of the material, it can simulate the deformation and damage process of the roll material under different working conditions. However, due to a large number of uncertain factors (such as changes in rolling force, temperature, etc.) in actual operation, the prediction of the finite element simulation model may be biased. In order to further improve the prediction accuracy, the model needs to be corrected through a residual driving algorithm.

[0114] Specifically, the residual-driven algorithm corrects the time-varying parameters in the energy dissipation rate model by analyzing the errors between the mapping relationship and the prediction relationship of the simulation model. These time-varying parameters include crack propagation coefficient, exponential factor, and crack propagation activation energy, which represent the damage accumulation characteristics of the roll material. This makes the model more accurately reflect the damage evolution process of the roll in the actual production process, improves the prediction accuracy of the crack propagation trend, and effectively avoids equipment failure and production downtime caused by excessive damage.

[0115] S1.4: generating a crack propagation trend prediction according to the corrected energy dissipation rate model, wherein the crack propagation trend prediction includes a probability distribution thermogram of crack initiation location and a time-varying propagation rate curve of crack length;

[0116] Specifically, first, the probability distribution thermogram of crack initiation location is predicted by the corrected model, which shows the possibility and risk of crack initiation at different locations. Through the thermogram, the high-risk areas on the roll surface where crack propagation is most likely to occur can be accurately identified, providing data support for subsequent maintenance and prevention. Secondly, a time-varying propagation rate curve of crack length is generated, which shows the propagation speed of cracks at different time periods. Through these data, production line managers can monitor the propagation of cracks in real time and take appropriate control measures based on the prediction results, thereby effectively extending the service life of the equipment and reducing equipment downtime and production losses caused by crack propagation. Through these prediction results, managers can better control the health of the roll to ensure the stability of the rolling process and the quality of the product.

[0117] The specific steps of S1.3 are as follows:

[0118] S1.3.1: frequency domain decomposition of the residual signal between the mapping relationship and the prediction relationship, extracting the energy dominant mode of the characteristic frequency band;

[0119] In this embodiment, the mapping relationship is the mapping between the energy threshold and the crack initiation location obtained by the foregoing steps, and the prediction relationship is the crack propagation prediction result provided by the finite element simulation model. There may be errors between the two relationships, which may not be easily observed directly in the time domain, therefore, the frequency domain analysis method is used to more clearly identify and separate these residual signals.

[0120] Specifically, frequency domain decomposition techniques such as Fast Fourier Transform (FFT) are used to convert the residual signal in the time domain into a frequency domain signal, and the energy distribution in different frequency bands is analyzed. In the frequency domain, the energy-dominant mode refers to the pattern in which the energy of the signal is concentrated within a specific frequency range. These modes are usually closely related to the damage progression of the roll. The purpose of extracting the energy-dominant mode of the characteristic frequency band is to identify the signal features that are most relevant to the crack propagation process. These features can help identify the dynamic process of crack propagation and provide accurate data support for subsequent correction steps.

[0121] S1.3.2: Perform sensitivity analysis on the time-varying parameters in the energy dissipation rate model based on the energy dominant mode, calculate the sensitivity weights of the time-varying parameters using the roll material constitutive equation, and generate a priority sequence;

[0122] In this example, after obtaining the energy-dominant mode, the next step is to perform a sensitivity analysis on the time-varying parameters in the energy dissipation rate model. Time-varying parameters such as the crack growth coefficient, exponential factor, and crack growth activation energy are key variables in characterizing the roll material damage process.

[0123] Specifically, the constitutive equation of the roll material is used to describe its mechanical behavior, and the influence of time-varying parameters on damage propagation is derived through factors such as the material's stress-strain relationship, temperature changes, and strain rate. The purpose of sensitivity analysis is to quantify the impact of these time-varying parameters on crack propagation prediction, and to determine which parameters have a significant impact on the crack propagation trend and which parameters may contribute less to the final prediction results. It can clarify which time-varying parameters should be adjusted first during the correction process, thereby improving the efficiency of the correction process. The generated priority sequence provides a reference for subsequent correction steps, ensuring that the optimization process can focus on the most influential parameters to achieve the best correction effect.

[0124] S1.3.3: Performing a staged Bayesian correction on the time-varying parameters according to the priority sequence, wherein the staged Bayesian correction includes material constraints, energy constraints, and set constraints;

[0125] In this embodiment, after obtaining the priority sequence, the next step is to perform a phased Bayesian correction of the time-varying parameters. Based on Bayes' theorem, the Bayesian correction method gradually updates the model parameters by combining prior knowledge with observed data. The core of this phased correction is that, at different stages of the correction, more data and constraints are gradually introduced to gradually approach the final corrected value.

[0126] Specifically, for time-varying parameters, material constraints are first used to limit the correction range of the parameters. These constraints come from the inherent properties of the material, such as elastic modulus, yield strength, etc., to ensure that the physical properties of the material itself will not be violated during the correction process. Secondly, energy constraints ensure that the energy dissipation rate does not change abnormally during the correction process, ensuring the physical rationality of the entire process. Finally, the introduction of set constraints is to coordinate the interactions between different time-varying parameters, ensuring that these parameters can remain consistent during the correction process, and avoiding unreasonable correction results. Through this staged Bayesian correction method, time-varying parameters can be accurately optimized, the prediction accuracy of the model can be improved, and the uncertainty of the model can be reduced, thereby obtaining a more accurate prediction of crack propagation trends.

[0127] S1.3.4: Iteratively modify the time-varying parameters until the number of iterations is satisfied;

[0128] Specifically, the iterative correction process is to optimize the model by continuously updating the parameter values ​​and gradually reducing the residual between the model prediction value and the actual observation value. In each iteration, the time-varying parameters will be adjusted according to the sensitivity weights generated in the previous steps and the priority of the Bayesian correction, so that each step of the correction can be closer to the actual damage state. As the number of iterations increases, the prediction accuracy of the model continues to improve, and eventually a stable parameter solution can be obtained. This solution fully considers the actual evolution process of the roll damage and can accurately predict the crack propagation trend based on the current rolling conditions. Through this iterative process, it is ensured that the time-varying parameters can be continuously optimized in a dynamically changing environment, ultimately providing reliable crack propagation prediction results.

[0129] The specific steps of S2 are as follows:

[0130] S2.1: performing a coupled analysis on the probability distribution heat map and the time-varying growth rate curve to extract the credibility weight of the crack initiation location and the probability density function of the growth rate;

[0131] In this example, the probability distribution heatmap reflects the probability distribution of crack initiation at different locations, while the time-varying growth rate curve shows the crack growth rate over different time periods. By combining these two data sources, a more detailed and accurate analysis of the crack growth process can be achieved.

[0132] Specifically, the system identifies areas with high crack initiation probability in the thermal map and, based on the locations of these areas, extracts the corresponding crack growth velocity data from the time-varying growth rate curve. In other words, when a crack at a specific location is marked as a high-risk area, the system extracts the crack growth velocity data at that location from the growth rate curve. These two data sources are combined through the corresponding relationship between position and time to derive a new crack growth prediction model. By dynamically combining the crack initiation probability at the crack location with the crack growth rate, a more detailed modeling of the crack evolution process can be achieved, rather than simply analyzing crack occurrence or growth. Therefore, the results obtained from this coupled analysis provide the probability density function of the crack initiation location and the crack growth rate at that location in a specific area, which is crucial for further remaining life prediction. This allows accurate capture of the initial crack growth trend during actual rolling and dynamic prediction of crack growth behavior in future time periods, providing a scientific basis for subsequent maintenance decisions.

[0133] S2.2: Generate a probability distribution interval of the remaining useful life through Monte Carlo simulation based on the credibility weight and the probability density function;

[0134] In this embodiment, Monte Carlo simulation is used to generate a probability distribution interval of the remaining life. This process involves generating multiple possible remaining life scenarios through multiple random sampling and iterative calculations, and deriving an overall probability distribution interval from them.

[0135] Specifically, the system performs multiple sampling based on the confidence weights of the crack initiation locations and the probability density function of the crack growth rate. The confidence weights reflect the probability of crack initiation at different locations, while the probability density function of the growth rate describes the distribution of crack growth rates. In the Monte Carlo simulation, multiple possible crack initiation locations are first randomly sampled from the probability distribution of crack initiation locations. Then, samples are taken based on the crack growth rate probability density function for each location. Each sampling generates a combination of a specific location and a corresponding rate, representing the crack growth at that specific location.

[0136] Furthermore, in each simulation, these sampled results are combined to construct a complete crack propagation path, including the crack's growth process over time. These paths are then used to calculate the remaining life of the crack. Specifically, each path is simulated over time, until the crack reaches a critical size or the roll surface reaches a preset fatigue threshold, thereby determining the remaining life of the crack. The result of each simulation represents a possible crack growth scenario, and ultimately all simulation results form a set of remaining life values.

[0137] Further, through multiple random sampling and multiple rounds of simulation, the system can obtain a set of distribution data of the remaining life, from which the probability distribution interval of the remaining life can be calculated. This probability distribution interval reflects the remaining service life of the roll under different production conditions and provides support for subsequent decision-making. For example, using these simulation results, a confidence interval of the remaining life can be generated, indicating the remaining life of the roll within a certain probability range under given production conditions.

[0138] Further, by using the Monte Carlo simulation process, various uncertain factors in actual production (such as stress fluctuations, temperature changes, rolling force differences, etc.) can be considered, thereby providing more comprehensive and accurate remaining life prediction.

[0139] S2.3: Match the probability distribution interval with the order rolling mileage requirements in the production plan database, calculate the remaining life confidence that meets the current production task requirements, and generate a dynamic remaining life index;

[0140] In this embodiment, the production plan database contains the requirements of specific production tasks, such as rolling mileage, time, and other information of each order. By matching the probability distribution interval of the remaining life with these production requirements, the remaining life confidence of the current roll in completing a specific production task can be determined.

[0141] Specifically, according to the rolling mileage required by the current production task, combined with the distribution of the remaining life, it is calculated whether the roll can meet the requirements when completing the task, and then the roll whether it needs to be maintained or replaced is evaluated. The core is to compare the probability distribution of the remaining life with the requirements of the actual production task to provide real-time decision support for the production process. Not only improves the operation efficiency of the production line, but also effectively reduces the cost and risk caused by early or delayed maintenance of equipment. In addition, the generated dynamic remaining life index can be used as a real-time updated indicator to provide reference for production managers.

[0142] The specific steps of S3 are as follows:

[0143] S3.1: Divide the pipe cross-section ovality data into N detection regions according to the circumferential angle, calculate the wall thickness deviation rate of each region, and generate an ovality feature matrix;

[0144] In this step, it is first necessary to divide the ovality data of the pipe cross section according to the circumferential angle, and divide the cross section of the entire pipe into several small areas. The number of these small areas is N. The specific setting of N can be determined by a technician in this field through a large number of repeated experiments. Each area covers a certain angle range to ensure that the ovality changes of each part of the pipe can be reflected in detail. Ovality refers to the deviation of the pipe cross section caused by uneven rolling force or other factors during the rolling process, which is usually manifested as the non-circular degree of the pipe cross-section shape. By dividing these areas, the ovality changes of each area can be accurately monitored. Next, for each divided area, the wall thickness deviation rate of each area is calculated by measuring the deviation between the actual value of the pipe wall thickness and the standard wall thickness value. This process can reflect the uneven wall thickness or deformation that may occur in the pipe during the rolling process, and provide accurate data support for subsequent rolling force compensation. By generating an ovality characteristic matrix, the comprehensive situation of the wall thickness deviation and ovality of each area can be obtained. This matrix becomes the key data source for subsequent rolling force adjustment and optimization. By accurately monitoring the geometric deformation of the pipe, the uniformity of the wall thickness during the rolling process is ensured, the occurrence of pipe quality defects is avoided, and production efficiency and product quality are improved.

[0145] S3.2: Convert the dynamic remaining life index into a roll health weight;

[0146] Specifically, the system converts the dynamic remaining life index (DLI) into a roll health weight. The DLI reflects the remaining service life of the roll under its current production conditions and takes into account the growth trend of microcracks on the roll surface and other influencing factors. Analysis of this index yields a quantitative indicator of the roll's health—the roll health weight. The roll health weight represents the current health level of the roll, with higher values ​​indicating a healthier roll and lower values ​​indicating a higher risk of damage. During this process, the system uses a weighting function based on the roll's dynamic remaining life and health status to convert the DLI into a health weight. This allows the actual health of the roll to be used as input for the subsequent calculation of the rolling force compensation model. By quantifying the roll's health, a scientific basis is provided for accurate rolling force compensation, preventing excessive wear or quality defects caused by roll damage during the rolling process.

[0147] S3.3: Construct a rolling force compensation model based on the equipment health status, use the ellipticity characteristic matrix and the roll health weight as inputs of the rolling force compensation model, perform optimization through the rolling force compensation model, and calculate a rolling force compensation parameter set, specifically including:

[0148] generating an initial compensated pressure distribution using the ellipticity characteristic matrix as input;

[0149] The roll health weight is taken as a health weight constraint condition;

[0150] A Pareto optimal solution with the minimum increment of total rolling force is calculated by a particle swarm optimization algorithm, and the Pareto optimal solution is taken as a rolling force compensation parameter set;

[0151] In this step, the system will construct a rolling force compensation model based on the ellipticity feature matrix and the roll health weight. The ellipticity feature matrix contains the wall thickness deviation and ellipticity data of each region of the pipe, which reveals the changes in the geometry of the pipe during rolling. The roll health weight provides a quantitative assessment of the current health status of the rolls. Both of them are used as input data into the rolling force compensation model. The core purpose of the rolling force compensation model is to dynamically adjust the distribution of rolling force according to the geometric changes of the pipe and the health status of the rolls, ensuring that the rolling force is evenly distributed throughout the rolling process, thereby achieving precise control of the wall thickness of the pipe. The compensation model optimizes the rolling force by adjusting the applied pressure of each roll, avoiding excessive compression or stress concentration in local areas, and ensuring the uniformity of the pipe wall thickness and the service life of the rolls.

[0152] Next, the system uses a particle swarm optimization algorithm (PSO) to optimize the rolling force compensation model and calculate the rolling force compensation parameter set. Particle swarm optimization is an optimization method that simulates the behavior of a particle swarm in nature, which can effectively search for optimal solutions in multi-dimensional parameter space. In this process, the particle swarm optimization algorithm will perform multiple iterative calculations in the parameter space of rolling force adjustment based on the input data of the rolling force compensation model (i.e., the ellipticity feature matrix and the roll health weight), to find the optimal solution that minimizes the total rolling force increment. Through this optimization, the system can calculate the rolling force compensation parameter set that best suits the current production status, which will be used to adjust the distribution of rolling force in real time, ensuring the quality of the pipe and the health of the rolls during production.

[0153] Further, the particle swarm optimization algorithm calculates a Pareto optimal solution through multiple iterations of the rolling force compensation model. The Pareto optimal solution refers to a solution in a multi-objective optimization problem that cannot be improved without sacrificing other objectives. In this embodiment, the Pareto optimal solution represents the optimal solution that balances the health of the rolls, the uniformity of the pipe wall thickness, and the minimum increment of rolling force in the rolling force compensation model. Through this optimal solution, the system can ensure that the pressure applied by each roll meets the production requirements while minimizing the increment of rolling force, thereby effectively reducing the wear and tear of the rolls and the pipe. Ultimately, these calculated parameter sets will be used to adjust the rolling force in real time to ensure the smooth progress of the entire rolling process.

[0154] The cooling intensity is adjusted according to the real-time feedback loop, including:

[0155] The amplitude and frequency domain characteristics of the Barkhausen noise signal are used to identify the local magnetic domain changes of the roller, and a mapping relationship is established between the abnormal phase change characteristics of the roller surface.

[0156] When the abnormal phase change characteristics indicated by the mapping relationship exceed the process allowable range, the specific location and extent of the abnormal roll area are determined by extracting the multimodal features of the Barkhausen noise signal;

[0157] According to the specific location and extent of the abnormal roll area, the cooling nozzle combination opening strategy is adjusted;

[0158] The adjusting the cooling intensity according to the real-time feedback loop further includes:

[0159] determining the effectiveness of the current cooling mode according to the cooling medium flow data;

[0160] When the cooling medium flow rate data deviates locally from the set value, the local change in the cooling medium flow rate is correlated to determine the spatial area with uneven cooling distribution;

[0161] According to the spatial area where the cooling is unevenly distributed, the cooling nozzle combination opening strategy is adjusted.

[0162] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. An equipment status monitoring and maintenance system, applied to a steel pipe continuous rolling production line, characterized in that: The system includes a data acquisition module, a data analysis module and an equipment maintenance module, wherein: The data acquisition module is used to collect three-dimensional stress energy density distribution data, Barkhausen noise signal and cooling medium flow data of the roller bearing seat; The data analysis module establishes an energy dissipation rate model based on the three-dimensional stress energy density distribution data, predicts the growth trend of microcracks on the roll surface and generates a dynamic remaining life index. At the same time, based on the amplitude and frequency domain characteristics of the Barkhausen noise signal, it establishes a mapping relationship between the local magnetic domain changes of the roll and the abnormal phase change characteristics. It also determines the effectiveness of the current cooling mode based on the cooling medium flow data and identifies the cooling abnormality area. The equipment maintenance module compensates for the roll reduction amount according to the output of the data analysis module and adjusts the opening strategy of the cooling nozzle combination.

2. The equipment status monitoring and maintenance system according to claim 1, characterized in that: The data analysis module includes: The crack prediction unit builds an energy dissipation rate model based on three-dimensional stress energy density distribution data, dynamically corrects time-varying parameters through a residual-driven algorithm, and generates a thermal map of the crack initiation location and a time-varying growth rate curve; The remaining life analysis unit is used to couple crack prediction results with production condition data, generate the remaining life probability distribution interval through Monte Carlo simulation, and generate a dynamic remaining life index based on order requirements; The cooling intensity analysis unit is used to analyze the magnetic domain-phase change mapping relationship of the Barkhausen noise signal and identify abnormal phase change areas and cooling failure areas based on the cooling medium flow distribution characteristics.

3. The equipment status monitoring and maintenance system according to claim 2, characterized in that: The crack prediction unit is configured with prediction logic, which includes: Calculating local energy density distribution based on the three-dimensional stress energy density distribution data; Extracting the energy dissipation rate peak and the spatial distribution entropy value based on the local energy density distribution, and establishing a mapping relationship between the energy threshold and the crack initiation position in combination with the fatigue crack growth rate test data of the roll material; Comparing the mapping relationship with the predicted relationship of the offline calibrated finite element simulation model, and correcting the time-varying parameters of the energy dissipation rate model through a residual-driven algorithm, wherein the time-varying parameters characterize the crack growth coefficient, exponential factor, and crack growth activation energy of the damage accumulation characteristics of the roll material; A crack growth trend prediction is generated based on the modified energy dissipation rate model, wherein the crack growth trend prediction includes a probability distribution heat map of the crack initiation position and a time-varying growth rate curve of the crack length.

4. The equipment status monitoring and maintenance system according to claim 3, characterized in that: The method of correcting the time-varying parameters of the energy dissipation rate model by using a residual driving algorithm includes: Perform frequency domain decomposition on the residual signal between the mapping relationship and the prediction relationship to extract the energy dominant mode of the characteristic frequency band; performing sensitivity analysis on the time-varying parameters in the energy dissipation rate model according to the energy dominant mode, calculating the sensitivity weights of the time-varying parameters through the constitutive equation of the roll material, and generating a priority sequence; Performing a staged Bayesian correction on the time-varying parameters according to the priority sequence, wherein the staged Bayesian correction includes material constraints, energy constraints, and set constraints; The time-varying parameters are iteratively modified until a predetermined number of iterations is met.

5. The equipment status monitoring and maintenance system according to claim 3, characterized in that: The remaining life analysis unit is configured with life analysis logic, and the life analysis logic includes: Couple the probability distribution heat map and the time-varying expansion rate curve to extract the credibility weight of the crack initiation position and the probability density function of the expansion rate; Generating a probability distribution interval of the remaining life through Monte Carlo simulation according to the credibility weight and the probability density function; The probability distribution interval is matched with the order rolling mileage requirements in the production plan database, the remaining life confidence level that meets the current production task requirements is calculated, and a dynamic remaining life index is generated.

6. The equipment status monitoring and maintenance system according to claim 2, characterized in that: The cooling intensity analysis unit is configured with cooling identification logic, which includes: The amplitude and frequency domain characteristics of the Barkhausen noise signal are used to identify the local magnetic domain changes of the roller, and a mapping relationship is established between the abnormal phase change characteristics of the roller surface. When the abnormal phase change characteristics indicated by the mapping relationship exceed the process allowable range, the position and extent of the abnormal roll area are determined by extracting the multimodal features of the Barkhausen noise signal.

7. The equipment status monitoring and maintenance system according to claim 6, characterized in that: The cooling identification logic further includes: determining the effectiveness of the current cooling mode according to the cooling medium flow data; When the cooling medium flow rate data locally deviates from the set value, the local change in the cooling medium flow rate is correlated to determine the spatial region with uneven cooling distribution.

8. The equipment status monitoring and maintenance system according to claim 1, characterized in that: The equipment maintenance module includes: A rolling force compensation unit, configured to receive ovality detection data and a dynamic remaining life index and generate a rolling force compensation parameter set; The cooling intensity adjustment unit adjusts the nozzle opening and closing combination and flow distribution according to the abnormal phase change area and cooling failure area.

9. The equipment status monitoring and maintenance system according to claim 8, characterized in that: The rolling force compensation unit is provided with compensation logic, which includes: The ovality data of the pipe section is divided into N detection areas according to the circumferential angle, the wall thickness deviation rate of each area is calculated, and the ovality feature matrix is ​​generated; converting the dynamic remaining life index into a roll health weight; A rolling force compensation model based on the health status of the equipment is constructed, the ellipticity characteristic matrix and the roll health weight are used as inputs of the rolling force compensation model, optimization is performed through the rolling force compensation model, and a rolling force compensation parameter set is calculated.

10. The equipment status monitoring and maintenance system according to claim 9, characterized in that: Optimizing the rolling force compensation model to calculate a rolling force compensation parameter set includes: generating an initial compensated pressure distribution using the ellipticity characteristic matrix as input; Using the health weight of the roll as a health weight constraint condition; The Pareto optimal solution with the minimum total rolling force increment is calculated by a particle swarm optimization algorithm, and the Pareto optimal solution is used as a rolling force compensation parameter set.

Images