Corrugated board processing equipment failure prediction and health management system

Abstract

This invention relates to the field of industrial intelligent manufacturing and equipment health management technology, specifically to a fault prediction and health management system for corrugated cardboard processing equipment. The system includes: a data acquisition terminal for acquiring production process operating parameters and real-time physical observation data of the target production equipment, and sending the real-time physical observation data to a residual analysis module; a baseline generation server configured with a physical information neural network model, and sending dynamic reference baseline data to the residual analysis module; the residual analysis module for calculating the deviation between the real-time physical observation data and the dynamic reference baseline data, and sending the equipment physical degradation residual to a decision execution terminal; and the decision execution terminal for analyzing the equipment physical degradation residual, generating a status assessment result and control commands for the target production equipment. This invention effectively resists the long-tail effect and predicts degradation caused by service aging.

Description

Technical Field

[0001] This invention relates to the field of industrial intelligent manufacturing and equipment health management technology, specifically to a fault prediction and health management system for corrugated cardboard processing equipment. Background Technology

[0002] Corrugated cardboard processing equipment is a core production link in the packaging manufacturing industry. The operating status of the processing equipment, especially the core components, directly affects the continuity of the production line and the quality of cardboard forming. How to accurately monitor the health status of the equipment and predict potential failures has become a key research direction in the packaging manufacturing industry.

[0003] The corrugated cardboard processing system mainly includes corrugated rollers, matching bearings, and rotary joints, which are thermo-mechanical coupling components. In actual production, the pressing and shaping of the cardboard is completed through complex mechanical rigid motion and heat transfer. Existing technologies rely mainly on subjective judgment based on manual observation of charts or traditional pure data network models when predicting faults. This cannot effectively isolate the data distribution deviation caused by high-frequency order switching and process changes, and it is difficult to adapt to complex and ever-changing thermo-mechanical coupling working conditions. It is also easy to misreport normal process fluctuation noise as physical damage to equipment, resulting in a bottleneck from data monitoring to business operations.

[0004] The operating state of corrugated processing equipment exhibits significant nonlinear characteristics and complex physical properties. Existing pure data networks, lacking constraints from physical laws, struggle to decouple from the underlying physical degradation state of the equipment. This is especially true when speeds surge or operating conditions change abruptly, leading to a significant drop in prediction accuracy, potentially resulting in conclusions that contradict physical principles and slow response times. Therefore, a solution is urgently needed to address the problems present in existing technologies. Summary of the Invention

[0005] The purpose of this invention is to provide a fault prediction and health management system for corrugated cardboard processing equipment, and to solve the following technical problems:

[0006] By stripping away the data distribution shifts caused by surface process changes, the true physical health status of the equipment at its underlying level is accurately decoupled; bridging the semantic gap between data features and underlying causality, the system outputs deterministic physical quantities that frontline engineers can directly understand and use; eliminating the differences in subjective human judgment, the system achieves automated status interception and deterministic closed-loop intervention; and ultimately, it opens up the path from deep digital diagnostics to autonomous hardware correction, achieving true intelligent closed-loop control.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] The fault prediction and health management system for corrugated cardboard processing equipment includes: a data acquisition terminal, a baseline generation server, a residual analysis module, and a decision execution terminal;

[0009] The data acquisition terminal is used to acquire the production process operation parameters and real-time physical observation data of the target production equipment, and send the production process operation parameters to the baseline generation server and the real-time physical observation data to the residual analysis module.

[0010] The baseline generation server is configured with a physical information neural network model. The baseline generation server is used to input the production process operating parameters into the physical information neural network model, calculate the dynamic reference baseline data of the target production equipment under the current production process operating parameters, and send the dynamic reference baseline data to the residual analysis module.

[0011] The residual analysis module is used to calculate the deviation between the real-time physical observation data and the dynamic reference baseline data, determine the deviation as the equipment physical degradation residual, and send the equipment physical degradation residual to the decision execution terminal.

[0012] The decision execution terminal is used to generate status assessment results and control instructions for the target production equipment based on the physical degradation residuals of the equipment.

[0013] Furthermore, the decision execution terminal is also used for:

[0014] Retrieve the preset degradation threshold from the storage space;

[0015] The configuration is as follows: if the physical degradation residual of the equipment is greater than or equal to the preset degradation threshold, then it is determined that the target production equipment is abnormal, and a status warning and maintenance suggestion instruction is generated as the control instruction; if the physical degradation residual of the equipment is less than the preset degradation threshold, then it is determined that the target production equipment is not abnormal.

[0016] Furthermore, the system also includes: a prior model initialization module;

[0017] The prior model initialization module is used to obtain the structural topology and physical attribute parameters of the target production equipment, and to convert the structural topology and physical attribute parameters into structured prior parameters and boundary conditions of the physical information neural network model.

[0018] Furthermore, the physical degradation residual of the equipment includes the true mechanical stiffness attenuation coefficient or the fluid anomalous accumulation index, which characterizes the actual mechanical stiffness after eliminating fluctuations in the production process.

[0019] Furthermore, the decision execution terminal is also used for:

[0020] Obtain actual wear data corresponding to the target production equipment where the anomaly exists;

[0021] The abnormal physical degradation residuals of the equipment and the actual wear conditions are stored as tagged data samples;

[0022] The structured prior parameters in the physical information neural network model are adaptively calibrated and updated using the labeled data samples.

[0023] Furthermore, the production process operating parameters include equipment operating speed, load parameters, and set energy input parameters; the real-time physical observation data includes vibration parameters and temperature parameters.

[0024] Furthermore, the prior model initialization module is also used for:

[0025] Obtain the thermodynamic and kinetic equations;

[0026] The thermodynamic equation and the kinetic equation are incorporated as penalty terms into the loss function of the physical information neural network model to constrain the generation of the dynamic reference baseline data using the thermodynamic equation and the kinetic equation.

[0027] Furthermore, generate status alerts and maintenance suggestion instructions, including:

[0028] Based on the physical degradation residual of the equipment, determine the target production process operating parameters corresponding to the anomaly;

[0029] Recommended production process adjustment parameters are generated as the status warning and maintenance suggestion instructions, wherein the recommended production process adjustment parameters include a suggested reduction in equipment operating speed reference value or a suggested fine adjustment in equipment pressure reference value.

[0030] Furthermore, the decision execution terminal is also used for:

[0031] The status warning and maintenance suggestion instructions are sent to the control terminal of the target production equipment to prompt adjustments to the working status of the target production equipment and to avoid the equipment resonance zone.

[0032] The beneficial effects of this invention are:

[0033] 1. This invention calculates dynamic reference baseline data by combining a physical information neural network model with production process operating parameters, and compares it with real-time physical observation data to obtain equipment physical degradation residuals. This mechanism effectively filters out interference caused by normal process fluctuations, decouples the true physical state of the equipment at the underlying level, and avoids the problem of traditional pure data-driven models misreporting process fluctuations as equipment abnormalities when frequently cutting orders.

[0034] 2. The prior model initialization module of this invention transforms the structural topology and physical properties of the target production equipment into structured prior parameters and boundary conditions, and incorporates thermodynamic and kinetic equations as penalty terms into the loss function; this provides a solid physical foundation for the artificial intelligence model, suppresses the random divergence of the neural network model under unexhaustive working conditions, and ensures that the baseline deduction strictly conforms to the laws of nature.

[0035] 3. This invention concretely decouples the physical degradation residual of equipment into a real mechanical stiffness attenuation coefficient or fluid abnormal accumulation index after eliminating production process fluctuations; this approach bridges the semantic gap between high-dimensional abstract data and underlying causality, outputs a deterministic physical quantity that front-line engineers can intuitively understand and use, and breaks through the subjective differences and technical bottlenecks of human judgment.

[0036] 4. The decision execution terminal of this invention can acquire actual wear data of the equipment and store it and degradation residual as labeled data samples, and adaptively calibrate and update the structured prior parameters in the physical information neural network model; this enables the model prediction boundary to fit the current life cycle of the equipment and effectively resist the long tail effect and prediction degradation caused by service aging.

[0037] 5. This invention generates process adjustment parameter information based on the physical degradation residuals of the equipment, which includes suggestions for fine-tuning the equipment pressure or speed, and sends it directly to the control terminal. This mechanism can seamlessly adjust the working state to avoid the equipment resonance zone, and suppress the degradation trend through automated optimization before irreversible damage occurs, thus ensuring the continuous operation of the production line. Attached Figure Description

[0038] The invention will now be further described with reference to the accompanying drawings.

[0039] Figure 1 A schematic diagram of the modules of the corrugated cardboard processing equipment fault prediction and health management system provided in the embodiments of this application. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Please see Figure 1 The fault prediction and health management system for corrugated cardboard processing equipment includes: a data acquisition terminal, a baseline generation server, a residual analysis module, and a decision execution terminal.

[0042] Among them, the data acquisition terminal is used to acquire the production process operation parameters and real-time physical observation data of the target production equipment, and send the production process operation parameters to the baseline generation server and the real-time physical observation data to the residual analysis module;

[0043] The baseline generation server is configured with a physical information neural network model. The baseline generation server is used to input the production process operating parameters into the physical information neural network model, calculate the dynamic reference baseline data of the target production equipment under the current production process operating parameters, and send the dynamic reference baseline data to the residual analysis module.

[0044] The residual analysis module is used to calculate the deviation between real-time physical observation data and dynamic reference baseline data, determine the deviation as equipment physical degradation residual, and send the equipment physical degradation residual to the decision execution terminal.

[0045] The decision execution terminal is used to generate status assessment results and control instructions for target production equipment based on the physical degradation residuals of the equipment.

[0046] This embodiment provides the basic operating mechanism of a fault prediction and health management system for corrugated cardboard processing equipment. Specifically, the system is deployed in a real industrial cardboard manufacturing environment to monitor the complex thermo-mechanical-fluid states of core components such as corrugated rolls. The data acquisition terminal collects production flow data in the first vector dimension in real time. Sensor observation data in the second vector dimension Such as bearing end vibration timing signal;

[0047] The baseline generation server has a built-in physical information neural network that receives... And it forward deduces the dynamic reference baseline data that a fully healthy device should exhibit under this specific thermo-mechanical coupling condition. The residual analysis module performs difference calculations between the actual observed values ​​and the theoretical baseline values, and then performs spatial dimensionality reduction processing such as L2 norm aggregation to obtain the scalar form of the equipment physical degradation residuals.

[0048]

[0049] The decision execution terminal is based on this residual Generate status assessment results and control commands; at the anomaly handling boundary, if the data acquisition terminal experiences a brief signal loss due to strong electromagnetic interference on site, the baseline generation server will freeze and maintain the previous moment's data. Maintain the baseline simulation until communication is restored;

[0050] If the missing time exceeds the preset communication heartbeat signal timeout threshold, a safety warning for system degradation will be directly output. For example, when corrugated cardboard production is switching orders frequently, such as when the speed of the machine increases and the moisture content of the raw paper rises sharply, the system will automatically increase the steam pressure. Traditional pure data models will mistakenly report the sudden change in thermal expansion vibration of the corrugated roll caused by this as bearing damage.

[0051] In this embodiment, the baseline generation server synchronously simulates and obtains a new baseline that includes thermal response characteristics. This makes the residual after the difference between the two It maintains stability and accurately filters out normal process fluctuation noise; the purpose of this mechanism is to remove the data distribution offset caused by surface process changes and decouple the true physical health status of the equipment.

[0052] In a preferred embodiment of the present invention, the decision execution terminal is further configured to: obtain a preset degradation critical threshold from the storage space; and be configured to: determine that the target production equipment is abnormal if the physical degradation residual of the equipment is greater than or equal to the preset degradation critical threshold, and generate a status warning and maintenance suggestion instruction as a control instruction; and determine that the target production equipment is not abnormal if the physical degradation residual of the equipment is less than the preset degradation critical threshold.

[0053] This embodiment provides a mechanism for automated judgment and alarm based on degradation residuals. If the system can only output continuously fluctuating physical degradation residual values, field operators often find it difficult to determine when maintenance should be intervened, creating a technical bottleneck of a gap between data monitoring and operational actions. Specifically, the decision execution terminal reads a preset degradation critical threshold for a specific equipment model from the local configuration library or the storage space of the cloud server. ;

[0054] Preset degradation threshold It is a constant scalar obtained by extracting the mean residual value within 5% of the time window before irreversible plastic deformation or fracture of the equipment based on the historical full life cycle fatigue failure test dataset of the same type of equipment.

[0055] Perform hard logic comparison: If the calculated equipment physical degradation residuals are... ,in Once the preset degradation threshold is reached, the system determines that the current target production equipment is abnormal and packages and generates a status warning and maintenance suggestion instruction containing the abnormality confidence level and the repair part.

[0056] like If the target production equipment is within the health tolerance range, no intervention will be taken; regarding boundary conditions, if the extracted threshold is reduced due to sector damage in the storage space... If the value is empty or an illegal negative value is parsed, the system will refuse to perform a regular comparison and automatically downgrade to the factory-programmed conservative safety threshold as a system safety redundancy mechanism to prevent missed detections; for example, when the system monitors wear on the non-drive end bearing of the upper corrugated roller, the calculated degradation residual... Climb to Exceeding the preset critical threshold At that time, the decision execution terminal immediately identifies the anomaly and generates an early warning instruction containing maintenance suggestions to prevent the generation of waste products in advance. The purpose of this step is to eliminate the subjective judgment difference of human chart reading and realize automated status interception and deterministic closed-loop intervention.

[0057] In a preferred embodiment of the present invention, the system further includes: a prior model initialization module; the prior model initialization module is used to obtain the structural topology and physical attribute parameters of the target production equipment, and to convert the structural topology and physical attribute parameters into structured prior parameters and boundary conditions of the physical information neural network model.

[0058] This embodiment provides a prior model initialization mechanism based on physical mechanisms; neural networks driven purely by massive time series lack specific physical constraints in the cold start phase, which can easily lead to the initial baseline being seriously contrary to the common sense of real dynamics and slow network convergence.

[0059] Specifically, at the beginning of the monitoring task, the prior model initialization module reads the structural topology matrix of the target production equipment. and physical property parameter set Includes corrugated roll quality scalar Scalar of rib type coefficient Thermal conductivity scalar The module converts these parameters into fixed structured prior parameters and optimization boundary conditions for the internal computation graph of the physical information neural network. If a certain physical property parameter is missing or seriously exceeds the range of real dimensions, such as a negative thermal conductivity, the prior model initialization module will intercept the initialization process, throw a parameter abnormal configuration log, and suspend it for manual verification and repair.

[0060] For example, the system pre-converts the thermal-mechanical digital topology of a specific type of corrugated unit and the normal mechanical clearance range into network connection weight constraints, giving the model a priori physical mechanism constraint for that specific device; the purpose of this mechanism is to build a solid physical prior foundation for the AI ​​model, so that the dynamic reference baseline data it generates has native high fidelity.

[0061] In a preferred embodiment of the present invention, the equipment physical degradation residual includes the true mechanical stiffness attenuation coefficient or the fluid anomalous accumulation index, which characterizes the actual mechanical stiffness after eliminating production process fluctuations.

[0062] This embodiment provides a mechanism for concretely decoupling physical degradation residuals of equipment; conventional feature extraction methods often output abstract high-dimensional vectors, which field operators cannot understand the actual engineering significance of hidden layer state shifts.

[0063] Specifically, the residual analysis module performs a second decomposition of the comprehensive deviation value in the frequency and time domains. The specific steps are as follows: the comprehensive deviation value is converted to the frequency domain using a fast Fourier transform, and the feature components with frequencies greater than the first threshold are extracted by a preset high-pass filter as high-frequency transient mechanical degradation components, and the feature components with frequencies less than the second threshold are extracted by a low-pass filter as low-frequency gradual heat transfer resistance components. The first threshold is calibrated based on the inherent mechanical resonance frequency of the target production equipment, and the second threshold is calibrated based on the highest frequency of the heat transfer cycle.

[0064] The inherent mechanical resonance frequency is determined by extracting the center frequency corresponding to the main resonance peak through a factory modal sweep frequency test on the target production equipment in a healthy state. The first threshold is specifically set to 1.2 to 1.5 times the center frequency. The empirical highest frequency is calculated based on the reciprocal of the duration of a single physical cycle of steam fluid in the rotary joint. The second threshold is specifically set to 0.8 times the reciprocal value.

[0065] The root mean square value of the time-domain energy of the high-frequency transient mechanical degradation component is multiplied by a preset stiffness conversion coefficient to output the true mechanical stiffness attenuation coefficient. The sliding window integral mean of the low-frequency gradually varying heat transfer resistance component is output as the fluid anomalous accumulation index. The preset stiffness conversion coefficient is obtained from the stiffness test calibration data of the target production equipment at the factory; the actual mechanical stiffness attenuation coefficient is calculated using the following formula:

[0066]

[0067] in, The preset stiffness conversion factor is... For the first The amplitude of the high-frequency transient mechanical degradation component at each discrete sampling point The total number of sampling points within the sampling window; Fluid anomalous accumulation index:

[0068]

[0069] in, The preset sliding window time step, For a moment The corresponding low-frequency gradual heat transfer resistance component amplitude;

[0070] In the exception handling branch, if the module finds that the eigenvalue matrix is ​​singular or non-convergent during the decoupling operation, the system will determine that the current data frame has been severely affected by external noise, automatically discard the calculation result, and smoothly extend the residual evaluation value of the previous cycle at the output.

[0071] For example, when scaling is detected inside the steam rotary joint after years of operation, the system, after successfully removing the steam pressure fluctuations caused by normal machine acceleration, directly and accurately outputs a continuously rising fluid abnormality accumulation index. The curve clearly points to the fluid stagnation inside the rotary joint; the purpose of this step is to establish a mapping relationship between the high-dimensional abstract data features and the real physical degradation causality of the underlying equipment, and to output deterministic physical quantities that front-line engineers can directly understand and use.

[0072] In a preferred embodiment of the present invention, the decision execution terminal is further configured to: acquire actual wear data corresponding to the target production equipment with anomalies; store the physical degradation residuals and actual wear data of the abnormal equipment as labeled data samples; and use the labeled data samples to adaptively calibrate and update the structured prior parameters in the physical information neural network model.

[0073] This embodiment provides a mechanism for adaptive model calibration based on real feedback. As the corrugated unit's service time increases, its inherent material properties will age. If the physical structural parameters at the initial stage are always adhered to, the systematic error in the model baseline extrapolation will gradually amplify. Specifically, after a round of shutdown maintenance, the decision execution terminal will obtain the actual wear data measured and entered by maintenance personnel through the human-computer interaction interface. For example, the measured increase in bearing clearance is ;

[0074] The system compares the high-frequency equipment physical degradation residual sequence recorded by the device before the failure with... Anchoring, encapsulation into high-quality labeled data samples The system initiates a background fine-tuning task, using the labeled sample to perform adaptive calibration and update of the structured prior parameters at the bottom layer of the physical information neural network model through backpropagation calculation.

[0075] In the abnormal data filtering mechanism, if manually entered... There is a serious logical contradiction with the residual trend recorded in the previous period. For example, the system records a very large residual, but the entered wear amount is a completely healthy zero value. The system marks the sample as conflicting and waiting to be updated, isolates it and does not participate in the automatic weight update, so as to prevent the introduction of noisy data due to human error, thereby reducing the accuracy of model prediction.

[0076] For example, after a real paper feed roller bearing replacement operation, the system completes a closed-loop learning using data with strong physical causal labels, making its prediction boundary for this specific wear mode more consistent with the current life cycle of the equipment; the purpose of this mechanism is to give the system long-term self-evolution capability and resist the prediction degradation caused by the long tail effect of the equipment's entire life cycle.

[0077] In a preferred embodiment of the present invention, the production process operating parameters include equipment operating speed, load parameters, and set energy input parameters; real-time physical observation data include vibration parameters and temperature parameters.

[0078] This embodiment provides a mechanism for cross-domain multimodal data classification and acquisition; specifically, the data acquisition terminal is equipped with dual concurrent acquisition channels: the industrial bus channel is responsible for directly retrieving process parameter streams characterizing system input attributes from the control layer, specifically including equipment operating speed. Load parameters determined by paper tension And the energy input parameters set on the control panel, such as the target steam pressure. ;

[0079] Meanwhile, the sensor channel is responsible for capturing observational data characterizing the system's output state from the physical environment layer, including vibration parameters acquired by the high-frequency accelerometer. Local temperature parameters of infrared array scanning If a sensor experiences a disconnection or communication protocol parsing failure, the data acquisition terminal will not stop working. Instead, it will immediately broadcast a sensor fault code to the bus and fill the specified placeholder, such as a specific NaN extreme value, into that feature dimension to notify the downstream module that the dimension is temporarily invalid.

[0080] For example, when production starts, the system continuously takes in process control instructions determined by the attributes of the current order, such as... Upgraded to Simultaneously, accurately record the real-time generation at the bearings at both ends of the corrugated roller. and Fluctuations; the purpose of this setting is to fully collect the constraint boundaries and actual responses of the system across all dimensions of thermo-mechanical-fluid-process, providing complete data elements for constructing digital twins of nonlinear systems.

[0081] In a preferred embodiment of the present invention, the prior model initialization module is further configured to: obtain thermodynamic equations and kinetic equations; incorporate the thermodynamic equations and kinetic equations as penalty terms into the loss function of the physical information neural network model, so as to use the thermodynamic equations and kinetic equations to constrain the generation of dynamic reference baseline data.

[0082] This embodiment provides a mechanism for using physical law equations to deeply constrain neural networks. In the corrugated packaging industry, the combination of working conditions brought about by high-frequency cutting is almost infinite. When faced with an unexhaustible working condition space, pure data-driven algorithms often output prediction results that defy common sense in physics, causing large-scale false alarms in the system.

[0083] Specifically, the prior model initialization module will characterize the thermodynamic equations of ultra-fast heat conduction. and the dynamic equations characterizing rigid body motion Extract and symbolize; wherein, the thermodynamic equation The specific mathematical expression is as follows:

[0084]

[0085] in, Temperature distribution on the surface of the corrugated roll. The thermal conductivity scalar value obtained above, The input heat source term is calculated from the set steam pressure; the kinetic equation The specific mathematical expression is as follows:

[0086]

[0087] in, This refers to the radial displacement vibration of the corrugated roller. The aforementioned quality scalar value of the corrugated roll. The equivalent stiffness of the system is the same as the aforementioned scalar coefficient of the rib type. A positive correlation is shown. The system damping coefficient is... The excitation force is determined by the load parameters;

[0088] The computational residuals are constructed as a physical mechanism loss term, and combined with a scalar of the weight hyperparameter penalty term. The total loss function, along with λ2, is incorporated into the physical information neural network model. In the middle; specifically, the total loss function The expression is:

[0089]

[0090] in, This represents the mean squared error loss from data fitting. To calculate the residuals for the thermodynamic equations, Calculate the residuals for the dynamic equations; force the model to not only fit historical data, but also minimize the penalty terms in the equations;

[0091] If, during the network training optimization phase, local minima cause the gradient of the overall loss value, including the penalty term, to expand rapidly or even overflow, the system will automatically trigger gradient pruning and dynamically reduce the weight coefficients to ensure stable convergence of model training.

[0092] For example, when a corrugating machine experiences a momentary paper break while operating at high speed, the system's process environment undergoes a sudden change. The constructed penalty term strongly constrains the network nodes, ensuring that the dynamic baseline predicted by the model fully conforms to the physical fading law of transient heat dissipation, rather than randomly oscillating data points. The purpose of this mechanism is to suppress the random divergence of the neural network model from the bottom layer of the AI ​​algorithm, ensuring that the deduction of the dynamic reference baseline always operates within the framework of real natural laws.

[0093] In a preferred embodiment of the present invention, generating a status warning and maintenance suggestion instruction includes: determining the target production process operating parameters corresponding to the anomaly based on the equipment physical degradation residual; generating recommended production process adjustment parameter information as the status warning and maintenance suggestion instruction, wherein the recommended production process adjustment parameter information includes a suggested reduction in equipment operating speed reference value or a suggested fine adjustment in equipment pressure reference value.

[0094] This embodiment provides a process adaptive adjustment calculation mechanism based on degradation state; specifically, the decision execution terminal has a built-in process control optimization algorithm. When it is determined that there is an abnormality in the target production equipment, the current target production process operation parameters that cause the degradation residual to exceed the limit are extracted; then, with the goal of reducing the physical degradation residual of the equipment, within the preset safety process tolerance range, the process adjustment amount used to alleviate the physical stress of the equipment is calculated.

[0095] For example, if the fluid abnormality accumulation index indicates that there is thermal stagnation in the rotary joint, and the system calculates that the heat cannot be effectively conducted under the current set steam pressure, then a status warning and maintenance suggestion instruction containing a suggested fine-tuning of the equipment pressure reference value will be generated.

[0096] If the actual mechanical stiffness attenuation coefficient indicates early wear in the bearing, the system generates a status warning and maintenance recommendation instruction that includes a suggested reduction in the equipment operating speed reference value. The purpose of this step is to suppress the physical degradation trend through automated process optimization before irreversible damage occurs, thus ensuring the continuity of the production line.

[0097] In a preferred embodiment of the present invention, the decision execution terminal is further configured to: send status warning and maintenance suggestion instructions to the control terminal of the target production equipment, so as to prompt adjustment of the working status of the target production equipment and avoid the equipment resonance zone.

[0098] This embodiment provides a system-level self-healing and anti-resonance avoidance mechanism for unmanned equipment. Even if high-precision process adjustment suggestions are output, if manual execution is carried out after layer-by-layer confirmation, the control delay may still lead to irreversible damage to the equipment or scrapping of the entire batch of cardboard. Specifically, the decision execution terminal sends the formatted status warning and maintenance suggestion instructions directly as control messages to the control terminal of the target production equipment through the industrial Ethernet interface. After authentication and reception, the control terminal directly drives the motor and valves, seamlessly adjusting the working state of the equipment and driving the equipment away from the dangerous frequency band to avoid the equipment resonance zone.

[0099] If the control terminal times out when executing the servo action after receiving the instruction, or if the underlying driver reports a hardware-level denial-of-service signal such as overload or jamming, the decision execution terminal will immediately terminate the automated tuning closed loop and trigger a high-level audible and visual emergency stop alarm, handing over the task to manual forcible takeover.

[0100] For example, during monitoring, the system discovered that although the mechanical components themselves were not damaged, the operating baseline of the equipment was abnormally approaching the inherent resonance frequency band of its mechanical structure under the currently selected combination of special machine speed and specific paper thickness. At this time, the command was sent directly to the control end in milliseconds, and the PLC automatically made a slight adjustment to the main motor speed, successfully avoiding the destructive resonance that was about to occur and effectively ensuring the flatness of the cardboard forming. The ultimate goal of this mechanism is to build a control link from deep digital diagnosis to hardware autonomous correction, and complete intelligent closed-loop control.

[0101] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A fault prediction and health management system for corrugated cardboard processing equipment, characterized in that, include: Data acquisition terminal, baseline generation server, residual analysis module, and decision execution terminal; The data acquisition terminal is used to acquire the production process operation parameters and real-time physical observation data of the target production equipment, and send the production process operation parameters to the baseline generation server and the real-time physical observation data to the residual analysis module. The baseline generation server is configured with a physical information neural network model. The baseline generation server is used to input the production process operating parameters into the physical information neural network model, calculate the dynamic reference baseline data of the target production equipment under the current production process operating parameters, and send the dynamic reference baseline data to the residual analysis module. The residual analysis module is used to calculate the deviation between the real-time physical observation data and the dynamic reference baseline data, determine the deviation as the equipment physical degradation residual, and send the equipment physical degradation residual to the decision execution terminal. The decision execution terminal is used to generate status assessment results and control instructions for the target production equipment based on the physical degradation residuals of the equipment.

2. The fault prediction and health management system for corrugated cardboard processing equipment according to claim 1, characterized in that, The decision execution terminal is also used for: Retrieve the preset degradation threshold from the storage space; The configuration is as follows: if the physical degradation residual of the equipment is greater than or equal to the preset degradation threshold, then it is determined that the target production equipment is abnormal, and a status warning and maintenance suggestion instruction is generated as the control instruction; if the physical degradation residual of the equipment is less than the preset degradation threshold, then it is determined that the target production equipment is not abnormal.

3. The fault prediction and health management system for corrugated cardboard processing equipment according to claim 1, characterized in that, The system also includes: a priori model initialization module; The prior model initialization module is used to obtain the structural topology and physical attribute parameters of the target production equipment, and to convert the structural topology and physical attribute parameters into structured prior parameters and boundary conditions of the physical information neural network model.

4. The fault prediction and health management system for corrugated cardboard processing equipment according to claim 1, characterized in that, The physical degradation residual of the equipment includes the true mechanical stiffness attenuation coefficient or the fluid anomalous accumulation index, which characterizes the actual mechanical stiffness after eliminating production process fluctuations.

5. The fault prediction and health management system for corrugated cardboard processing equipment according to claim 2, characterized in that, The decision execution terminal is also used for: Obtain actual wear data corresponding to the target production equipment where the anomaly exists; The abnormal physical degradation residuals of the equipment and the actual wear conditions are stored as tagged data samples; The structured prior parameters in the physical information neural network model are adaptively calibrated and updated using the labeled data samples.

6. The fault prediction and health management system for corrugated cardboard processing equipment according to claim 1, characterized in that, The production process operating parameters include equipment operating speed, load parameters, and set energy input parameters; the real-time physical observation data includes vibration parameters and temperature parameters.

7. The fault prediction and health management system for corrugated cardboard processing equipment according to claim 3, characterized in that, The prior model initialization module is also used for: Obtain the thermodynamic and kinetic equations; The thermodynamic equation and the kinetic equation are incorporated as penalty terms into the loss function of the physical information neural network model to constrain the generation of the dynamic reference baseline data using the thermodynamic equation and the kinetic equation.

8. The fault prediction and health management system for corrugated cardboard processing equipment according to claim 2, characterized in that, The instructions for generating status alerts and maintenance recommendations include: Based on the physical degradation residual of the equipment, determine the target production process operating parameters corresponding to the anomaly; Recommended production process adjustment parameters are generated as the status warning and maintenance suggestion instructions, wherein the recommended production process adjustment parameters include a suggested reduction in equipment operating speed reference value or a suggested fine adjustment in equipment pressure reference value.

9. The fault prediction and health management system for corrugated cardboard processing equipment according to claim 8, characterized in that, The decision execution terminal is also used for: The status warning and maintenance suggestion instructions are sent to the control terminal of the target production equipment to prompt adjustments to the working status of the target production equipment and to avoid the equipment resonance zone.

Images