A dynamic control method and system for feeding a card shuffler

Abstract

The application belongs to the technical field of industrial equipment control, and discloses a kind of card crusher feed dynamic control method and system, the method includes: obtaining the multidimensional state feature of card crusher, and using automatic feed control strategy generation logic to generate recommended feed control action;Monitoring the actual feed control action that operator orders card crusher to execute, compared with recommended feed control action, to identify the corrective action of operator;When identifying the corrective action of operator, monitor the card crusher equipment operation data in the observation period after executing actual feed control action, to evaluate multidimensional equipment performance improvement information;According to multidimensional state feature, recommended feed control action, actual feed control action and multidimensional equipment performance improvement information, adjust the automatic feed control strategy generation logic, to carry out feed control to card crusher;So as to be able to learn the experience of operator, improve the adaptive ability and performance of card crusher under complex working conditions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial equipment control, in particular to a card breaking machine feeding dynamic control method and system. BACKGROUND

[0002] In the waste plastic recycling production line, the plastic sheet card breaking machine as a key equipment is responsible for breaking the waste plastic sheet to the required size. Its performance and safe operation are closely related to the control of the feeding device. The ideal feeding control should be able to dynamically adjust according to the material characteristics and equipment state, in order to balance efficiency and safety.

[0003] The existing card breaking machine usually adopts an automatic feeding control system based on key parameters such as motor current and main shaft speed. These systems rely on preset logic or simple feedback algorithms, such as current threshold reduction. However, the actual processing of waste plastic sheet material characteristics is complex and variable, with great differences in material, thickness, shape, size, humidity, impurities, etc., resulting in complex changes in resistance and energy demand during the breaking process. The automatic system based on simple parameter feedback is difficult to accurately capture this complex material-equipment interaction, and is prone to control lag or inadequate control under complex conditions, making it difficult to continuously maintain optimal or safe state.

[0004] In actual production, experienced operators can compensate for the shortcomings of automatic systems by observing multi-dimensional signals of the equipment. For example, the operator may be able to predict potential risks through abnormal features in the device sound spectrum or current waveform, even if the key parameters do not reach the threshold, and can perform manual intervention. These intervention strategies are a manifestation of the operator's experience, which can more effectively deal with complex situations.

[0005] However, existing automatic systems can only record the operator's simple speed adjustment and cannot identify and record the more complex intervention behavior of the operator based on multi-dimensional signal judgment, such as temporary pause, mode switching, etc. More importantly, the existing system cannot systematically correlate and learn these operator intervention behaviors with the device multi-modal operating state data at the time of intervention and the multi-dimensional improvement of the device operating state after intervention (such as efficiency, energy consumption, stability). This makes it difficult for the automatic system to learn and reproduce the advanced experience strategies of the operator, limiting the performance ceiling and adaptive ability of the card breaking machine when processing complex materials.

[0006] In view of the above problems, the existing technology needs to be improved. SUMMARY

[0007] The purpose of the present application is to provide a card breaking machine feeding dynamic control method and system that can learn the experience of the operator and improve the adaptive ability and performance of the card breaking machine under complex conditions.

[0008] In a first aspect, the application provides a method for dynamically controlling the feeding of a plastic sheet shredder, comprising the steps of:

[0009] A1. obtaining multi-dimensional device operation signals of the shredder, and extracting multi-dimensional state features therefrom;

[0010] A2. generating a recommended feeding control action using an automated feeding control strategy generation logic based on the multi-dimensional state features;

[0011] A3. monitoring actual feeding control actions performed by the operator on the shredder, and comparing the actual feeding control actions with the recommended feeding control action to identify the corrective behavior of the operator;

[0012] A4. upon identifying the corrective behavior of the operator, monitoring device operation data of the shredder in an observation time period after the actual feeding control action is performed, to evaluate multi-dimensional device performance improvement information;

[0013] A5. generating a corrective feedback signal based on the multi-dimensional state features, the recommended feeding control action, the actual feeding control action, and the multi-dimensional device performance improvement information;

[0014] A6. adjusting the automated feeding control strategy generation logic based on the corrective feedback signal;

[0015] A7. controlling the feeding of the shredder based on the adjusted automated feeding control strategy generation logic.

[0016] Preferably, step A1 comprises:

[0017] A101. obtaining multi-dimensional device operation signals of the shredder; the multi-dimensional device operation signals comprising motor current signals, device sound signals, and device vibration signals;

[0018] A102. performing waveform analysis on the motor current signals to extract impact load features;

[0019] A103. performing frequency spectrum analysis on the device sound signals to extract sound spectrum features;

[0020] A104. performing frequency spectrum analysis on the device vibration signals to extract vibration spectrum features;

[0021] A105. combining the impact load features, the sound spectrum features, and the vibration spectrum features to form the multi-dimensional state features.

[0022] Preferably, step A3 comprises:

[0023] A301. Obtain actual control instructions and parameter settings executed by the operator through the card crusher control interface;

[0024] A302. Analyze the actual control instructions and parameter settings to identify the type and parameters of the actual feeding control action;

[0025] A303. Obtain the type and parameters of the recommended feeding control action;

[0026] A304. Compare the type and parameters of the actual feeding control action with the type and parameters of the recommended feeding control action to determine whether there is a difference between the actual feeding control action and the recommended feeding control action;

[0027] A305. When there is a difference, determine that there is a corrective action, and take the actual feeding control action as the operator's corrective action.

[0028] Preferably, step A4 comprises:

[0029] A401. When the operator's corrective action is identified, determine the observation time period;

[0030] A402. Monitor multi-dimensional equipment operation data of the card crusher within the observation time period; the multi-dimensional equipment operation data includes average motor power, motor current, equipment sound signal energy in a specific frequency range, equipment vibration signal amplitude in a specific frequency range, discharge amount of the card crusher, and discharge particle size distribution;

[0031] A403. According to the multi-dimensional equipment operation data, evaluate multi-dimensional equipment performance improvement information; the multi-dimensional equipment performance improvement information includes efficiency improvement indicators, energy consumption reduction indicators, and stability enhancement indicators.

[0032] Preferably, step A401 comprises:

[0033] Extract state features reflecting current processing material characteristics from the multi-dimensional state features as material characterization features;

[0034] According to the material characterization features and the type of the actual feeding control action, determine the observation time period according to a pre-set observation time period determination rule or an observation time period lookup table.

[0035] Preferably, step A403 comprises:

[0036] B1. According to the multi-dimensional equipment operation data, calculate evaluation values of the efficiency improvement indicators, the energy consumption reduction indicators, and the stability enhancement indicators;

[0037] B2. Identify information reflecting the correction behavior target according to the actual feed control action, denoted as correction intention information;

[0038] B3. Determine the weight coefficients of the efficiency improvement index, the energy consumption reduction index and the stability enhancement index according to the material characterization features and the correction intention information;

[0039] B4. Calculate the comprehensive performance improvement index according to the evaluation values of the efficiency improvement index, the energy consumption reduction index and the stability enhancement index and the corresponding weight coefficients;

[0040] B5. Combine the evaluation values of the efficiency improvement index, the energy consumption reduction index, the stability enhancement index and the comprehensive performance improvement index to form the multi-dimensional equipment performance improvement information.

[0041] Preferably, step B2 comprises:

[0042] Based on a preset intention recognition rule or intention lookup table, the correction intention information is identified according to the type and parameters of the actual feed control action and the multi-dimensional state features.

[0043] Preferably, step A6 comprises:

[0044] A601. Locate the strategy parameters or rules corresponding to the multi-dimensional state features in the automatic feed control strategy generation logic as target strategy parameters or target rules according to the multi-dimensional state features in the correction feedback signal;

[0045] A602. Calculate the feedback intensity of the actual feed control action according to the evaluation value of the efficiency improvement index, the evaluation value of the energy consumption reduction index, the evaluation value of the stability enhancement index and the comprehensive performance improvement index in the multi-dimensional equipment performance improvement information in the correction feedback signal;

[0046] A603. Adjust the target strategy parameters or target rules according to the feedback intensity to increase the probability of generating the actual feed control action in a state similar to the multi-dimensional state features.

[0047] Preferably, step A602 comprises:

[0048] Determine a preliminary feedback intensity according to the comprehensive performance improvement index;

[0049] Determine an intention compliance degree according to the correction intention information, the evaluation value of the efficiency improvement index, the evaluation value of the energy consumption reduction index and the evaluation value of the stability enhancement index;

[0050] The preliminary feedback intensity is corrected according to the intention conformity degree, to obtain a final feedback intensity.

[0051] In a second aspect, the application provides a feeding dynamic control system for a plastic sheet shredder, which is used for controlling the feeding action of the plastic sheet shredder, and the system comprises:

[0052] A state feature acquisition module is configured to acquire multi-dimensional equipment running signals of the plastic sheet shredder and extract multi-dimensional state features from the multi-dimensional equipment running signals.

[0053] A recommended action generation module is configured to generate a recommended feeding control action by using an automatic feeding control strategy generation logic according to the multi-dimensional state features.

[0054] A corrective behavior recognition module is configured to monitor actual feeding control actions performed by the operator on the plastic sheet shredder, compare the actual feeding control actions with the recommended feeding control action, and recognize corrective behaviors of the operator.

[0055] A performance evaluation module is configured to monitor the plastic sheet shredder equipment running data in an observation time period after the actual feeding control actions are performed, when the corrective behaviors of the operator are recognized, and evaluate multi-dimensional equipment performance improvement information.

[0056] A feedback signal generation module is configured to generate a corrective feedback signal according to the multi-dimensional state features, the recommended feeding control action, the actual feeding control action, and the multi-dimensional equipment performance improvement information.

[0057] A strategy adjustment module is configured to adjust the automatic feeding control strategy generation logic according to the corrective feedback signal.

[0058] A control execution module is configured to control the feeding of the plastic sheet shredder according to the adjusted automatic feeding control strategy generation logic.

[0059] Beneficial effects: The feeding dynamic control method and system for the plastic sheet shredder provided by the application can solve the problem that the prior art is difficult to cope with complex materials and learn the experience of the operator by acquiring multi-dimensional equipment signals, recognizing corrective behaviors of the operator, and learning the experience of the operator, and have the advantages of being capable of learning the experience of the operator and improving the self-adaptive ability and performance of the plastic sheet shredder under complex working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 A flowchart of the feeding dynamic control method for the plastic sheet shredder provided by the embodiments of the application.

[0061] Figure 2 A structural schematic diagram of the feeding dynamic control system for the plastic sheet shredder provided by the embodiments of the application.

[0062] Label description: 1, state feature acquisition module; 2, recommended action generation module; 3, corrective behavior identification module; 4, performance evaluation module; 5, feedback signal generation module; 6, strategy adjustment module; 7, control execution module. DETAILED DESCRIPTION

[0063] The technical solutions in the present application will be described clearly and completely below in combination with the drawings in the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. The components of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0064] It should be noted that: similar labels and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present application, the terms "first", "second" and the like are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0065] Reference Figure 1 The present application proposes a dynamic control method for the feeding of a card breaking machine, for controlling the feeding action of a plastic sheet card breaking machine, the steps of the method comprising:

[0066] A1. Obtain multi-dimensional equipment operation signals of the card breaking machine, and extract multi-dimensional state features therefrom;

[0067] A2. According to the multi-dimensional state features, generate a recommended feeding control action using an automatic feeding control strategy generation logic;

[0068] A3. Monitor the actual feeding control action performed by the operator to the card breaking machine, and compare the actual feeding control action with the recommended feeding control action to identify the corrective behavior of the operator;

[0069] A4. When the corrective behavior of the operator is identified, monitor the card breaking machine equipment operation data in an observation time period after the actual feeding control action is performed, to evaluate multi-dimensional equipment performance improvement information;

[0070] A5. According to the multi-dimensional state features, the recommended feeding control action, the actual feeding control action and the multi-dimensional equipment performance improvement information, generate a corrective feedback signal;

[0071] A6. Adjust the automated feed control strategy generation logic according to the correction feedback signal;

[0072] A7. Perform feed control on the card shuffler according to the adjusted automated feed control strategy generation logic.

[0073] Wherein, the multi-dimensional device running signal refers to multiple types of sensor data or monitoring data reflecting the current working state of the card shuffler, which can be obtained by current sensors, sound sensors, vibration sensors, speed sensors, temperature sensors, etc., such as motor current signal, device sound signal, device vibration signal, etc., which is mainly to comprehensively perceive the real-time running state of the card shuffler and the characteristics of the processed materials.

[0074] Wherein, the multi-dimensional state feature refers to the key information extracted from the multi-dimensional device running signal, which can represent the state of the device and the material, which can be realized by signal processing and feature extraction technology, such as waveform analysis of motor current signal to extract impact load characteristics, frequency spectrum analysis of device sound signal and vibration signal to extract frequency spectrum characteristics, which is mainly to convert the original running data into structured information that can be used for control decision and state judgment.

[0075] Wherein, the automated feed control strategy generation logic refers to an algorithm or model used to automatically determine the feed action according to the state of the device, which can be realized by rule-based expert system, fuzzy control algorithm, machine learning model (such as reinforcement learning, neural network), etc., such as outputting recommended feed speed or mode according to pre-set threshold or trained model, which is mainly to realize the automatic feed control of the card shuffler without human intervention.

[0076] Wherein, the recommended feed control action refers to the feed operation suggested by the automated feed control strategy generation logic under the current state, which can take the form of feed speed adjustment, feed mode switching, pause feed, etc., which is mainly to reflect the optimal control suggestion given by the automated system based on the current strategy.

[0077] Wherein, the actual feed control action refers to the feed operation performed by the operator on the card shuffler through the control interface, which can take the form of manual adjustment of feed speed, emergency stop, switching to a specific mode, etc., which is mainly to reflect the operator's intervention based on experience to the automated system control.

[0078] Wherein, the correction behavior refers to the difference between the actual feed control action performed by the operator and the recommended feed control action generated by the automated system, which is mainly to identify the experienced correction of the operator to the automated control.

[0079] The observation time period refers to a period of time for monitoring equipment operation data to evaluate the effect of the actual feeding control action performed by the operator after the operator performs the actual feeding control action, which can be preset or dynamically determined according to material characteristics and correction action types, and is mainly used to obtain data required for evaluating the effect of the operator's intervention.

[0080] The multi-dimensional equipment performance improvement information refers to a quantitative indicator for evaluating the influence of the operator's correction behavior on the performance of the card shuffler, which can take the form of an efficiency improvement indicator, an energy consumption reduction indicator, a stability enhancement indicator, etc., and is mainly used to quantify the effectiveness of the operator's experiential intervention.

[0081] The correction feedback signal refers to information integrating the situation when the operator's correction behavior occurs, the behavior itself, and the effect thereof, and is mainly used to provide a basis for adjusting the automatic feeding control strategy generation logic.

[0082] The core innovation of the present application is that by monitoring the actual feeding control action performed by the operator in addition to the recommended action of the automatic system, identifying the operator's correction behavior, and monitoring the equipment performance improvement information after the behavior is performed, the experiential intervention of the operator and its effect are converted into a correction feedback signal, so as to dynamically adjust the automatic feeding control strategy generation logic using the feedback signal, thereby solving the problem that the existing system cannot learn the operator's experience and is difficult to cope with complex materials, and achieving the effect of improving the performance and self-adaptation ability of the card shuffler.

[0083] Specifically, the method first acquires multi-dimensional equipment operation signals of the card shuffler, and extracts multi-dimensional state features therefrom to comprehensively perceive the device and material states. Then, according to these state features, a recommended feeding control action is generated using the current automatic feeding control strategy generation logic. At the same time, the system monitors the actual feeding control action performed by the operator. By comparing the actual action of the operator with the action recommended by the automatic system, the operator's correction behavior is identified. When the correction behavior is identified, the system monitors the equipment operation data of the card shuffler within an observation time period after the actual action is performed, and evaluates the multi-dimensional equipment performance improvement information brought by the correction behavior based on these data. Then, the identified multi-dimensional state features, the recommended feeding control action, the actual feeding control action of the operator, and the evaluated multi-dimensional equipment performance improvement information are integrated into a correction feedback signal. Finally, the automatic feeding control strategy generation logic is adjusted according to the correction feedback signal. The adjusted strategy generation logic will be used for subsequent feeding control of the card shuffler. The entire process forms a closed loop, so that the automatic system can continuously learn the experience of the operator under complex working conditions and integrate it into its own control strategy.

[0084] By the above scheme, the application can identify and learn the experiential feeding control strategy of the operator in the complex and changeable material processing process, convert the experience of the operator into the ability of the automatic system, thereby overcoming the limitation that the existing automatic system is difficult to cope with complex working conditions, improving the processing efficiency, energy consumption performance and operation stability of the card breaking machine in processing complex materials, and enhancing the self-adaptation ability of the equipment.

[0085] In some embodiments, step A1 comprises:

[0086] A101. Obtain a multi-dimensional equipment operation signal of the card breaking machine; the multi-dimensional equipment operation signal comprises a motor current signal, an equipment sound signal and an equipment vibration signal;

[0087] A102. Perform waveform analysis on the motor current signal to extract an impact load feature;

[0088] A103. Perform frequency spectrum analysis on the equipment sound signal to extract a sound spectrum feature;

[0089] A104. Perform frequency spectrum analysis on the equipment vibration signal to extract a vibration spectrum feature;

[0090] A105. Combine the impact load feature, the sound spectrum feature and the vibration spectrum feature to form the multi-dimensional state feature.

[0091] Wherein, waveform analysis refers to a technology for researching and processing the shape of signal change over time, which can be realized by time domain analysis, feature point detection, instantaneous value calculation and the like. The impact load feature refers to a value or mode extracted from the waveform of the motor current signal, which can represent the state of the cutter or equipment bearing instantaneous impact or high stress, which can be represented by current peak amplitude, current change rate, current integral in a certain time window and the like.

[0092] Wherein, frequency spectrum analysis refers to a technology for decomposing a signal into different frequency components, which can be realized by fast Fourier transform (FFT), short-time Fourier transform (STFT), wavelet analysis and the like. The sound spectrum feature refers to a value or mode extracted from the frequency spectrum of the equipment sound signal, which can represent the energy distribution or specific frequency component of the equipment sound at different frequencies, which can be represented by specific frequency band energy, resonance peak frequency, harmonic content and the like. The vibration spectrum feature refers to a value or mode extracted from the frequency spectrum of the equipment vibration signal, which can represent the energy distribution or specific frequency component of the equipment vibration at different frequencies, which can be represented by specific frequency band vibration amplitude, main vibration frequency, harmonic vibration intensity and the like.

[0093] The multi-dimensional state feature refers to a numerical value or vector set formed by combining features of different dimensions, which can comprehensively represent the current running state of the card crusher.

[0094] Based on the above technical features, the method of the present application achieves its functions and solves the technical problems in the following way. First, by obtaining the motor current signal, the equipment sound signal and the equipment vibration signal, which are three dimensions of equipment running signals, the method breaks through the limitation of relying on only a single or simple signal, providing basic data for comprehensively sensing the running state of the card crusher. Then, the motor current signal is analyzed to extract the impact load feature, which can capture the instantaneous stress change during the material entering or breaking process; the equipment sound signal is analyzed to extract the sound spectrum feature, which can reflect the auditory information of the interaction between the cutter and the material and the internal mechanical state of the equipment; the equipment vibration signal is analyzed to extract the vibration spectrum feature, which can sense the mechanical stability, balance and vibration response caused by material impact of the equipment. Through these refined analysis methods, the features extracted from different dimensions can more accurately and meticulously represent the real working conditions of the card crusher when processing complex and variable materials, including the hardness, toughness, shape of the material and potential problems such as wear and loosening of the equipment cutter. Finally, these impact load features, sound spectrum features and vibration spectrum features are combined to form a multi-dimensional state feature vector. This multi-dimensional state feature vector contains more rich and detailed state information than traditional methods. When this more detailed multi-dimensional state feature is applied to the subsequent automatic feeding control strategy generation, operator corrective behavior identification, performance evaluation and strategy adjustment process, the automatic system can more accurately understand the complex working conditions of the current equipment, identify the corrective behavior of the operator in a specific subtle state and the performance improvement it brings, so as to more effectively learn the experience strategy of the operator in handling complex materials. This learning and strategy adjustment mechanism based on multi-dimensional and refined state features enables the automatic feeding control strategy to better adapt to complex and variable material properties and equipment working conditions, overcoming the control lag and poor control problems caused by insufficient state information in traditional systems, thereby improving the overall performance and self-adaptation ability of the card crusher when processing complex materials.

[0095] In some embodiments, step A3 comprises:

[0096] A301. Obtain the actual control instructions and parameter settings executed by the operator through the card crusher control interface;

[0097] A302. Analyze the actual control instructions and parameter settings to identify the type and parameters of the actual feeding control action;

[0098] A303. Obtain the type and parameters of the recommended feeding control action;

[0099] A304. Comparing the type and parameter of the actual feeding control action with the type and parameter of the recommended feeding control action to determine whether there is a difference between the actual feeding control action and the recommended feeding control action;

[0100] A305. When there is a difference, determining that there is a corrective action, and taking the actual feeding control action as the operator's corrective action.

[0101] In the above steps, the actual control instructions and parameter settings executed by the operator through the card crusher control interface are obtained, which can be any user interface or input device used by the operator to directly intervene in the card crusher feeding process, such as a physical control panel, a touch screen interface, a remote monitoring terminal, or a specific software application. The actual control instructions and parameter settings refer to the original operation commands and related values or options input by the operator through these interfaces, such as "pause" instructions, values for setting feeding speed, options for selecting specific feeding modes, etc.

[0102] Among them, the actual control instructions and parameter settings are analyzed to identify the type and parameter of the actual feeding control action. The analysis process is to convert the original and diversified instructions and parameters input by the operator into a standardized representation that the system can understand and process. The type of actual feeding control action can include but is not limited to start, stop, pause, resume, set speed, adjust torque limit, switch feeding mode, etc. The parameters can be specific values related to these action types, such as the set speed value, the torque limit value, the selected mode identifier, etc. The type and parameter of the recommended feeding control action are obtained, which is generated by the automatic feeding control strategy generation logic according to the current state of the card crusher, and its representation form corresponds to that of the actual feeding control action to facilitate comparison.

[0103] Among them, the type and parameter of the actual feeding control action are compared with the type and parameter of the recommended feeding control action to determine whether there is a difference between the actual feeding control action and the recommended feeding control action. The comparison process can include comparing whether the action types are consistent, and when the action types are consistent, comparing whether the related parameter values are within the pre-set allowed deviation range. When there is a difference, it is determined that there is a corrective action, and the actual feeding control action is taken as the operator's corrective action, which is based on the comparison result. Any inconsistency in type or parameter is considered as the operator's intervention on the automation recommendation, i.e. corrective action. Taking the actual feeding control action itself as the corrective action means that the subsequent learning process will directly use the specific operation actually executed by the operator as the learning sample.

[0104] By the above steps, the method of the present application can realize accurate identification of the operator's corrective behavior. Specifically, first, by obtaining the operator's original input on the control interface, it ensures that the detailed operation intention and specific value of the operator are captured. Then, the original input is parsed and standardized into action types and parameter representations that the system can understand. At the same time, the types and parameters of the recommended actions generated by the automation system are obtained. Then, the actual action of the operator is accurately compared with the recommended action, whether the action types are different or the parameter values are different, it can be identified. Once the difference is found, it is determined that the operator's corrective behavior, and the actual action performed by the operator is taken as the identification result. This method not only can identify simple speed adjustment, but also can capture more complex intervention behavior such as pause and mode switching performed by the operator. Compared with the existing method of only monitoring the change of device state or simple speed adjustment, the present method can more comprehensively and accurately reflect the real operation intention and specific strategy of the operator by directly obtaining and parsing the original control instructions and parameters of the operator. The accurate identification of the detailed intervention behavior of the operator provides high-quality input data for the subsequent system to learn the experience of the operator, so as to more effectively optimize the automation control strategy and improve the performance and adaptive ability of the card shuffler in handling diversified and complex materials.

[0105] In some embodiments, step A4 comprises:

[0106] A401. Upon identifying the corrective behavior of the operator, determining the observation time period;

[0107] A402. During the observation time period, monitoring multi-dimensional device running data of the card shuffler; the multi-dimensional device running data comprises motor average power, motor current, device sound signal energy in a specific frequency range, device vibration signal amplitude in a specific frequency range, discharge amount of the card shuffler, and discharge granularity distribution;

[0108] A403. According to the multi-dimensional device running data, evaluating multi-dimensional device performance improvement information; the multi-dimensional device performance improvement information comprises efficiency improvement index, energy consumption reduction index, and stability enhancement index.

[0109] Wherein, the observation time period can be determined according to the characteristics of the current processing material, the type of the operator's corrective action, or the change of the device state, etc. according to the pre-set rules or lookup table.

[0110] Wherein, the multi-dimensional device running data refers to a set of signals or parameters collected from the card shuffler, which can reflect the running state and effect of the device in multiple aspects, which can include electrical parameters reflecting motor load and stability, acoustic and vibration parameters reflecting cutter interaction with material or potential abnormalities, discharge parameters reflecting processing capacity and reflecting breaking quality.

[0111] wherein the equipment sound signal energy in a specific frequency range refers to the total energy or average energy of the equipment running sound signal in a preset frequency interval (the specific frequency interval can be determined in advance by data statistics) related to a specific physical phenomenon (such as tool impact, material friction) after frequency spectrum analysis, which can be obtained by Fourier transform or other frequency spectrum analysis method on the original sound signal and integrating or averaging in the specified frequency range.

[0112] wherein the equipment vibration signal amplitude in a specific frequency range refers to the vibration amplitude peak value or root mean square value of the equipment running vibration signal in a preset frequency interval (the specific frequency interval can be determined in advance by data statistics) related to a specific mechanical state (such as bearing wear, imbalance) after frequency spectrum analysis, which can be obtained by frequency spectrum analysis on the original vibration signal and extracting characteristic values in the specified frequency range.

[0113] wherein the discharge particle size distribution refers to the distribution of the particle size of the material after crushing by the card crusher, which can be measured and characterized by sieve analysis, image recognition or laser particle size instrument, etc.

[0114] wherein the efficiency improvement indicator refers to an indicator measuring the change degree of the processing capacity or effective output of the card crusher after the intervention of the operator relative to before the intervention. For example, the discharge efficiency in the observation period can be obtained by dividing the discharge amount in the observation period by the time length of the observation period, and then the increase amount or increase proportion of the discharge efficiency after the intervention relative to the discharge efficiency before the intervention is calculated as the efficiency improvement indicator.

[0115] wherein the energy consumption reduction indicator refers to an indicator measuring the change degree of the energy consumption per unit processing amount or per unit time of the card crusher after the intervention of the operator relative to before the intervention. For example, the total energy consumption can be obtained by multiplying the average power of the motor by the time length of the observation period, and then the unit discharge amount energy consumption is obtained by dividing the total energy consumption by the discharge amount, and then the reduction amount or reduction proportion of the unit discharge amount energy consumption after the intervention relative to the unit discharge amount energy consumption before the intervention is calculated as the energy consumption reduction indicator.

[0116] wherein the stability enhancement indicator refers to an indicator measuring the change degree of the running state stability of the card crusher after the intervention of the operator relative to before the intervention. The stability enhancement indicator can be a multi-dimensional indicator, which can include the motor current standard deviation (calculated according to the motor current), the coefficient of variation of the sound energy integral value (calculated according to the equipment sound signal energy in a specific frequency range), the coefficient of variation of the vibration amplitude peak value (calculated according to the equipment vibration signal amplitude in a specific frequency range), and the standard deviation of the discharge particle size distribution (calculated according to the discharge particle size distribution), and the stability enhancement indicator can also be the weighted average of the normalized values of these indicators.

[0117] The present scheme initiates a performance evaluation process after identifying the operator's corrective behavior. First, a suitable observation time period is determined, which sets a clear time limit for subsequent data collection and effectiveness evaluation, ensuring that the evaluation results are closely related to the operator's intervention behavior. During this observation time period, the system monitors multi-dimensional equipment operation data of the card shuffler. These data not only include traditional motor average power, current and discharge volume, but also introduce more detailed and comprehensive parameters such as sound energy and vibration signal amplitude in a specific frequency range, and discharge particle size distribution. These multi-dimensional data can capture the real running state and material processing effect of the equipment under complex working conditions in detail, reflecting the subtle changes that may be brought by the operator's intervention. Subsequently, according to these multi-dimensional equipment operation data, multi-dimensional equipment performance improvement information is evaluated, including efficiency improvement indicators, energy consumption reduction indicators and stability enhancement indicators. This multi-dimensional evaluation method can quantify the specific contributions of the operator's intervention in different aspects, for example, the operator's adjustment may maintain high efficiency while reducing energy consumption, or significantly improve stability at the expense of a small amount of efficiency. By correlating the operator's corrective behavior, the equipment state at the time of the behavior (reflected by the multi-dimensional state features obtained in the previous step), and the multi-dimensional performance improvement information after the behavior, the present scheme can generate high-quality corrective feedback signals. The feedback signal contains the essence of the operator's experience and can be used to adjust the automatic feeding control strategy generation logic, enabling the automatic system to learn and reproduce the effective intervention strategy of the operator under specific complex working conditions, thereby improving the adaptive ability and overall performance of the automatic control. This mechanism of converting operator experience into learnable, multi-dimensional quantitative feedback information and optimizing automatic strategies is the key to solving the problems of the prior art.

[0118] The present scheme can comprehensively and accurately capture the impact of the operator's corrective behavior on the running performance of the card shuffler by monitoring multi-dimensional equipment operation data and evaluating multi-dimensional equipment performance improvement information based on these data. Specifically, by monitoring multi-dimensional data such as motor average power, motor current, equipment sound signal energy in a specific frequency range, equipment vibration signal amplitude in a specific frequency range, discharge volume of the card shuffler, and discharge particle size distribution, the actual running state of the equipment after the operator's intervention can be understood in detail. Based on these data, multi-dimensional indicators such as efficiency improvement indicators, energy consumption reduction indicators and stability enhancement indicators can be evaluated to quantify the contributions of the operator's intervention in different aspects. This multi-dimensional and detailed evaluation overcomes the shortcomings of traditional methods that rely on limited parameters, enabling a more accurate understanding of the value of the operator's experience strategy and providing high-quality feedback information for subsequent learning and adjustment of automatic strategies, thereby improving the adaptive ability and overall performance of the card shuffler in processing complex materials.

[0119] Preferably, step A401 can include:

[0120] extracting, from the multi-dimensional state features, a state feature reflecting a current processing material characteristic as a material characterization feature;

[0121] determining, according to the material characterization feature and the type of actual feeding control action, the observation time period according to a preset observation time period determination rule or an observation time period lookup table.

[0122] wherein the material characterization feature refers to a state feature closely related to the physical characteristics (such as hardness, toughness, humidity, size, etc.) of the material being processed, which is selected or extracted from the multi-dimensional state features and can be determined from the multi-dimensional state features by using principal component analysis, feature selection algorithm or expert knowledge-based feature combination, etc.

[0123] wherein the preset observation time period determination rule refers to a logic or algorithm for calculating or deriving the corresponding observation time period according to the input material characterization feature and the type of actual feeding control action, which is established based on historical data analysis, machine learning model training or domain expert experience, and can be implemented by using decision tree, regression model or fuzzy logic-based reasoning system. The observation time period lookup table refers to a data structure that is pre-established and stores the recommended observation time period corresponding to different combinations of material characterization features (or their classifications) and different types of actual feeding control actions, which can be implemented by using two-dimensional or multi-dimensional array, hash table or database table. Determining the observation time period refers to calculating or obtaining a specific time length as a time window for monitoring the subsequent equipment operation data by applying the preset determination rule or querying the preset lookup table according to the current material characterization feature and the type of actual feeding control action.

[0124] The scheme is not simply fixed or only based on the observation period of action type after identifying the correction behavior of the operator, but first identifies and extracts the state features closely related to the current processing material characteristics from the multi-dimensional state features reflecting the real-time state of the equipment, and takes these features as material characterization features, thereby obtaining the cognition of the material state under the current working condition. Subsequently, combined with the specific type of actual feeding control action performed by the operator, the system determines the observation time period according to the pre-established observation time period determination rules or queries the pre-constructed observation time period lookup table. These rules or lookup tables associate different material characteristics (reflected by material characterization features) and different correction action types with the corresponding effect appearing time. For example, for brittle and dry materials, the operator's slight deceleration can quickly see the effect of reduced current fluctuations or stable sound, at which time the rules or lookup tables will determine a shorter observation time period; while for strong, wet or impurity-containing materials, or the operator performs actions such as reversing or long pauses, the stability or improvement of the equipment state may take longer to reflect, at which time the rules or lookup tables will determine a longer observation time period. In this way, the determination process of the observation time period fully considers the actual characteristics of the current processing material and the nature of the operator's correction action, so that the determined time period can more accurately match the real appearing period of the influence of the correction behavior on the equipment performance. Limiting subsequent equipment operation data monitoring within such a more targeted observation time period can collect data that more accurately reflects the actual effect of the correction behavior, thereby laying a solid foundation for subsequent accurate evaluation of multi-dimensional equipment performance improvement information. This dynamic and adaptive observation time period determination mechanism based on material characteristics and action type significantly improves the accuracy and effectiveness of operator experience learning, enabling the system to more accurately understand the reasons for the operator's specific actions under specific complex working conditions and the actual effects brought about by the actions.

[0125] Preferably, step A403 can include:

[0126] B1. Calculate the evaluation values of the efficiency improvement index, the energy consumption reduction index and the stability enhancement index according to the multi-dimensional equipment operation data;

[0127] B2. Identify information reflecting the correction behavior target according to the actual feeding control action, denoted as correction intention information;

[0128] B3. Determine the weight coefficients of the efficiency improvement index, the energy consumption reduction index and the stability enhancement index according to the material characterization features and the correction intention information;

[0129] B4. Calculate a comprehensive performance improvement index according to the evaluation values of the efficiency improvement index, the energy consumption reduction index and the stability enhancement index and the corresponding weight coefficients;

[0130] B5. The evaluation values of the efficiency improvement indicator, the energy consumption reduction indicator, and the stability enhancement indicator are combined to form the multi-dimensional device performance improvement information.

[0131] The calculation method of the evaluation values of the efficiency improvement indicator, the energy consumption reduction indicator, and the stability enhancement indicator can refer to the foregoing.

[0132] The correction intention information refers to the potential purpose or goal of the operator's intervention behavior inferred by analyzing the actual feed control action of the operator, which can be represented by a pre-set intention category (e.g., improving efficiency, reducing energy consumption, enhancing stability, avoiding jamming, optimizing discharge, etc.) or a more refined intention description.

[0133] The comprehensive performance improvement indicator refers to a holistic evaluation value calculated by combining the evaluation values of each performance indicator with their corresponding weight coefficients, which can be calculated using weighted summation, weighted average, or other forms of aggregation functions.

[0134] The scheme defines the specific process of evaluating multi-dimensional equipment performance improvement information in detail, aiming to more accurately and intelligently evaluate the effect of operator's corrective behavior, thereby providing high-quality feedback signals for the learning of automation strategies. Specifically, according to the multi-dimensional equipment operation data, the evaluation values of the efficiency improvement index, the energy consumption reduction index and the stability enhancement index are calculated. This is the basis for quantitative evaluation of the effect of operator's corrective behavior, directly reflecting the performance of the equipment after intervention in these key performance dimensions. According to the multi-dimensional equipment operation data, the evaluation results are ensured to be objective and data-supported. At the same time, the information reflecting the target of corrective behavior is identified according to the actual feed control action, which is recorded as the corrective intention information. This step is the key to understanding the operator's experience. By analyzing the specific control action type and parameters performed by the operator, the system attempts to infer the potential purpose of the operator's intervention, such as improving processing efficiency, reducing energy consumption or enhancing the stability of equipment operation. Identifying the corrective intention information enables the subsequent performance evaluation to be more targeted, understanding which performance goal the operator values more in a specific situation. On this basis, according to the material characterization features and the corrective intention information, the weight coefficients of the efficiency improvement index, the energy consumption reduction index and the stability enhancement index are determined. This step further improves the intelligent level of evaluation. It recognizes that the importance of each performance index is different when processing materials with different characteristics, and when the operator has different corrective intentions. For example, when processing materials prone to jamming, stability may be more critical; if the operator's intention is to solve the overload problem, the weights of energy consumption reduction and stability enhancement may be higher. Specifically, the weight coefficients of each performance index can be determined by consulting a pre-set weight lookup table or applying weight calculation rules according to the identified corrective intention information and material characterization features. Adjusting the weights of each performance index dynamically according to the material characterization features (reflecting the characteristics of the current processed material) and the corrective intention information makes the performance evaluation more consistent with the actual working conditions and the operator's experience, avoiding the shortcomings of simple averaging or fixed weights. Subsequently, the comprehensive performance improvement index is calculated according to the evaluation values of the efficiency improvement index, the energy consumption reduction index and the stability enhancement index and the corresponding weight coefficients. This step combines the evaluation values of each performance index calculated in the previous steps with the weights determined according to the material characteristics and the operator's intention to calculate a comprehensive evaluation index. This comprehensive index can more comprehensively and accurately reflect the overall effect of the operator's corrective behavior, especially when different performance goals need to be weighed. Calculating the comprehensive performance improvement index provides a quantitative, context and intention considering feedback signal for the subsequent learning of automation strategies. Finally, the evaluation values of the efficiency improvement index, the energy consumption reduction index, the stability enhancement index and the comprehensive performance improvement index are combined to form the multi-dimensional equipment performance improvement information. This step integrates the evaluation values and the comprehensive index calculated in the previous steps into a structured information package.The multi-dimensional equipment performance improvement information not only contains the evaluation of each original performance, but also contains the comprehensive evaluation considering the material and intention weight, providing rich and valuable feedback information for subsequent generation of correction feedback signals and adjustment of automation strategy, so that the automation system can learn the advanced experience of the operator more effectively.

[0135] By integrating this more accurate and more contextually aware performance evaluation information into the adjustment process of the automation feed control strategy, the system can better understand the decision-making logic of the operator under complex working conditions, thereby learning and replicating these advanced experiences, improving the adaptability and robustness of the automation strategy, and ultimately optimizing the overall performance of the card shuffler when handling diversified materials.

[0136] Preferably, step B2 can include:

[0137] Based on the pre-set intention recognition rules or intention lookup table, the correction intention information is identified according to the type and parameters of the actual feed control action and the multi-dimensional state characteristics.

[0138] Among them, the pre-set intention recognition rules or intention lookup table refers to a pre-established knowledge structure for mapping a specific input combination to correction intention information. The rules can be a series of conditional-action statements, and the lookup table can be a multi-dimensional mapping table. They are constructed based on the analysis of operator experience and equipment behavior.

[0139] The scheme defines a specific implementation of identifying the correction intention information, aiming to solve the problem that it is difficult to accurately determine the real intention of the operator only according to the actual feeding control action of the operator. By combining the multi-dimensional state features of the equipment when the operator performs the action, the purpose of the operator's correction behavior can be more accurately inferred. After identifying the correction behavior of the operator, the system obtains the multi-dimensional state features at the time when the behavior occurs. The type and parameter of the actual feeding control action are combined with the multi-dimensional state features as input. Based on the pre-set intention recognition rule or intention lookup table, the input is matched or reasoned, so as to identify the real correction intention information of the operator performing the action. For example, if the operator reduces the feeding speed when a specific abnormal frequency appears in the vibration spectrum of the equipment, the intention recognition rule may identify it as the intention of processing the jam or protecting the cutter. This more accurate intention recognition enables the subsequent determination of the performance evaluation weight according to the intention to more accurately reflect the key performance indicators that the operator focuses on under specific working conditions. For example, if the intention is identified as protecting the equipment, the weight of the stability enhancement indicator may be increased when evaluating the performance improvement; if it is identified as improving efficiency, the weight of the efficiency improvement indicator may be increased. In this way, the scheme makes the performance improvement evaluation of the operator's correction behavior more in line with the real purpose and actual effect of the operator, thereby providing more accurate and more meaningful feedback signals for subsequent strategy adjustment, and ultimately improving the self-adaptation ability and performance of the automatic feeding control strategy under complex working conditions.

[0140] In one embodiment, an intention lookup table can be constructed, the input of which includes the actual feeding action type, the speed adjustment parameter, the impact load feature, the sound spectrum feature, and the vibration spectrum feature. The lookup table can contain multiple entries, for example, one entry can map the combination of "speed adjustment" action, 20% speed reduction parameter, high impact load feature, normal sound spectrum feature, and normal vibration spectrum feature to the correction intention information of "protecting the equipment". Another entry can map the combination of "speed adjustment" action, 10% speed reduction parameter, medium impact load feature, specific high frequency enhanced sound spectrum feature, and normal vibration spectrum feature to the correction intention information of "processing hard points of material". For another example, an entry can map the combination of "pause" action, no applicable parameter, medium impact load feature, normal sound spectrum feature, and specific low frequency enhanced vibration spectrum feature to the correction intention information of "processing jam". In actual application, when the operator performs an action, the system obtains the action type, parameter, and multi-dimensional state features at that time, then queries the lookup table to find the matching entry, and outputs the corresponding correction intention information.

[0141] In some embodiments, step A6 comprises:

[0142] A601. According to the multi-dimensional state feature in the corrective feedback signal, locate the strategy parameter or rule in the automated feed control strategy generation logic corresponding to the multi-dimensional state feature as the target strategy parameter or target rule;

[0143] A602. According to the evaluation value of the efficiency improvement indicator, the evaluation value of the energy consumption reduction indicator, the evaluation value of the stability enhancement indicator, and the comprehensive performance improvement indicator in the multi-dimensional device performance improvement information in the corrective feedback signal, calculate the feedback intensity of the actual feed control action;

[0144] A603. According to the feedback intensity, adjust the target strategy parameter or target rule to increase the probability of generating the actual feed control action in a similar state to the multi-dimensional state feature.

[0145] Wherein, the strategy parameter or rule of the multi-dimensional state feature refers to the adjustable part of the automated feed control strategy generation logic, for example, in a rule-based system, it can be the condition, priority or associated action of the rule; in a machine learning model, it can be the weight, bias or other hyperparameters of the model. The target strategy parameter or target rule refers to the strategy parameter or rule that needs to be adjusted according to the current multi-dimensional state feature in a specific strategy adjustment process.

[0146] Wherein, the feedback intensity refers to a quantitative value calculated according to the degree of improvement of the device performance brought by the operator's actual feed control action, which is used to measure the effectiveness and importance of the operator's intervention, which can be calculated by weighted combination of performance improvement indicators, mapping based on lookup table or through learning algorithm.

[0147] Wherein, adjusting the target strategy parameter or target rule to increase the probability of generating the actual feed control action in a similar state to the multi-dimensional state feature refers to modifying the corresponding part of the automated feed control strategy generation logic, so that when the automated system encounters a similar device state in the future as the recorded operator intervention, it is more likely to generate the same or similar recommended action as the operator's actual corrective action, which can be achieved by modifying the trigger condition or priority of the rule, adjusting the weight or bias of the model output layer, or updating the policy function in reinforcement learning.

[0148] By the combination of the above technical means, the method of the present application can realize the following working principle: first, when the system identifies that the operator corrects the intervention of the automatic recommended action and generates a correction feedback signal containing multi-dimensional state features, actual actions and performance improvement information, the method uses the multi-dimensional state features in the feedback signal to accurately find the strategy parameters or rules most relevant to the specific state in the complex automatic feeding control strategy generation logic. This ensures that the adjustment of the strategy is made for the specific working conditions where the operator actually intervenes, improving the efficiency and accuracy of learning. Then, according to the multi-dimensional device performance improvement information contained in the feedback signal, especially the evaluation values of the efficiency, energy consumption, stability and comprehensive performance indicators quantifying the effect of operator intervention, the method calculates the feedback strength of the operator's actual feeding control action. The more significant the performance improvement, the greater the calculated feedback strength, indicating that the operator's experience is more valuable. Finally, the method adjusts the target strategy parameters or target rules located in the front according to the calculated feedback strength. The adjustment range is related to the feedback strength, and the greater the feedback strength, the greater the adjustment range. The goal of adjustment is to increase the probability of the same recommended action as the operator's actual correction action when the automatic strategy generation encounters similar working conditions as the recorded multi-dimensional state features in the future. This probabilistic adjustment mechanism based on state similarity and performance improvement degree enables the automatic strategy to selectively and selectively absorb the successful experience of the operator, gradually improving the self-adaptation ability and control performance in handling complex materials and dealing with complex working conditions.

[0149] The method makes full use of the rich and detailed feedback information provided by the previous steps (acquiring multi-dimensional state features, identifying correction behavior, evaluating multi-dimensional performance improvement, and generating correction feedback signals), so that the strategy adjustment is no longer a simple and blind attempt, but based on a deep understanding of the specific situation and behavior effect when the operator's behavior occurs, so as to more effectively learn the wisdom of the operator's experience and improve the intelligent level of automatic control.

[0150] To illustrate the method of the present application more specifically, reference can be made to the following example: Assume that the automated feed control strategy generation logic is a rule-based expert system, containing a series of feed control rules triggered according to equipment state. When the system identifies that an operator has performed a corrective action under a certain specific multi-dimensional state feature, and that the action has brought about significant multi-dimensional equipment performance improvement, a corrective feedback signal is generated. Step A601 can search in the rule base of the expert system for the rule or rule set that matches the multi-dimensional state feature in the feedback signal most closely, and determine these rules as the target rules. For example, if the state feature indicates that there is a risk of entanglement of the material, and the operator has performed the action of a short pause in feeding, the system can locate the feed rules related to the "high entanglement risk" state. Step A602 can calculate the feedback intensity of the pause in feeding action this time according to the evaluation values of the efficiency improvement, energy consumption reduction, stability enhancement, and comprehensive performance improvement indicators in the feedback signal, through a pre-set function or lookup table. For example, if the motor current fluctuation is significantly reduced after the pause, the sound spectrum returns to normal, and the comprehensive performance indicator is greatly improved, the calculated feedback intensity will be high. Step A603 can adjust the target rules located by step A601 according to the calculated high feedback intensity. This adjustment can be through increasing the trigger priority of the target rules under similar states, or modifying the conditions of the rules to make them more easily triggered under similar states, or associating a higher probability weight for the rules, so as to increase the probability of the automated system generating the pause in feeding recommended action when encountering similar entanglement risk states in the future.

[0151] By adopting the method of the present application, the following technical effects can be obtained: By accurately locating the strategy adjustment target according to the multi-dimensional state feature in the corrective feedback signal, invalid adjustments to irrelevant parts of the strategy are avoided, making the strategy learning targeted. By calculating the feedback intensity according to the multi-dimensional equipment performance improvement information, the value of the operator's intervention is quantified, ensuring that the strategy learning is based on the operator's behavior that has a positive impact on the equipment performance. By adjusting the target strategy parameters or rules according to the feedback intensity, the automated strategy can learn the operator's successful experience with emphasis and probability, improving the self-adaptation ability and control performance when dealing with complex materials and complex working conditions.

[0152] Preferably, step A602 can include:

[0153] determining a preliminary feedback intensity according to the comprehensive performance improvement indicator;

[0154] determining an intention compliance degree according to the corrective intention information, the evaluation value of the efficiency improvement indicator, the evaluation value of the energy consumption reduction indicator, and the evaluation value of the stability enhancement indicator;

[0155] The preliminary feedback intensity is corrected according to the intention conformity to obtain a final feedback intensity.

[0156] The preliminary feedback intensity refers to an initial feedback amount calculated based on the overall performance improvement of the device, which can be achieved by converting the comprehensive performance improvement index into a basic intensity value through a mapping function (such as a linear function or a nonlinear function).

[0157] The intention conformity refers to the matching degree between the performance improvements brought by the actual operation of the operator and the correction intention of the operator, which can be achieved by calculating the similarity or distance between the actual performance evaluation vector and the reference performance improvement vector representing the intention of the operator (the reference performance improvement vector corresponding to each correction intention information can be determined in advance).

[0158] The final feedback intensity refers to the feedback amount adjusted by the intention conformity, which is used to finally guide the adjustment of the automated strategy, which can be achieved by multiplying or adding the preliminary feedback intensity and the intention conformity.

[0159] In the calculation of the feedback intensity of the actual feed control action, a preliminary feedback intensity is first calculated based on the overall performance improvement of the device observed after the correction behavior of the operator, which provides a basis for feedback based on actual effect. On this basis, the correction intention information of the operator is introduced, and the actual observed evaluation values of each specific performance (efficiency, energy consumption, stability) are combined to calculate an intention conformity. This intention conformity quantifies the degree to which the actual operation of the operator achieves its specific goal. For example, if the intention of the operator is to improve stability, and the stability is significantly improved after the actual operation, the intention conformity is high. Finally, the preliminary feedback intensity is corrected using the calculated intention conformity. If the intention conformity is high, it means that the correction behavior of the operator is very effective in achieving its specific goal, and the feedback intensity can be appropriately enhanced so that the automated strategy is more likely to learn and adopt such behavior in similar situations. On the contrary, if the intention conformity is low, even if the comprehensive performance index is acceptable, it may mean that the behavior is not the best way the operator expects, and the feedback intensity can be appropriately reduced to avoid the strategy from learning and adopting the behavior that is inconsistent with the intention of the operator. By combining the correction intention information of the operator with the actual performance improvement and correcting the feedback intensity, the present scheme can more accurately quantify the value of the correction behavior of the operator, so that the automated feed control strategy generation logic can more effectively learn the experience preference of the operator in a specific situation based on a specific goal. This feedback mechanism combined with the intention of the operator makes the strategy adjustment process more intelligent and targeted, and can better adapt to the changing and complex material characteristics and working conditions, thereby improving the adaptability and optimization effect of the automated system.

[0160] Reference Figure 2 The application provides a feeding dynamic control system for a plastic sheet shredder, which comprises:

[0161] a state feature acquisition module 1 for acquiring multi-dimensional device running signals of the shredder and extracting multi-dimensional state features therefrom (the specific process can refer to step A1 in the foregoing);

[0162] a recommended action generation module 2 for generating a recommended feeding control action by using an automatic feeding control strategy generation logic according to the multi-dimensional state features (the specific process can refer to step A2 in the foregoing);

[0163] a corrective behavior identification module 3 for monitoring actual feeding control actions performed by the operator on the shredder, comparing the actual feeding control actions with the recommended feeding control action, and identifying corrective behaviors of the operator (the specific process can refer to step A3 in the foregoing);

[0164] a performance evaluation module 4 for monitoring device running data of the shredder in an observation period after the actual feeding control actions are performed when the corrective behaviors of the operator are identified, and evaluating multi-dimensional device performance improvement information (the specific process can refer to step A4 in the foregoing);

[0165] a feedback signal generation module 5 for generating a corrective feedback signal according to the multi-dimensional state features, the recommended feeding control action, the actual feeding control action, and the multi-dimensional device performance improvement information (the specific process can refer to step A5 in the foregoing);

[0166] a strategy adjustment module 6 for adjusting the automatic feeding control strategy generation logic according to the corrective feedback signal (the specific process can refer to step A6 in the foregoing);

[0167] a control execution module 7 for performing feeding control on the shredder according to the adjusted automatic feeding control strategy generation logic (the specific process can refer to step A7 in the foregoing).

[0168] The above merely describes the embodiments of the application and is not used to limit the protection scope of the application. For those skilled in the art, the application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1. A method for dynamically controlling the feeding of a sheet shredder, for controlling the feeding action of a plastic sheet shredder, characterized in that, The method comprises the following steps: A1. Obtain multi-dimensional equipment operation signals of the card crusher, and extract multi-dimensional state features therefrom; A2. According to the multi-dimensional state features, generate a recommended feeding control action by using an automatic feeding control strategy generation logic; A3. Monitor an actual feeding control action performed by an operator on the card crusher, and compare the actual feeding control action with the recommended feeding control action to identify a corrective behavior of the operator; A4. When the corrective behavior of the operator is identified, monitor card crusher equipment operation data in an observation time period after the actual feeding control action is performed, to evaluate multi-dimensional equipment performance improvement information; A5. According to the multi-dimensional state features, the recommended feeding control action, the actual feeding control action and the multi-dimensional equipment performance improvement information, generate a corrective feedback signal; A6. According to the corrective feedback signal, adjust the automatic feeding control strategy generation logic; A7. Perform feeding control on the card crusher according to the adjusted automatic feeding control strategy generation logic; Step A4 comprises: A401. When the corrective behavior of the operator is identified, determine the observation time period; A402. Monitor multi-dimensional equipment operation data of the card crusher in the observation time period; the multi-dimensional equipment operation data comprises motor average power, motor current, equipment sound signal energy in a specific frequency range, equipment vibration signal amplitude in a specific frequency range, discharge amount of the card crusher and discharge granularity distribution; A403. According to the multi-dimensional equipment operation data, evaluate multi-dimensional equipment performance improvement information; the multi-dimensional equipment performance improvement information comprises an efficiency improvement index, an energy consumption reduction index and a stability enhancement index; Step A401 comprises: Extract state features reflecting current processing material characteristics from the multi-dimensional state features as material characterization features; According to the material characterization features and the type of the actual feeding control action, determine the observation time period according to a preset observation time period determination rule or an observation time period lookup table; Step A403 comprises: B1. According to the multi-dimensional equipment operation data, calculate evaluation values of the efficiency improvement index, the energy consumption reduction index and the stability enhancement index; B2. According to the actual feeding control action, identify information reflecting a corrective behavior target, which is referred to as corrective intention information; B3. According to the material characterization features and the corrective intention information, determine weight coefficients of the efficiency improvement index, the energy consumption reduction index and the stability enhancement index; B4. According to the evaluation values of the efficiency improvement index, the energy consumption reduction index and the stability enhancement index and the corresponding weight coefficients, calculate a comprehensive performance improvement index; B5. Combine the evaluation values of the efficiency improvement index, the energy consumption reduction index and the stability enhancement index and the comprehensive performance improvement index to form the multi-dimensional equipment performance improvement information.

2. The method of claim 1, wherein, Step A1 comprises: A101. Obtain multi-dimensional equipment operation signals of the card crusher; the multi-dimensional equipment operation signals comprise motor current signals, equipment sound signals and equipment vibration signals; A102. performing waveform analysis on the motor current signal to extract impact load characteristics; A103. performing spectrum analysis on the device sound signal to extract sound spectrum characteristics; A104. performing spectrum analysis on the device vibration signal to extract vibration spectrum characteristics; A105. combining the impact load characteristics, the sound spectrum characteristics, and the vibration spectrum characteristics to form the multi-dimensional state characteristics.

3. The method of claim 1, wherein, Step A3 comprises: A301. obtaining actual control instructions and parameter settings executed by the operator through the card crusher control interface; A302. parsing the actual control instructions and parameter settings to identify the type and parameters of the actual feeding control action; A303. obtaining the type and parameters of the recommended feeding control action; A304. comparing the type and parameters of the actual feeding control action with the type and parameters of the recommended feeding control action to determine whether there is a difference between the actual feeding control action and the recommended feeding control action; A305. when there is a difference, determining that there is a corrective behavior, and taking the actual feeding control action as the operator's corrective behavior.

4. The method of claim 1, wherein, Step B2 comprises: based on a preset intention recognition rule or intention lookup table, identifying the corrective intention information according to the type and parameters of the actual feeding control action and the multi-dimensional state characteristics.

5. The method of claim 1, wherein, Step A6 comprises: A601. locating the strategy parameters or rules corresponding to the multi-dimensional state characteristics in the automatic feeding control strategy generation logic according to the multi-dimensional state characteristics in the corrective feedback signal as target strategy parameters or target rules; A602. calculating the feedback strength of the actual feeding control action according to the evaluation value of the efficiency improvement index, the evaluation value of the energy consumption reduction index, the evaluation value of the stability enhancement index, and the comprehensive performance improvement index in the multi-dimensional device performance improvement information in the corrective feedback signal; A603. adjusting the target strategy parameters or target rules according to the feedback strength to increase the probability of generating the actual feeding control action in a similar state to the multi-dimensional state characteristics.

6. A method of dynamic control of the feed of a card shuffler according to claim 5, characterized in that, Step A602 comprises: determining a preliminary feedback strength according to the comprehensive performance improvement index; determining an intention compliance degree according to the corrective intention information, the evaluation value of the efficiency improvement index, the evaluation value of the energy consumption reduction index, and the evaluation value of the stability enhancement index; correcting the preliminary feedback strength according to the intention compliance degree to obtain a final feedback strength.

7. A dynamic control system for the feeding action of a sheet shredder, for controlling the feeding action of a plastic sheet shredder, characterized in that, The system comprises: a state characteristic acquisition module for acquiring multi-dimensional device running signals of a card crusher and extracting multi-dimensional state characteristics therefrom; a recommended action generation module for generating a recommended feeding control action using an automatic feeding control strategy generation logic according to the multi-dimensional state characteristics; a corrective behavior identification module for monitoring actual feeding control actions performed by the operator on the card crusher, comparing the actual feeding control actions with the recommended feeding control actions, and identifying the operator's corrective behavior; The performance evaluation module is configured to monitor the running data of the card shuffler during an observation period after the actual feeding control action is performed to evaluate multi-dimensional device performance improvement information when the corrective behavior of the operator is identified. The feedback signal generation module is configured to generate a corrective feedback signal according to the multi-dimensional state feature, the recommended feeding control action, the actual feeding control action, and the multi-dimensional device performance improvement information. The strategy adjustment module is configured to adjust the automated feeding control strategy generation logic according to the corrective feedback signal. The control execution module is configured to perform feeding control on the card shuffler according to the adjusted automated feeding control strategy generation logic. The performance evaluation module monitors the running data of the card shuffler during an observation period after the actual feeding control action is performed to evaluate multi-dimensional device performance improvement information when the corrective behavior of the operator is identified, and specifically performs the following steps: A401. Determine the observation period when the corrective behavior of the operator is identified. A402. Monitor the multi-dimensional running data of the card shuffler during the observation period; the multi-dimensional running data of the card shuffler includes motor average power, motor current, device sound signal energy in a specific frequency range, device vibration signal amplitude in a specific frequency range, output amount of the card shuffler, and output particle size distribution. A403. Evaluate multi-dimensional device performance improvement information according to the multi-dimensional running data; the multi-dimensional device performance improvement information includes efficiency improvement indicators, energy consumption reduction indicators, and stability enhancement indicators. Step A401 includes: Extracting state features reflecting the characteristics of the current processing material from the multi-dimensional state features as material representation features; According to the material representation features and the type of the actual feeding control action, determine the observation period according to the preset observation period determination rule or observation period lookup table; Step A403 includes: B1. Calculate the evaluation values of the efficiency improvement indicators, the energy consumption reduction indicators, and the stability enhancement indicators according to the multi-dimensional running data; B2. Identify information reflecting the target of corrective behavior according to the actual feeding control action, denoted as corrective intention information; B3. Determine the weight coefficients of the efficiency improvement indicators, the energy consumption reduction indicators, and the stability enhancement indicators according to the material representation features and the corrective intention information; B4. Calculate the comprehensive performance improvement indicators according to the evaluation values of the efficiency improvement indicators, the energy consumption reduction indicators, and the stability enhancement indicators and the corresponding weight coefficients; B5. The efficiency improvement indicators, the energy consumption reduction indicators, the stability enhancement indicators, and the comprehensive performance improvement indicators constitute the multi-dimensional device performance improvement information.

Images