Shot blasting machine intelligent control system based on multi-source data acquisition

Abstract

The invention discloses a shot blasting machine intelligent control system based on multi-source data acquisition, and the system comprises a data acquisition module group which is used for collecting shot blasting machine operation related data in a multi-dimensional manner, and comprises a shot blasting machine body information acquisition module which is used for collecting an equipment model and a part identifier; the environment information acquisition module is used for acquiring environment temperature, humidity and dust concentration in real time; the image information acquisition module is used for synchronously acquiring a workpiece surface image and an operator behavior image; the running state acquisition module is used for acquiring equipment three-dimensional axial vibration waveforms, running voiceprints and key mechanism action images; the shot blasting quality acquisition module is used for detecting the surface cleanliness and roughness of the workpiece in real time; and the data processing module is used for analyzing and optimizing the collected data. Through multi-dimensional data fusion, intelligent algorithm modeling and a man-machine cooperation mechanism, accurate control and dynamic optimization of the shot blasting process are achieved, and the production efficiency and the process stability are improved.

Description

Technical Field

[0001] The present invention relates to the field of control systems, and particularly to an intelligent control system for shot blasting machines based on multi-source data acquisition. Background Art

[0002] A shot blasting machine is an industrial equipment that realizes surface cleaning, rust removal, strengthening or roughening treatment by high-speed projecting shot materials (such as steel shots, cast iron shots, etc.) onto the surface of workpieces, and is widely used in fields such as mechanical manufacturing, shipbuilding, and automotive. Its core principle is to use the high-speed rotation of the impeller of the shot blasting machine to project the shot materials onto the surface of the workpiece, and remove surface impurities or change the surface topography through the impact force to meet the subsequent processing or use requirements.

[0003] The control system of the shot blasting machine is the core module to ensure the accurate execution of the shot blasting process. Its main functions include: real-time monitoring of the equipment operation status (such as vibration, sound pattern, component movement), adjustment of process parameters (such as shot blasting pressure, shot material flow rate, projection time), ensuring the workpiece processing quality (such as cleanliness, roughness) and equipment maintenance management. With the development of industrial intelligence, traditional control systems are gradually difficult to meet the production requirements of high precision and high efficiency due to the single dimension of data acquisition and the insufficient ability of algorithm optimization.

[0004] The existing control systems of shot blasting machines mainly have the following technical problems: single dimension of data acquisition, insufficient intelligence, low fault diagnosis efficiency, poor environmental adaptability and weak human-machine collaboration ability. Therefore, an intelligent control system for shot blasting machines based on multi-source data acquisition is proposed. Summary of the Invention

[0005] The technical problem to be solved by the present invention is: how to solve the problems of single dimension of data acquisition, insufficient intelligence, low fault diagnosis efficiency, poor environmental adaptability and weak human-machine collaboration ability in the existing technology, and provide an intelligent control system for shot blasting machines based on multi-source data acquisition.

[0006] The present invention solves the above technical problems through the following technical solutions. The present invention includes: A data acquisition module group for multi-dimensionally acquiring data related to the operation of the shot blasting machine, including: A shot blasting machine body information acquisition module for acquiring the equipment model and component identification; An environmental information acquisition module for real-time acquiring the environmental temperature, humidity and dust concentration; An image information acquisition module for synchronously acquiring the surface image of the workpiece and the behavior image of the operator; An operation status acquisition module for acquiring the three-dimensional axial vibration waveform of the equipment, the operation sound pattern and the action image of the key mechanism; A shot blasting quality acquisition module for real-time detecting the cleanliness and roughness of the workpiece surface; A data processing module for analyzing and optimizing the collected data, including: A time synchronization controller that, based on Aligns the data timestamps of each module according to the protocol and outputs a synchronized data packet with a time deviation less than 1 ms; A process optimization engine that generates process parameter optimization instructions based on the shot peening quality detection value and the synchronized data packet; A fault collaborative diagnosis unit that locates abnormal components of the equipment based on vibration, acoustic fingerprint, and image features; A human-machine collaborative optimization module that integrates manual intervention parameters to generate final optimization instructions; An information sending module for writing the final optimization instructions into The controller and pushing the fault location information and operation suggestions.

[0007] Furthermore, the operating state acquisition module includes: A vibration sensor group arranged on the bearing seat for outputting a three-dimensional axial vibration spectrum; An omnidirectional microphone array installed on the side wall of the shot peening chamber for collecting the operating acoustic fingerprint from 200 Hz to 10 kHz; A high-speed imaging unit for collecting dynamic images of the impeller and the blast wheel of the shot peener.

[0008] Furthermore, the data processing module also includes a time synchronization controller, which is used to: receive the original data streams of each acquisition module; Based on Align all data timestamps according to the protocol; Output a synchronized data packet with a time deviation less than 1 ms to the process optimization engine and the fault collaborative diagnosis unit.

[0009] Furthermore, the process optimization engine executes: Receives the measured values of the actual cleanliness and roughness of the workpiece surface, i.e., the shot peening quality detection value And the synchronized data packet; Calculates the quality prediction value The specific process is: ; In the formula, Is the workpiece surface texture feature matrix, which is obtained by performing Feature extraction on the workpiece surface image collected by the image information acquisition module; Is the three-dimensional axial vibration spectrum feature vector of the equipment, which is obtained by performing Time series coding on the data collected by the vibration sensor group of the operating state acquisition module; Is the Mel cepstral coefficient of the operating acoustic fingerprint , obtained through feature extraction by the microphone array of the operating status acquisition module ; is the model weight, called from the process parameter library based on the device model; Solve for the parameter set that makes the smallest .

[0010] Further, the process optimization engine includes an adaptive feature extraction unit. When the environmental dust concentration changes beyond the threshold or the temperature and humidity deviate from the standard range, it reconstructs the voiceprint features and replaces the model input, where: The specific process of reconstructing the voiceprint features is: ; In the formula, is the compensation function of the environmental parameters (temperature, humidity) for the voiceprint; The specific process of updating the model input is: .

[0011] Further, the fault collaborative diagnosis unit performs the following process: Extract the vibration spectrum features in the synchronization data packet and the voiceprint features ; Establish a spatio-temporal alignment model: ; Output the fault component number: ; In the formula is the Pearson correlation coefficient (range [-1, 1]), used to calculate the linear correlation intensity between vibration and voiceprint; is the spatio-temporal overlap degree (range [0, 1]), used to quantify the matching ratio between the vibration peak region and the image motion region; is the weight, is the optical flow change amount of the key mechanism action.

[0012] Further, the update mechanism of the process optimization engine is: When the fault component number of the severity coefficient ; Update the weight: ; Among them, represents the update amount of the model weight in the process optimization engine, where The weights corresponding to different features in the model, is the learning rate parameter for weight update, used to control the step size of weight update and determine the amplitude of each weight adjustment. Loss is the loss function in the process optimization engine, used to quantify the deviation between the predicted value of shot peening quality and the actual detected value. is used to represent the th weight of the loss function, reflecting the rate and direction of change of the loss function with respect to this weight, and used to indicate the direction and basic amplitude of weight adjustment; is a preset value. When triggers the weight update mechanism.

[0013] Furthermore, the human-machine collaborative optimization module performs: When it detects that the operator manually adjusts the parameters, it synchronously extracts the behavioral operation action feature vector of the image information acquisition module through the network to predict the confidence level of the adjustment intention ; Records the adjustment amount and generates the final optimization instruction: ; Among them, is calculated based on the correlation between the gesture trajectory in the behavioral image and the parameter adjustment history, is the weight coefficient of the manually intervened parameter, used to adjust the influence degree of the operator's manual adjustment amount in the final instruction.

[0014] Furthermore, the human-machine collaborative optimization module responds to the output of the fault collaborative diagnosis unit: Generates a three-dimensional fault mark according to the faulty component number and superimposes the abnormal action hot zone in the operator's behavioral image; The spatial coordinates are determined by the component identification and the three-dimensional mapping function , and the abnormal action hot zone is generated by analyzing the pixel motion intensity distribution in the behavioral image through the optical flow method ; The pixel motion intensity distribution in the image is correlated with the spatio-temporal distribution of the operation action feature vector ; The information sending module performs: Writes into the control register; Generates a fault report containing the component coordinates and pushes operation suggestions.

[0015] Furthermore, the operator behavior images collected by the image information acquisition module are processed by computer vision algorithms. The specific processing process includes: using the model to perform human pose recognition on the behavior images and extract operation action feature vectors ; The operation action feature vectors are dynamically time warped with the preset compliance operation template library to generate an operation compliance score ; When the sends a warning signal to the human-machine collaboration optimization module, where

[0016] The present invention has the following advantages compared with the prior art: The intelligent control system of the shot blasting machine based on multi-source data acquisition realizes precise optimization of the shot blasting process and efficient diagnosis of equipment failures by collecting shot blasting machine operation data in multiple dimensions and performing intelligent processing. The system comprehensively obtains information such as equipment operation status, environmental parameters, and workpiece quality through the multi-source data acquisition module group, ensures data consistency by combining time synchronization technology, dynamically optimizes process parameters with the help of intelligent algorithms, and can adjust parameters in real time according to the actual shot blasting quality to improve the processing effect. At the same time, the fault collaborative diagnosis mechanism based on vibration, voiceprint, and image features can accurately locate abnormal components of the equipment and improve the efficiency of fault troubleshooting. The human-machine collaboration optimization module integrates manual intervention and intelligent prediction, giving full play to the optimization ability of artificial intelligence and combining the experience of operators to improve the adaptability and operation convenience of the system. In addition, the system also has the ability to adaptively process the environment, can automatically adjust the model input when environmental parameters such as dust concentration, temperature, and humidity change, and ensure the stable operation of the system; monitors the operator's behavior through computer vision algorithms to ensure operation compliance and reduce the risk of human error. Overall, the system realizes intelligent and precise control of the shot blasting process, improves production efficiency, processing quality, and equipment maintenance level, making the system more worthy of popularization and use. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is the overall structure diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The following details the embodiments of the present invention. The embodiments are implemented on the premise of the technical solution of the present invention, and detailed implementation manners and specific operation processes are given. However, the protection scope of the present invention is not limited to the following embodiments.

[0019] As Figure 1 shown, this embodiment provides a technical solution: An intelligent control system of a shot blasting machine based on multi-source data acquisition, including: A data acquisition module group for multi-dimensionally acquiring data related to the operation of a shot blasting machine, including: A shot blasting machine body information acquisition module for acquiring equipment model and component identification; An environmental information acquisition module for real-time acquisition of environmental temperature, humidity and dust concentration; An image information acquisition module for synchronously acquiring the surface image of the workpiece and the behavior image of the operator; An operating state acquisition module for acquiring the three-dimensional axial vibration waveform, operating sound pattern and key mechanism action images of the equipment; A shot blasting quality acquisition module for real-time detection of the surface cleanliness and roughness of the workpiece; A data processing module for analyzing and optimizing the acquired data, including: A time synchronization controller, based on Protocol to align the data timestamps of each module and output a synchronized data packet with a time deviation less than 1 ms; A process optimization engine for generating process parameter optimization instructions according to the shot blasting quality detection value and the synchronized data packet; A fault collaborative diagnosis unit for locating abnormal components of the equipment based on vibration, sound pattern and image features; A human-machine collaborative optimization module for integrating manual intervention parameters to generate final optimization instructions; An information sending module for writing the final optimization instructions into The controller and push the fault location information and operation suggestions.

[0020] The operating state acquisition module includes: A vibration sensor group arranged on the bearing seat for outputting a three-dimensional axial vibration spectrum; A directional microphone array installed on the side wall of the shot blasting chamber for acquiring an operating sound pattern of 200 Hz - 10 kHz; A high-speed imaging unit for acquiring dynamic images of the impeller and the dosing wheel of the shot blasting machine; By setting a vibration sensor group, a directional microphone array and a high-speed imaging unit in the operating state acquisition module, key operating data of the shot blasting machine can be acquired in real time from multiple dimensions. Among them, the vibration sensor group can capture the three-dimensional axial vibration spectrum of the bearing seat for monitoring the operating stability of the mechanical components of the equipment; the directional microphone array can acquire the operating sound pattern of the shot blasting chamber to assist in identifying mechanical abnormal sounds; the high-speed imaging unit acquires dynamic images of key mechanisms such as the impeller and the dosing wheel of the shot blasting machine, intuitively reflecting the action state of the components. The collaborative arrangement of multiple types of sensors realizes the comprehensive monitoring of equipment vibration, sound and mechanical actions, provides multi-source data support for subsequent fault location and operating state analysis, improves the perception accuracy and real-time performance of the system for equipment abnormalities, and ensures the comprehensiveness and accuracy of the operating state monitoring of the shot blasting machine.

[0021] The data processing module further includes a time synchronization controller, which is used for: receiving the original data streams of each acquisition module; Based on the protocol, align all data timestamps; Output a synchronization data packet with a time deviation less than 1 ms to the process optimization engine and the fault collaborative diagnosis unit; By eliminating the time deviation of the data of each acquisition module (the time deviation of the output synchronization data packet is less than 1 ms), it ensures the consistency of multi-dimensional data such as vibration spectrum, voiceprint, and image in the time dimension, providing an accurate time reference for the parameter calculation of the subsequent process optimization engine and the abnormal positioning of the fault collaborative diagnosis unit. This mechanism avoids analysis errors caused by data asynchronization, enabling the system to perform more accurate process parameter optimization and equipment fault feature extraction based on the spatio-temporal aligned data streams, and improving the response accuracy and reliability of the intelligent control system to the dynamic changes in the shot peening process.

[0022] The process optimization engine performs: Receiving the measured values of the actual cleanliness and roughness of the workpiece surface, i.e., the shot peening quality detection values and the synchronization data packet; Calculating the quality prediction value The specific process is as follows: ; In the formula, is the workpiece surface texture feature matrix, which is obtained by performing feature extraction on the workpiece surface image collected by the image information acquisition module; is the device three-dimensional axial vibration spectrum feature vector, which is obtained by performing time series coding on the data collected by the vibration sensor group of the operating state acquisition module; is the Mel frequency cepstral coefficient (MFCC) of the running voiceprint, which is obtained by performing feature extraction through the microphone array of the operating state acquisition module; is the model weight, which is called from the process parameter library based on the device model; Solve the parameter set that makes the smallest ; Combining the workpiece surface image ( extracting texture features), the device vibration spectrum ( processing time series features), and the running voiceprint ( extracting acoustic features), construct a comprehensive quality prediction model to avoid analysis biases in a single data dimension and make the process parameter optimization more in line with the actual production requirements.

[0023] Intelligent prediction and precise parameter tuning: By weighted fusion of the outputs of multiple models, that is , calculate the deviation between the quality prediction value and the actual value in real time, and automatically solve the optimal parameter set , realizing closed-loop control from detection, analysis to optimization, and reducing the cost of manual trial and error.

[0024] For example, the surface cleanliness of a workpiece after shot peening does not meet the standard, such as residual scale.

[0025] The image information acquisition module captures uneven surface texture of the workpiece, the vibration sensor detects abnormal axial vibration of the shot peening machine, and the microphone array collects abnormal sound patterns, such as high-frequency noise.

[0026] Extract the texture feature matrix of the residual scale from the image ; Analyze the vibration spectrum and find the abnormal vector of the sequential vibration when the impeller of the shot peening machine rotates ; Extract the features of the sound pattern and identify the acoustic mode related to component wear.

[0027] The process optimization engine passes the model to calculate the quality prediction deviation under the current parameters (such as shot peening pressure, shot material flow rate), automatically adjust the parameters, and generate the optimal parameter set , such as increasing the shot peening pressure by 10% and extending the shot peening time by 5s to meet the cleanliness standard.

[0028] This process requires no manual intervention. Through the collaboration of multi-source data and intelligent algorithms, it realizes the adaptive optimization of the shot peening process, improving production efficiency and quality stability.

[0029] The process optimization engine includes an adaptive feature extraction unit. When the environmental dust concentration changes beyond the threshold or the temperature and humidity deviate from the standard range, it reconstructs the sound pattern features and replaces the model input, where:[[]] The specific process of reconstructing the sound pattern features is:[[]] ; In the formula, is the compensation function of the environmental parameters (temperature, humidity) for the sound pattern; The specific process of updating the model input is:[[]] ; Through the adaptive feature extraction unit, the system can automatically reconstruct the voiceprint features and adjust the model input when the dust concentration suddenly changes or the temperature and humidity deviate from the standard range, avoiding the deviation of process parameter optimization caused by environmental interference and ensuring the stability of shot peening quality under different working conditions.

[0030] Without manual intervention, the system can dynamically correct the voiceprint features according to real-time environmental parameters, such as filtering dust noise and compensating for the influence of temperature and humidity on sound waves, so that the quality prediction model can continuously maintain high accuracy and improve the reliability of intelligent control.

[0031] For example, due to a malfunction of the dust removal system in the shot peening workshop, the dust concentration suddenly rises from 50mg / m 3 to 200mg / m 3 , exceeding the threshold of 100mg / m 3 , which affects the accuracy of voiceprint acquisition.

[0032] The environmental information acquisition module detects that the dust concentration exceeds the threshold and triggers the adaptive feature extraction unit.

[0033] Voiceprint feature reconstruction: The system applies a Wiener filter to the original voiceprint to filter out dust noise interference and obtain the corrected voiceprint ; At the same time, it calls the environmental compensation function to perform parameter compensation for the influence of the current temperature, such as 35°C, and humidity, such as 70%, on the propagation of the voiceprint, further optimizing the features.

[0034] The process optimization engine switches the model input and no longer uses the original voiceprint , but recalculates the quality prediction value based on the corrected and vibration spectrum to ensure that the parameter optimization instruction is not interfered by dust and maintain the stability of shot peening quality.

[0035] This mechanism enables the system to have self-purification ability in a complex industrial environment, avoiding fluctuations in shot peening effects caused by environmental fluctuations and reducing production abnormalities and quality defects caused by environmental factors.

[0036] The fault collaborative diagnosis unit performs the following process: Extract the vibration spectrum features and voiceprint features from the synchronous data packet; Establish a spatio-temporal alignment model: ; Output the fault component number: ; where is the Pearson correlation coefficient (range [-1, 1]), which is used to calculate the linear correlation strength between vibration and voiceprint; is the spatio-temporal overlap degree (range [0, 1]), which is used to quantify the matching ratio between the vibration peak region and the image motion region; is the weight, is the optical flow change of the key mechanism action; Multi-source data collaborative location fault: By fusing vibration spectrum, voiceprint features and image optical flow changes, a spatio-temporal alignment fault diagnosis model is constructed to break through the detection limitations of a single sensor, achieve precise location of abnormal components of the device, and reduce the misjudgment rate.

[0037] Dynamic correlation features improve diagnostic accuracy: Using the Pearson correlation coefficient to quantify the linear correlation between vibration and voiceprint, combined with the spatio-temporal overlap degree to match the vibration peak with the image motion region, making the fault location have both the synchronization in the time dimension and the correlation in the space dimension, and improving the diagnostic efficiency.

[0038] For example, abnormal operation occurs due to wear of the bearing of the shot blasting machine impeller.

[0039] The vibration sensor group detects the three-dimensional axial vibration spectrum of the bearing seat showing high-frequency abnormal fluctuations; The directional microphone array collects periodic abnormal sound features in the voiceprint of 200 Hz - 10 kHz ; The high-speed camera unit captures the small offset during the rotation of the impeller of the shot blasting machine, and the optical flow method analyzes the abnormal motion intensity.

[0040] The system calculates the optimal alignment time through the formula , making the difference between vibration and voiceprint and the optical flow change amount the smallest, ensuring the synchronization of multi-source data in the time dimension; Calculate the Pearson correlation coefficient between vibration and voiceprint , and find that the linear correlation between the two is significant, such as , and at the same time, the overlap degree between the vibration peak region and the motion region of the bearing part in the image reaches 0.78.

[0041] The fault collaborative diagnosis unit integrates according to the weight and values, calculates the fault probability of each component, and finally outputs the bearing component number , locating the fault source.

[0042] This mechanism avoids misjudgment of single signals through cross-verification of multi-dimensional data. For example, relying solely on vibration or voiceprint may lead to missed diagnoses. It realizes fault location with a triple evidence chain of vibration, sound, and images, and is especially suitable for abnormal detection of equipment in high-vibration and high-noise environments such as shot blasting machines, improving maintenance efficiency and reducing downtime.

[0043] The update mechanism of the process optimization engine is as follows: When the fault component number has a severity coefficient ; Update the weight: ; where represents the update amount of the model weight in the process optimization engine, where corresponds to the weights of different features in the model, is the learning rate parameter for weight update, used to control the step size of weight update and determine the amplitude of each weight adjustment, is the loss function in the process optimization engine, used to quantify the deviation between the predicted value of shot blasting quality and the actual detected value, is the partial derivative of the loss function with respect to the th weight , reflecting the rate and direction of change of the loss function with respect to this weight, and used to indicate the direction and basic amplitude of weight adjustment; is a preset value. When triggers the weight update mechanism; By binding the severity coefficient of the fault component to the model weight update mechanism, the system can dynamically adjust the parameter sensitivity of the process optimization model according to the severity of equipment abnormalities, avoiding optimization deviations of process parameters caused by component failures, and improving the robustness of the system under non-ideal working conditions.

[0044] Model self-evolution ability: When a fault occurs in the equipment, the system automatically increases the weights of data dimensions related to the fault, such as vibration and voiceprint features, making the process optimization model pay more attention to the state changes of abnormal components, forming a closed-loop adaptive mechanism from fault detection to weight adjustment to parameter optimization, reducing the cost of manual intervention; For example, the impeller of the shot blasting machine in the shot blasting machine wears due to long-term use, and the severity coefficient exceeds the threshold , resulting in unstable shot blasting force.

[0045] The fault collaborative diagnosis unit locates the impeller component number , and calculates its according to the abnormal degree of the vibration spectrum, the offset of the voiceprint feature, etc.

[0046] The process optimization engine triggers the update mechanism. Based on the gradient descent formula, it increases the model weights related to the impeller state.

[0047] For example: the weight corresponding to the vibration characteristics (original weight 0.3), due to abnormal vibration caused by impeller wear, its gradient is positive. Combining , it is calculated that , after update ; The weight of the voiceprint feature (original weight 0.2), due to significant features related to abnormal noise, the gradient is positive, , after update .

[0048] After the model is updated, when the process optimization engine calculates the quality prediction value , it relies more on the vibration spectrum and the voiceprint features, automatically adjusts the shot blasting parameters (such as increasing the shot supply by 15%), compensates for the insufficient shot blasting force caused by impeller wear, and maintains the surface cleanliness of the workpiece up to standard.

[0049] The human-machine collaborative optimization module performs: When it detects that the operator manually adjusts the parameters, it synchronously extracts the behavioral operation action feature vector of the image information acquisition module , through the network to predict the confidence level of the adjustment intention ; Records the adjustment amount , generates the final optimization instruction: ; Among them, is calculated based on the correlation between the gesture trajectory in the behavioral image and the parameter adjustment history, is the weight coefficient of the manually intervened parameter, which is used to adjust the influence degree of the operator's manual adjustment amount in the final instruction; By combining the operator's manual adjustment and the system's intelligent prediction, it realizes the dynamic complementarity of human experience and algorithm optimization, gives full play to the intuitive judgment ability of humans for complex working conditions, and also utilizes the accurate calculation advantage of the system to improve the flexibility and adaptability of shot blasting process adjustment. Through the network analyzes the operator's behavioral characteristics (such as gesture trajectory, operation habits), predicts the confidence level of the manual adjustment intention, enables the system to automatically judge the rationality of manual intervention, avoids blindly following ineffective operations, and reduces the trial-and-error cost of parameter adjustment.

[0050] The operator finds that the surface roughness of a certain batch of workpieces is too high after shot blasting and manually increases the rotational speed of the shot blasting motor 。

[0051] The image information acquisition module captures the gesture trajectory of the operator rotating the speed adjustment knob, and the model extracts the action feature vector (such as gesture speed and rotation amplitude).

[0052] Based on the and historical adjustment data (such as the adjustment mode when the roughness of similar workpieces was insufficient in the past), the network calculates the confidence level of the adjustment intention (a high confidence level indicates that the operation intention is highly correlated with roughness optimization).

[0053] Final optimization instruction generation: The original optimization parameter set of the system is that the rotational speed is increased by 3% and the shot flow rate remains unchanged. Combining the manual adjustment amount with a confidence level of 0.9, the final optimization instruction is generated: , that is, the rotational speed is increased by 3% + 0.8×5%×0.9 = 3% + 3.6% = 6.6%, and at the same time, the shot flow rate is synchronously adjusted to match the rotational speed change.

[0054] After shot peening, the roughness of the workpiece meets the standard. By integrating manual experience and algorithm optimization, the system reduces the number of parameter iterative adjustments by 2 compared to relying solely on automatic optimization, improving production efficiency.

[0055] The human-machine collaborative optimization module responds to the output of the fault collaborative diagnosis unit: Generates a three-dimensional fault marker based on the faulty component number and superimposes the abnormal action hot zone in the operator's behavior image; The spatial coordinates are determined by the component identification and the three-dimensional mapping function The abnormal action hot zone is generated by analyzing the pixel motion intensity distribution in the behavior image through the optical flow method ; The pixel motion intensity distribution in the said image is correlated with the spatio-temporal distribution of the operation action feature vector ; The information sending module executes: Writes the into the PLC control register; Generates a fault report containing the component coordinates and pushes operation suggestions; By converting the faulty component number into three-dimensional spatial coordinates and superimposing the abnormal action hot zone, an intuitive mapping from the fault location to the operation behavior is achieved, helping the operator quickly locate the physical fault point and the area related to human operation, reducing the maintenance troubleshooting time.

[0056] Analyze the operator's behavior image in combination with the optical flow method, dynamically associate equipment failures with abnormal operation actions, generate a fault report with spatial coordinates and operation suggestions, provide dual support of data-driven and visualization guidance for maintenance, and lower the technical threshold of maintenance.

[0057] For example, the bearing of the shot blasting machine's shot blasting device is worn (fault component number ), and at the same time, the operator frequently touches the nearby sensor by mistake during debugging.

[0058] Fault diagnosis and 3D marking generation: The fault collaborative diagnosis unit outputs the bearing component number , and the system calculates its spatial coordinates (x = 2.5m, y = 1.8m, z = 0.5m) through the 3D mapping function , and superimposes a red 3D mark (such as a highlighted cube at the bearing position) on the operator's behavior image.

[0059] Analysis of abnormal action hotspots: The optical flow method analyzes the behavior image and finds that the pixel motion intensity distribution in the area corresponding to the bearing of the operator is abnormal. For example, the area where the hand frequently approaches the sensor has a high brightness, and a yellow hotspot mark is generated and displayed overlapping with the 3D fault mark.

[0060] The system associates the operation action feature vector with the spatio-temporal distribution of the hotspot to identify the abnormal operation mode of touching the sensor by mistake.

[0061] Information push and maintenance guidance: The information sending module writes the final optimization instruction into , and at the same time generates a fault report, including: highlighting the equipment structure diagram corresponding to the 3D coordinates ; operation video clips of the abnormal action hotspot (such as a 10s video of the hand touching the sensor by mistake); operation suggestions: Check the lubrication status of the bearing and avoid touching sensor A (coordinates x = 2.6m, y = 1.7m) during debugging.

[0062] According to the visualization guidance, the operator locates the bearing wear problem within 10 minutes and corrects the operation habit, reducing the troubleshooting time compared with the traditional maintenance process and avoiding secondary failures caused by incorrect operations.

[0063] The operator's behavior image collected by the image information acquisition module is processed through computer vision algorithms. The specific processing process includes: using the model to perform human pose recognition on the behavior image and extract the operation action feature vector ; The operation action feature vector and preset compliance operation template library Dynamic Time Warping Matching, generating operational compliance scores ; when When the warning signal is sent to the human-machine collaborative optimization module, is the preset threshold; Computer vision algorithms are used to analyze operator behavior in real time, automatically identify illegal operations (such as failure to debug according to the process, dangerous actions), generate compliance scores and issue warnings, reduce the risk of equipment failure or quality defects caused by human errors, dynamically match preset compliance operation templates with actual actions, provide real-time feedback to operators, and promote the standardization of operating habits. This is especially suitable for new employee training or complex process scenarios, and can reduce production anomalies caused by non-standard operations.

[0064] If the operator does not close the protective door according to the standard process when the shot blasting machine is running, and directly reaches out to adjust the shot blasting machine parameters.

[0065] Behavior image acquisition and feature extraction: The image information acquisition module captures the operator's hand movement images. The model recognizes the gesture features of the hand approaching the running shot blasting machine and extracts the action feature vector , such as hand movement trajectory and distance parameters from the device.

[0066] Compliance score calculation: The system will and preset compliance template library Dynamic time regularization using the shutdown debugging action template in Matching, calculating the deviation between the operation trajectory and the standard process. Because the operator did not close the protective door and operated while the equipment was running, the matching similarity is low, and a compliance score is generated. (Below the preset threshold ).

[0067] The system sends an early warning signal to the human-machine collaborative optimization module, and displays a red warning box on the operation interface, marking the illegal action area, such as the video clip of the hand approaching the shot blasting wheel, and pushes a text prompt: Please close the protective door before performing debugging operations.

[0068] The operator stopped the illegal action in time and followed the standard process, avoiding downtime accidents caused by mechanical damage or accidental contact of equipment. At the same time, the system recorded the violation and provided data support for subsequent training.

[0069] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0070] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0071] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. An intelligent control system for a shot blasting machine based on multi-source data acquisition, characterized in that, Including: A data acquisition module group for multi-dimensionally acquiring data related to the operation of a shot blasting machine, including: A shot blasting machine body information acquisition module for acquiring the equipment model and component identification; An environmental information acquisition module for real-time acquiring the environmental temperature, humidity, and dust concentration; An image information acquisition module for synchronously acquiring the workpiece surface image and the operator behavior image; An operation status acquisition module for acquiring the three-dimensional axial vibration waveform, operation sound pattern, and key mechanism action images of the equipment; A shot blasting quality acquisition module for real-time detecting the surface cleanliness and roughness of the workpiece; A data processing module for analyzing and optimizing the acquired data, including: A time synchronization controller for aligning the data timestamps of each module based on a preset type of protocol and outputting a synchronization data packet with a time deviation less than a preset value; A process optimization engine for generating a process parameter optimization instruction according to the shot blasting quality detection value and the synchronization data packet; A fault collaborative diagnosis unit for locating the abnormal components of the equipment based on vibration, sound pattern, and image features; A human-machine collaborative optimization module for integrating manual intervention parameters to generate a final optimization instruction; An information sending module, which is used to write the final optimization instruction into the controller and push the fault location information and operation suggestions.

2. The intelligent control system of a shot blasting machine based on multi-source data acquisition according to claim 1, wherein: The operation status acquisition module includes: A vibration sensor group arranged on the bearing seat for outputting a three-dimensional axial vibration spectrum; An omnidirectional microphone array installed on the side wall of the shot blasting chamber for acquiring the operation sound pattern of a preset frequency; A high-speed camera unit for acquiring the dynamic images of the impeller and the blast wheel of the shot blasting machine.

3. An intelligent control system for a shot blasting machine based on multi-source data acquisition according to claim 1, characterized in that: The time synchronization controller is used to: receive the original data streams of each acquisition module; Align all data timestamps based on a preset type of protocol; Output a synchronization data packet with a time deviation less than a preset value to the process optimization engine and the fault collaborative diagnosis unit.

4. The intelligent control system of a shot blasting machine based on multi-source data acquisition according to claim 1, characterized in that: The process optimization engine executes: Receive the measured values of the actual cleanliness and roughness of the workpiece surface, i.e., the shot peening quality inspection values and synchronous data packets; Calculate the predicted mass value , and the specific process is as follows: ； In the formula, is the workpiece surface texture feature matrix, which is obtained by performing feature extraction on the workpiece surface image collected by the image information acquisition module; It is the three-dimensional axial vibration spectrum feature vector of the device, which is obtained by performing time series coding on the data collected by the vibration sensor group of the operating state acquisition module; The Mel cepstral coefficients of the running voiceprint are obtained through feature extraction by the microphone array of the running state acquisition module; ​ is the model weight, which is called from the process parameter library based on the device model; Solve for the parameter set that minimizes the minimum .

5. The intelligent control system of a shot blasting machine based on multi-source data acquisition according to claim 4, characterized in that: The process optimization engine includes an adaptive feature extraction unit. When the environmental dust concentration changes beyond the threshold or the temperature and humidity deviate from the standard range, it reconstructs the sound pattern features and replaces the model input, where: The specific process of reconstructing the sound pattern features is: ； The specific process of replacing the model input is: 。 6. The intelligent control system of a shot blasting machine based on multi-source data acquisition according to claim 1, characterized in that: The fault collaborative diagnosis unit executes the following process: Extract the vibration spectrum features in the synchronization data packet and voiceprint features ; Establish a spatio-temporal alignment model: ； Output the fault component number: 。 7. The intelligent control system of a shot blasting machine based on multi-source data acquisition according to claim 6, wherein: The update mechanism of the process optimization engine is: When the failure component number has a severity coefficient ; Update the weight: ； Among them, represents the update amount of the model weights in the process optimization engine, is the learning rate parameter for weight update, is the loss function in the process optimization engine, represents the partial derivative of the loss function with respect to the i-th weight ωi, is a preset value. When the weight update mechanism is triggered.

8. An intelligent control system for a shot blasting machine based on multi-source data acquisition according to claim 1, characterized in that: The human-machine collaborative optimization module executes: When it is detected that the operator manually adjusts the parameters, the behavioral operation action feature vector of the image information acquisition module is synchronously extracted , through the network to predict the confidence level of the adjustment intention ; Record adjustment amount Generate the final optimization instruction: ； Among them, The calculation is based on the correlation between the gesture trajectory in the behavior image and the parameter adjustment history, which is the weight coefficient of the manual intervention parameter.

9. An intelligent control system for a shot blasting machine based on multi-source data acquisition according to claim 8, characterized in that: The human-machine collaborative optimization module responds to the output of the fault collaborative diagnosis unit: Generates a three-dimensional fault mark according to the fault component number k and superimposes the abnormal action hot zone in the operator behavior image; The spatial coordinates are determined by the component identifier and the three-dimensional mapping function The abnormal action hot zone is generated by analyzing the pixel motion intensity distribution in the behavior image through the optical flow method ; The pixel motion intensity distribution in the image is correlated with the spatio-temporal distribution of the operation action feature vector; The information sending module executes: Write to the control register; Generate a fault report containing component coordinates and push operation suggestions.

10. The intelligent control system of a shot blasting machine based on multi-source data acquisition according to claim 1, characterized in that: The operator behavior images collected by the image information acquisition module are processed by computer vision algorithms. The specific processing process includes: using the model to perform human pose recognition on the behavior images and extract operation action feature vectors ; The operation action feature vector is dynamically time warped and matched with a preset compliance operation template library to generate an operation compliance score ; When a warning signal is sent to the human-machine collaborative optimization module, which is a preset threshold value.

Images