Brick quality monitoring method and system based on data analysis and storage medium

Abstract

The invention relates to the technical field of data analysis, and discloses a brick quality monitoring method and system based on data analysis and a storage medium. The method comprises the following steps: collecting vibration, pressure, displacement, image and air pressure data on a brick production line to obtain a data set; performing domain synchronous averaging and edge detection on the data set to obtain a brick surface defect feature frequency, a brick contour feature vector and a quality anomaly index; analyzing through a convolutional neural network to obtain a brick defect type judgment result and a defect severity score; and technological parameter optimization is executed, an equipment control strategy is obtained and transmitted to each execution mechanism, and a motor control signal, a servo positioning instruction, a contour adjustment instruction and management layer optimization decision information are obtained. According to the method, the sensitivity and the accuracy of brick defect detection are improved, so that a brick production system can be continuously self-optimized to effectively cope with internal changes and external disturbance, and the quality stability and the production efficiency of brick production are remarkably improved.

Description

Technical Field

[0001] This application relates to the field of data analysis technology, and in particular, to a brick quality monitoring method, system, and storage medium based on data analysis. Background Art

[0002] In the traditional brick production process, quality monitoring mainly relies on manual visual inspection and simple physical measurements. This method is not only inefficient but also has a high misjudgment rate, making it difficult to detect potential problems such as polygonal defects and contour errors on the brick surface in a timely manner. On high-speed production lines, manual inspection can no longer meet the refined requirements for quality control in modern production, resulting in a large number of unqualified products flowing into the market, increasing subsequent construction risks and material waste.

[0003] Although existing automated monitoring technologies have achieved mechanization in some detection links, they still have obvious defects. These systems often only focus on optimizing mechanical parameters during the conveying process, lacking data collection, analysis, and feedback control for the entire production process. At the same time, existing equipment cannot effectively handle the problem of breaking up stacked materials, resulting in uneven stress on the conveyor belt, which gradually deforms during long-term use and affects the transmission efficiency. More seriously, materials are prone to backward movement and backflow during upward transportation, not only causing material waste but also making it difficult to ensure the consistent quality of the final product. Summary of the Invention

[0004] This application provides a brick quality monitoring method, system, and storage medium based on data analysis. This application improves the sensitivity and accuracy of brick defect detection, enables the brick production system to continuously self-optimize, effectively responds to internal changes and external disturbances, and significantly improves the quality stability and production efficiency of brick production.

[0005] In a first aspect, this application provides a brick quality monitoring method based on data analysis. The brick quality monitoring method based on data analysis includes:

[0006] Collect vibration, pressure, displacement, image, and air pressure data on the brick production line to obtain a data set;

[0007] Perform domain synchronous averaging and edge detection on the data set to obtain the brick surface defect characteristic frequency, brick contour feature vector, and quality anomaly index;

[0008] Input the brick surface defect characteristic frequency, the brick contour feature vector, and the quality anomaly index into a convolutional neural network for analysis to obtain a brick defect type determination result and a defect severity score;

[0009] Execute process parameter optimization based on the defect type determination result and the defect severity score to obtain an equipment control strategy;

[0010] Transmit the device control strategy to each actuator to perform closed-loop feedback, obtaining a motor control signal, a servo positioning command, a contour adjustment command, and management optimization decision-making information.

[0011] In a second aspect, the present application provides a brick quality monitoring system based on data analysis. The brick quality monitoring system based on data analysis includes:

[0012] An acquisition module, configured to acquire vibration, pressure, displacement, image, and air pressure data on a brick production line to obtain a data set;

[0013] A detection module, configured to perform domain synchronous averaging and edge detection on the data set to obtain brick surface defect characteristic frequencies, brick contour feature vectors, and quality anomaly indicators;

[0014] An analysis module, configured to input the brick surface defect characteristic frequencies, the brick contour feature vectors, and the quality anomaly indicators into a convolutional neural network for analysis to obtain a brick defect type determination result and a defect severity score;

[0015] A parameter optimization module, configured to perform process parameter optimization based on the defect type determination result and the defect severity score to obtain a device control strategy;

[0016] A transmission module, configured to transmit the device control strategy to each actuator to perform closed-loop feedback, obtaining a motor control signal, a servo positioning command, a contour adjustment command, and management optimization decision-making information.

[0017] In a third aspect, a computer-readable storage medium is provided. Instructions are stored in the computer-readable storage medium, and when running on a computer, cause the computer to execute the above-mentioned brick quality monitoring method based on data analysis.

[0018] In the technical solution provided by this application, the present invention configures a multi-sensor network on the brick production line to collect multi-dimensional data such as vibration, pressure, displacement, image, and air pressure, forming a complete production process data set. Compared with the traditional monitoring method that only relies on a single parameter, it realizes the comprehensive monitoring of the entire brick production process and effectively avoids misjudgment caused by abnormal single parameters. The present invention applies advanced algorithms such as domain synchronous averaging to process vibration signals, Canny edge detection to analyze images, and Fourier descriptors to extract contour features, which can accurately identify the defect characteristic frequencies, contour feature vectors, and quality anomaly indicators on the brick surface, greatly improving the sensitivity and accuracy of brick defect detection. The present invention inputs the extracted features into a convolutional neural network for in-depth analysis. Through multi-layer feature extraction and classification regression, it can not only accurately determine various different types of brick defects but also give a defect severity score, providing an accurate basis for subsequent process parameter optimization. Based on the defect analysis results, the present invention constructs an association model between brick quality and process parameters, searches in the subspace of high-contribution parameters through the differential evolution algorithm, and designs specific compensation strategies for different types of defects, realizing the precise dynamic adjustment of process parameters during the brick production process. The present invention adopts a three-level control architecture of the device layer, process layer, and production management layer, realizing hierarchical collaboration from basic control to advanced decision-making. Through the iterative contour error compensation algorithm and dynamic production plan adjustment strategy, it ensures the accurate execution of control instructions. At the same time, it establishes a three-level alarm mechanism of early warning, warning, and emergency warning, improving the stability and security of the system. Through the calculation of control performance indicators and parameter dynamic adjustment, the present invention forms a complete closed-loop of data collection, analysis and processing, defect diagnosis, parameter optimization, and control execution, enabling the brick production system to continuously self-optimize, effectively respond to internal changes and external disturbances, and significantly improving the quality stability and production efficiency of brick production. Description of the Drawings

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0020] Figure 1 It is a schematic diagram of an embodiment of the brick quality monitoring method based on data analysis in the embodiments of this application;

[0021] Figure 2 It is a schematic diagram of an embodiment of the brick quality monitoring system based on data analysis in the embodiments of this application. Detailed Embodiments

[0022] The embodiments of the present application provide a method, a system and a storage medium for monitoring the quality of bricks based on data analysis. The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and the above-mentioned drawings of the present application are used to distinguish similar objects and do not necessarily need to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order different from that illustrated or described here. In addition, the term "comprising" or "having" and any variation thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0023] For ease of understanding, the specific process of the embodiments of the present application will be described below. Please refer to Figure 1 , an embodiment of the method for monitoring the quality of bricks based on data analysis in the embodiments of the present application includes:

[0024] Step S101: Collect vibration, pressure, displacement, image and air pressure data on the brick production line to obtain a data set;

[0025] It can be understood that the execution subject of the present application can be a system for monitoring the quality of bricks based on data analysis, or a terminal or a server. Specifically, it is not limited here. The embodiments of the present application will be described by taking the server as the execution subject as an example.

[0026] Specifically, a number of vibration sensors are evenly installed on the surface of the conveyor belt of the brick production line. These sensors are evenly distributed in the length direction of the conveyor belt to capture the vibration signals generated during the operation of the conveyor belt. The vibration signals reflect the dynamic changes of the bricks during transportation, including the force condition of the conveyor belt, the stability of the brick movement, and the interaction characteristics between the conveyor belt and the bricks. A number of pairs of pressure sensors are respectively installed on the first extrusion plate and the second extrusion plate on the production line to monitor the force state of the extrusion plates in real time, obtain the extrusion force on the bricks, the uniformity of the material during the pushing process, and the pressure change of the pushing component, which helps to analyze the force characteristics of the bricks during the processing. At the same time, the motion state of the vibrating bar is monitored by a displacement sensor. The displacement sensor is fixed at the bottom of the vibrating bar to accurately collect the displacement change data of the vibrating bar at different time points. This data reflects the displacement amplitude, motion frequency, and the change process of the vibration state of the vibrating bar. This information helps to analyze the relationship between the surface defects of the bricks and the vibration frequency. In terms of visual data acquisition, two high-speed industrial cameras are installed at the end of the conveyor belt. These two cameras take multi-angle images of the brick surface from different angles to capture the image data of the brick surface. The image data can reflect the contour characteristics, surface defects, and geometric shape deviations of the bricks. In order to obtain the dynamic pressure change inside the cavity, a barometric pressure sensor is installed inside the cavity. This sensor continuously monitors the dynamic data of the air pressure inside the cavity changing with time, providing information for analyzing the material flow, air pressure fluctuation, and internal environment change during the brick production process. Time synchronization processing is performed on the vibration signals, force data, displacement data, brick surface images, and barometric pressure data to ensure the accurate matching of different types of data at the same time point. By adding a unified time stamp to each type of data during the data acquisition process and performing time alignment and synchronization during data storage, a multi-modal data set is formed.

[0027] Step S102: Perform domain synchronous averaging and edge detection on the data set to obtain the characteristic frequency of the brick surface defects, the contour feature vector of the bricks, and the quality anomaly index;

[0028] Specifically, the vibration signals collected from the dataset are segmented according to the operating cycle of the conveyor belt. By averaging the vibration signals of each cycle, the domain synchronous average signal of the periodic defects on the brick surface is obtained. The domain synchronous average signal can eliminate random noise and non-periodic interference, thereby extracting the characteristic signals related to the periodic defects on the brick surface. The residual signal is extracted from the synchronous average signal, and the envelope spectrum of the residual signal is calculated. The characteristic frequencies of the envelope spectrum can accurately characterize the characteristics of the brick surface defects. In terms of image processing, the brick surface images in the dataset are preprocessed, including operations such as gray conversion, histogram equalization, and image enhancement, to improve the contrast and detail features of the images. After image preprocessing, the Canny edge detection algorithm is used to detect the edges of the target surface image, and the edge contour information of the brick is extracted from it. The Fourier descriptor is used to analyze the characteristics of the extracted brick contour, and the shape characteristics of the 1st to 20th orders are calculated. These shape characteristics can reflect the geometric shape changes of the brick. And the Hausdorff distance between the actual contour and the standard contour is calculated. This distance can quantify the error between the brick contour and the standard contour, forming a brick contour feature vector. Feature extraction is performed on the air pressure data and force data. For the air pressure data, by calculating its first derivative and second derivative, the inflection points, extreme points, and stable intervals of the internal pressure change of the cavity are identified. These features can reflect important information such as material flow, air pressure fluctuation, and cavity stability during the brick production process. At the same time, for the force data, torque analysis is performed to analyze the force conditions of the pusher component at different time points, and the force imbalance degree is calculated. The force imbalance degree reflects the balance degree of the forces between the first pressing plate and the second pressing plate during the pusher process. This parameter is directly related to the uniformity of the brick material distribution. The characteristic frequencies of the brick surface defects, the brick contour feature vector, and the brick material distribution characteristics are normalized, and different sources of data are integrated into the target feature vector through feature fusion. The feature fusion process ensures the consistency of various types of data in the same feature space and eliminates the scale differences brought by different data types. The target feature vector is detected by setting an anomaly score threshold through the isolation forest algorithm. The isolation forest algorithm constructs multiple randomly divided decision trees, calculates the anomaly score for each sample, and sets a threshold based on the anomaly score to distinguish normal samples and abnormal samples. By detecting the target feature vector, various types of quality anomalies occurring during the brick production process are identified, forming a quality anomaly index.

[0029] For the barometric pressure data, noise suppression and smoothing processing are carried out. By introducing the methods of moving average filtering or Gaussian filtering, the random noise in the signal is eliminated, and the smoothed barometric pressure data sequence is obtained. The five-point central difference method is used to calculate the derivative of the barometric pressure data sequence. The first-order derivative and the second-order derivative are calculated through the weighted difference of five adjacent points, and the derivative matrix representing the barometric pressure change rate and acceleration in the cavity is obtained. The first-order derivative reflects the speed of the barometric pressure changing with time, while the second-order derivative can reveal the acceleration characteristics of the barometric pressure change. Feature point detection is performed on the derivative matrix to identify the key barometric pressure change feature points, including inflection points, extreme points, and stable intervals. By identifying the inflection points at the intersection of the first-order derivative and the second-order derivative, and judging the extreme points according to the sign change of the second-order derivative, and at the same time identifying the stable region by analyzing the interval where the first-order derivative is close to zero, a set of barometric pressure change feature points is constructed. These feature points reflect the key stages of the barometric pressure change in the brick production process. At the same time, for the force data in the dataset, spectral analysis is carried out to extract the force amplitudes on the first extrusion plate and the second extrusion plate. The force data is transformed from the time domain to the frequency domain through Fourier transform, so as to identify the main frequency components and amplitude information of the force signal on the extrusion plate. This spectral analysis process can reveal the periodic changes in the force signal and the energy distribution at different frequencies. The extracted force amplitudes on the extrusion plate are used to construct the force time-series characteristics. Based on the force time-series characteristics, the force imbalance is calculated. By comparing the forces on the first extrusion plate and the second extrusion plate, the balance of the forces on both sides during the material pushing process is evaluated, and a curve of the force imbalance changing with time is plotted to obtain the force distribution characteristic map, which reflects the uniformity of the material distribution and the dynamic change of the extrusion force during the brick material pushing process. The set of barometric pressure change feature points and the force distribution characteristic map are subjected to time-synchronized superposition analysis. Based on the timestamps, the two sets of data are aligned and ensured to be matched on the same time axis. Through the weighted fusion algorithm, the barometric pressure feature points and the force imbalance features are comprehensively quantified. The weighted fusion algorithm dynamically adjusts the contribution degree of the feature points according to the influence weights of each feature point on the material distribution uniformity, and calculates the material distribution uniformity index. The closer the material distribution uniformity index is to the ideal value, the more uniform the material distribution is. On the contrary, it indicates that there are abnormalities or non-uniformities in the material distribution.

[0030] Step S103: Input the brick surface defect characteristic frequency, the brick contour feature vector, and the quality anomaly index into a convolutional neural network for analysis to obtain the brick defect type determination result and the defect severity score;

[0031] Specifically, the brick surface defect characteristic frequencies, brick contour feature vectors, and quality anomaly indicators are combined into a unified feature matrix. The feature matrix is normalized and standardized. By performing mean-variance normalization or min-max scaling on each column of features, the data is transformed into the same numerical range, eliminating the impact of scale differences between different features on model training, and obtaining the normalized network input data for neural network analysis. The normalized network input data is input into a convolutional neural network for deep feature extraction. The structure of the convolutional neural network consists of three convolutional layers, and the convolutional kernel sizes of each convolutional layer are 3×3, 5×5, and 7×7 respectively. Convolutional kernels of different scales can capture local patterns of different scales in the brick feature matrix, thereby extracting more representative deep mapping features. The convolutional layer performs layer-by-layer convolution on the input data through a sliding window and uses a non-linear activation function for feature mapping to obtain the deep representations of brick surface defects, contour features, and anomaly indicators in the high-dimensional feature space. Through convolutional operations, the local correlations of the feature matrix are effectively extracted, improving the model's ability to express complex features. The deep mapping features are input into two max-pooling layers of the convolutional neural network for dimensionality reduction. The pooling window size of the max-pooling layer is 2×2. Through pooling operations, the information in the feature space is reduced in dimension and compressed, thereby retaining key features and removing redundant information, minimizing the feature dimension while retaining the ability to express brick defect features, and obtaining the brick feature representation vector. The brick feature representation vector is input into the defect classifier of the convolutional neural network. Through this classifier, F predefined brick defect types are classified. The output layer of the classifier uses the softmax activation function to calculate the probability distribution of each defect type, and optimizes and adjusts the classification error of the model through the cross-entropy loss function. The softmax activation function can convert the multiple class scores output by the network into a probability distribution, making the sum of the predicted probabilities of each class equal to 1, so as to determine the defect type of the brick according to the maximum probability. At the same time, by calculating the cross-entropy loss function to evaluate the difference between the model prediction result and the true label, and updating the gradient of the model through the backpropagation algorithm, the performance of the classification model is continuously optimized, and the recognition accuracy of different brick defect types is improved. To achieve the scoring of the defect severity, the brick feature representation vector is simultaneously input into the modified output layer of the convolutional neural network. This modified output layer is different from the defect classifier and uses a regression network with the sigmoid activation function. Through the sigmoid activation function, the network output is restricted between 0 and 1, obtaining a continuous score representing the defect severity. This regression network optimizes the error of the defect severity score through the root mean square error loss function, and adjusts the network parameters by comparing the mean square error between the predicted score output by the model and the true defect severity label to improve the accuracy of defect severity prediction.Through the joint training of the convolutional neural network, the determination result of the defect type of the brick and the defect severity score are obtained simultaneously.

[0032] Step S104: Perform process parameter optimization based on the defect type determination result and the defect severity score to obtain an equipment control strategy;

[0033] Specifically, based on the defect type determination result and the defect severity score, correlation analysis is carried out between the brick quality and process parameters. During the analysis process, defect characteristics are matched with key process parameters to obtain a process parameter vector containing multiple key process parameters. This parameter vector includes the conveyor belt speed, the initial positions of the first and second extrusion plates, the spring pressure, the baffle height, the vibration bar height, and the opening and closing time of the electronic stop valve. These process parameters directly affect the distribution state of materials, the forming accuracy, and the generation of surface defects during the brick production process. Based on the process parameter vector, a mathematical model for optimizing process parameters is constructed. By establishing a mapping relationship between the brick quality objective function and process parameters, the model quantifies the influence of different parameters on the quality objective, and calculates the contribution degree of each process parameter to the quality objective through sensitivity analysis to generate a parameter contribution degree matrix. This matrix reflects the comprehensive influence of different process parameters on brick quality, production efficiency, and material utilization rate. By analyzing the contribution degree matrix, the process parameters are divided into a high contribution degree group and a low contribution degree group. High contribution degree parameters have a stronger influence on the quality objective. Therefore, during the parameter optimization process, the differential evolution algorithm is executed to search in the subspace composed of high contribution degree parameters. The differential evolution algorithm continuously iterates and updates the individuals in the population, combines mutation, crossover, and selection operations to find the optimal solution in the high-dimensional parameter space, and obtains the optimal process parameter configuration. During the optimization process, the objective function value is calculated after each iteration, and the search range is dynamically adjusted based on the improvement degree of the quality objective to ensure that the final parameter combination that meets the quality objective requirements is obtained. Based on the optimal process parameter configuration, corresponding compensation strategies are designed for different types of brick defects to enable differential control according to the specific defect type during the production process. For polygon defects, the conveyor belt speed and the vibration bar height are adjusted to change the brick movement state, thereby compensating for the shape abnormality caused by material accumulation or uneven conveying; for contour errors, the initial positions of the first extrusion plate and the second extrusion plate and the partition position are adjusted to precisely control the contour characteristics during brick forming, thereby eliminating the contour deviation caused by uneven extrusion; for the problem of uneven material mixing, the opening and closing time of the electronic stop valve is adjusted to control the uniform distribution of materials in the cavity and prevent internal defects of bricks caused by material accumulation or uneven mixing. Through the defect-specific compensation scheme, the finished product rate and quality consistency of bricks are effectively improved. The defect-specific compensation scheme is converted into an equipment control strategy to ensure that the compensation strategy is accurately implemented at each equipment level.The device control strategy includes various specific control instructions. Among them, the speed control instruction adjusts the motor output torque through a PID controller to ensure that the conveyor belt speed reaches the expected optimized value; the position control instruction controls the positions of the first and second pressing plates through the precise positioning signal of the servo motor to achieve contour error compensation; the spring pressure control instruction adjusts the spring pressure through a pressure regulating mechanism to maintain the force balance during the material pushing process; the baffle height control instruction dynamically adjusts the baffle height by controlling the pneumatic system to ensure the best pressure state during the material conveying process; and the timing control instruction of the electronic stop valve controls the material flow by precisely setting the switching time to optimize the material distribution uniformity.

[0034] Step S105: Transmit the device control strategy to each actuator to perform closed-loop feedback, and obtain the motor control signal, servo positioning instruction, contour adjustment instruction, and management optimization decision information.

[0035] Specifically, the device control strategy is decomposed into three levels, including device-level control, process-level control, and production management-level control. Among them, the device-level control is responsible for directly controlling each actuator on the brick production line to ensure that the production equipment operates according to the set process parameters; the process-level control is responsible for quality monitoring and dynamic compensation of the production process based on real-time data to ensure that the brick quality meets the expected standards; the production management-level control focuses on the optimization and adjustment of the overall production plan, and conducts parameter optimization, plan adjustment, and quality warning based on the long-term production data trend. In the device-level control, the control strategy is decomposed into instructions, converted into specific control commands, and the instructions are transmitted to each actuator through a fieldbus (such as PROFINET or EtherCAT). This process constructs a device basic control network, which connects and controls actuators such as motors, extrusion plates, vibration components, electronic shut-off valves, and baffles to achieve real-time transmission and feedback of control instructions. In terms of motor control, a vector control algorithm is applied to precisely control the speed and torque of the motor. The vector control algorithm converts the three-phase current of the AC motor into DC components for decoupling control by constructing a three-phase current control model, thereby achieving precise adjustment of the motor speed and torque. This control signal dynamically adjusts the motor output through a closed-loop feedback mechanism to ensure that the conveyor belt speed reaches the optimal value. At the same time, for the extrusion plate position control, servo positioning control is implemented. By tracking the position of the servo motor of the pusher assembly, high-precision pusher positioning is achieved. The servo positioning instruction is based on the optimized initial position of the extrusion plate and combines real-time displacement feedback for position correction, so as to ensure that the extrusion plate always maintains the best position during the pusher process. In the process-level control, an iterative contour error compensation algorithm is used to dynamically adjust the contour error on the surface of the brick. The motion synchronization points of the pusher assembly and the vibration assembly are calculated. The calculation of the synchronization points is based on the principle of motion synchronization control to ensure that the pusher action and the vibration action are consistent in time and space. A brick contour error compensation function is designed based on the motion synchronization points. The compensation function generates a compensation signal by analyzing the change law of the contour error and superimposes the compensation signal on the original control signal, thereby achieving real-time compensation of the contour error. This process generates a contour adjustment instruction, which is transmitted to the pusher assembly and the vibration assembly through a servo control system to achieve precise control of the brick contour shape, thereby reducing the contour deviation and improving the contour accuracy of the finished bricks. In the production management-level control, a dynamic production plan adjustment strategy is used to set up conditional response mechanisms such as quality trend prediction, parameter self-optimization program, and emergency adjustment program. When the same type of defect appears in consecutive B bricks, the parameter self-optimization program is triggered to recalculate the optimal process parameter configuration and dynamically adjust the device control parameters, thereby achieving adaptive optimization of the production process. When a serious defect is detected, the emergency adjustment program is triggered. This program quickly corrects abnormal situations in the production process by means of suspending production, checking the equipment status, and resetting key parameters.According to the results of quality trend prediction, dynamically adjust the production plan and give early warnings of potential quality risks during the production process, so as to achieve preventive control. This dynamic production plan adjustment mechanism identifies potential quality fluctuations through trend analysis and prediction of historical production data, and adjusts key process parameters in advance to prevent the occurrence of quality anomalies. Through a multi-level control architecture of the equipment layer, process layer, and production management layer, closed-loop feedback control of the brick production process is achieved. The motor control signal adjusts the motor speed in real time through a vector control algorithm, the servo positioning instruction ensures precise position control of the extrusion plate and the pusher assembly, the contour adjustment instruction realizes dynamic compensation of contour errors based on the motion synchronization point, and the management optimization decision information ensures the stability and efficiency of the production process through mechanisms such as quality trend prediction, parameter self-optimization, and emergency adjustment.

[0036] In the embodiment of the present application, the present invention configures a multi-sensor network on the brick production line to collect multi-dimensional data such as vibration, pressure, displacement, image, and air pressure, forming a complete production process data set. Compared with the traditional monitoring method that only relies on a single parameter, it realizes comprehensive monitoring of the entire brick production process and effectively avoids misjudgment caused by abnormal single parameters. The present invention applies advanced algorithms such as domain synchronous averaging to process vibration signals, Canny edge detection to analyze images, and Fourier descriptors to extract contour features, and can accurately identify the characteristic frequencies of brick surface defects, contour feature vectors, and quality anomaly indicators, greatly improving the sensitivity and accuracy of brick defect detection. The present invention inputs the extracted features into a convolutional neural network for in-depth analysis. Through multi-layer feature extraction and classification regression, it can not only accurately determine various types of brick defects, but also give defect severity scores, providing an accurate basis for subsequent process parameter optimization. Based on the defect analysis results, the present invention constructs an association model between brick quality and process parameters, searches in the subspace of high contribution degree parameters through a differential evolution algorithm, and designs specific compensation strategies for different types of defects, realizing precise dynamic adjustment of process parameters during the brick production process. The present invention adopts a three-level control architecture of the equipment layer, process layer, and production management layer, realizing hierarchical collaboration from basic control to high-level decision-making. Through an iterative contour error compensation algorithm and a dynamic production plan adjustment strategy, it ensures the precise execution of control instructions. At the same time, a three-level alarm mechanism of early warning, warning, and emergency warning is established, improving the stability and security of the system. Through the calculation of control performance indicators and dynamic parameter adjustment, the present invention forms a complete closed loop of data collection, analysis and processing, defect diagnosis, parameter optimization, and control execution, enabling the brick production system to continuously self-optimize, effectively respond to internal changes and external disturbances, and significantly improving the quality stability and production efficiency of brick production.

[0037] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0038] Uniformly install n vibration sensors on the surface of the conveyor belt on the brick production line, and collect vibration signals characterizing the dynamic characteristics of the conveyor belt through the vibration sensors;

[0039] Install m pairs of pressure sensors on the first extrusion plate and the second extrusion plate on the brick production line respectively, and collect force data characterizing the stress state of the extrusion plates through the pressure sensors;

[0040] Fix the displacement sensor at the bottom of the vibrating bar, and collect displacement data characterizing the motion state of the vibrating bar through the displacement sensor;

[0041] Collect multi-angle surface images of the bricks through two high-speed industrial cameras installed at the end of the conveyor belt, and use the air pressure sensor installed inside the cavity to collect air pressure data characterizing the dynamic change of the air pressure inside the cavity;

[0042] Perform time synchronization and storage on the vibration signals, force data, displacement data, brick surface images and air pressure data to obtain a data set.

[0043] Specifically, according to the length of the conveyor belt and the actual situation of the brick conveying path, determine the installation quantity n of vibration sensors and their evenly distributed positions. The vibration sensors are installed at different positions on the surface of the conveyor belt to ensure real-time monitoring of the dynamic characteristics of the conveyor belt during the brick transportation process, including key information such as the vibration amplitude, frequency change, and abnormal vibration state of the conveyor belt. By evenly distributing the vibration sensors in the length direction of the conveyor belt, ensure that the data covers all key stages of the entire conveying process, preventing inaccurate monitoring results caused by local data loss. The vibration signals collected by the vibration sensors reflect the force condition of the conveyor belt, the friction change between the bricks and the conveyor belt during the conveying process, and the uniformity of material distribution. At the same time, install pressure sensors at the key force-bearing parts of the first extrusion plate and the second extrusion plate of the pusher assembly, and ensure that their positions can cover the force-bearing area. The installation positions of the pressure sensors are accurately calculated according to the force distribution of the extrusion plates, so as to accurately monitor the acting forces received by the extrusion plates during the pusher process. Each pair of pressure sensors is respectively installed at the symmetrical positions of the first extrusion plate and the second extrusion plate. By measuring the force condition on the extrusion plates during the pusher process in real time, analyze whether the force on the extrusion plates is uniform, and monitor the problem of unbalanced force occurring during the pusher process. The force data collected by the pressure sensors reflect the magnitude of the acting force received by the material during the extrusion process, the change of the pusher pressure, and the fluctuation of the material density and hardness. Fix the displacement sensor at the bottom of the vibrating bar, and collect displacement data characterizing the motion state of the vibrating bar through the displacement sensor. The installation position of the displacement sensor is fixed at the center position of the bottom of the vibrating bar to capture the displacement change information of the vibrating bar at different time points in real time. Monitor the motion state of the vibrating bar through the displacement sensor, and obtain the displacement amplitude, motion period, and transient change situation of the vibrating bar. These data help to analyze the relationship between the operating state of the vibrating bar and the surface defects of the bricks. At the same time, the displacement data reflects the abnormal vibration of the vibrating bar during the pusher process. Install a high-speed industrial camera on both sides of the end of the conveyor belt to image the surface of the bricks from different angles respectively. The high-speed industrial camera has a high frame rate and resolution to ensure capturing the detailed features of the brick surface during the high-speed movement process and avoiding the decline of image quality caused by motion blur. The installation angle of the camera is accurately adjusted according to the geometric shape of the brick surface to cover different areas of the brick surface, so as to ensure obtaining a complete image of the brick surface. Through multi-angle image acquisition, extract the contour information, edge features, and surface defect information of the brick surface. At the same time, in order to monitor the pressure change inside the cavity, use a barometric pressure sensor installed inside the cavity to collect barometric pressure data characterizing the dynamic pressure change inside the cavity, reflecting the pressure fluctuation inside the cavity during the pusher process, the density change of the material, and the expansion and compression of the gas. These data help to analyze the uniformity of material distribution and the stability of brick forming. Synchronize and store the vibration signals, force data, displacement data, brick surface images, and barometric pressure data in time to obtain a data set.Use a high-precision time synchronization mechanism to add a unified timestamp to all data acquisition devices and align different data types in terms of time. The time synchronization mechanism adopts a high-precision network clock protocol (such as PTP) or GPS time synchronization technology to uniformly time-stamp the data of each sensor, ensuring that data from different sources are stored under the same time reference. The vibration signals, force data, displacement data, brick surface images, and air pressure data collected after time synchronization are transmitted to the central data acquisition controller through a high-speed communication bus and stored in the data management system. The data storage format adopts a block storage method, where the data for each sampling period is block-marked and stored as a complete data set.

[0044] In a specific embodiment, the process of performing step S102 may specifically include the following steps:

[0045] Segment the vibration signals in the data set according to the conveyor belt operation cycle and calculate the synchronous average signal for N cycles to obtain the domain synchronous average signal characterizing the periodic defects on the brick surface;

[0046] Extract the residual signal from the domain synchronous average signal and calculate its envelope spectrum to obtain the defect characteristic frequency on the brick surface;

[0047] Perform grayscale conversion, histogram equalization, and image enhancement on the brick surface images in the data set to obtain the target surface images, and perform Canny edge detection on the target surface images to obtain the brick contours;

[0048] Use Fourier descriptors to calculate the shape features of the 1st to 20th orders for the brick contours and calculate the Hausdorff distance between the actual contour and the standard contour to obtain the brick contour feature vector;

[0049] Calculate the first derivative and the second derivative based on the air pressure data in the data set and extract the inflection points, extreme points, and stable intervals. At the same time, perform moment analysis through the force data in the data set and calculate the force imbalance degree to obtain the brick material distribution characteristics;

[0050] Perform normalization processing and feature fusion on the brick surface defect characteristic frequency, brick contour feature vector, and brick material distribution characteristics to obtain the target feature vector, and detect the target feature vector by setting an anomaly score threshold through the isolation forest algorithm to obtain the quality anomaly index.

[0051] Specifically, preprocess the vibration signals collected in the data set and segment the vibration signals based on the operation cycle T of the conveyor belt. The operation cycle of the conveyor belt is calculated by monitoring the speed signal of the conveyor belt or according to the time interval when the brick passes through a specific sensor position. The vibration signals are segmented according to the cycle T, and each segment contains a fixed number of sampling points. The segmented signal is denoted as Vb(t) segmented sequence. To eliminate random noise and periodic interference, after obtaining the segmented signals of N cycles, synchronous averaging calculation is performed on the signals of all cycles. The signals of each cycle are accumulated and averaged to obtain the domain synchronous average signal, which can effectively retain the features related to the periodic defects on the brick surface while suppressing the non-periodic noise interference. Extract the residual signal from the domain synchronous average signal and calculate its envelope spectrum to obtain the defect characteristic frequencies on the brick surface. The calculation of the residual signal is performed by taking the difference between the original vibration signal and the domain synchronous average signal to obtain the residual signal that has nothing to do with the periodic characteristics. These residual signals contain the high-frequency characteristic information generated by the defects on the brick surface. By performing envelope spectrum analysis on the residual signal, the signal is transformed into the frequency domain and the characteristic frequencies related to the brick surface defects are extracted. These characteristic frequencies reflect the periodic changes of the polygon defects, contour errors, and microscopic defects on the brick surface. At the same time, preprocess the brick surface images in the dataset, including grayscale conversion, histogram equalization, and image enhancement, to obtain the target surface images. Convert the original brick image from the RGB color space to a grayscale image to eliminate the interference of color information and thus focus more on the edge features of the image. Adjust the grayscale distribution of the image by the histogram equalization method to enhance the contrast and detail features of the image, making the defects and contours on the brick surface more prominent. Apply image enhancement technology to process the image using adaptive filtering or edge enhancement algorithms to make the boundaries on the brick surface clearer. After the preprocessing operations, input the target surface image into the Canny edge detection algorithm. By calculating the gradient of the image and performing double-threshold screening, extract the contour edges of the brick surface and generate a binary image reflecting the contour characteristics of the brick. After obtaining the contour information of the brick, use Fourier descriptors to extract the features of the brick contour and calculate the shape features of the 1st to 20th order. Fourier descriptors are a mathematical tool for transforming contour shape information into the frequency domain for analysis. Convert the extracted contour boundary curve into a set of Fourier coefficients and describe the geometric shape characteristics of the brick by analyzing the first 20 Fourier coefficients. Fourier descriptors capture the overall shape information of the contour and the changes in local features. Quantify the size of the brick contour error by calculating the Hausdorff distance between the actual contour and the standard contour. The Hausdorff distance is a metric for calculating the maximum-minimum distance between two point sets, measuring the similarity between the actual contour and the standard contour, and generating a brick contour feature vector. At the same time, calculate the first derivative and the second derivative according to the air pressure data in the dataset, and extract the inflection points, extreme points, and stable intervals to characterize the dynamic change characteristics of the air pressure inside the cavity. After smoothing the air pressure data, use the central difference method to calculate the derivative of the air pressure data. The first derivative reflects the air pressure change rate, and the second derivative reflects the acceleration of the air pressure change.The extraction of inflection points is judged based on the sign changes of the first derivative and the second derivative. The identification of extreme points is determined by judging the points where the derivative is zero, and the identification of stable intervals is extracted by analyzing the time periods when the air pressure change rate is close to zero. At the same time, torque analysis is performed on the force data in the dataset, and the force imbalance degree is calculated to characterize the characteristics of the brick material distribution. Torque analysis is achieved by calculating the force changes of the first extrusion plate and the second extrusion plate during the material pushing process. The force imbalance degree is calculated by comparing the force differences between the two extrusion plates, which effectively reflects the uniformity of the material distribution during the material pushing process, thereby providing a basis for analyzing the internal material state of the brick. The surface defect characteristic frequencies of the brick, the brick contour feature vectors, and the brick material distribution characteristics are normalized, and are integrated into a target feature vector through feature fusion. The normalization process linearly scales the feature data to map data of different dimensions to the same numerical range, thereby eliminating the differences between different feature scales and preventing high-dimensional features from causing biases in the calculation results. Feature fusion is to combine different types of normalized features, and through methods such as feature splicing or weighted fusion, integrate the information of different features into a unified feature vector to form a brick feature representation. The target feature vector is input into the isolation forest algorithm for anomaly detection. The isolation forest algorithm calculates the anomaly score of each sample by constructing multiple randomly divided decision trees, and judges whether the sample is abnormal according to the set anomaly score threshold. By detecting the target feature vector through the isolation forest algorithm, the quality anomaly situations occurring in the brick production process are identified, and the quality anomaly index is obtained.

[0052] In a specific embodiment, the process of performing steps to calculate the first derivative and the second derivative based on the air pressure data in the dataset and extract inflection points, extreme points, and stable intervals, and at the same time perform torque analysis on the force data in the dataset and calculate the force imbalance degree to obtain the brick material distribution characteristics may specifically include the following steps:

[0053] Perform noise suppression and smoothing processing on the air pressure data in the dataset to obtain a preprocessed air pressure data sequence;

[0054] Calculate the first derivative and the second derivative of the preprocessed air pressure data sequence through the five-point central difference method to obtain a derivative matrix representing the air pressure change rate and acceleration in the cavity;

[0055] Perform a feature point detection algorithm on the derivative matrix to identify inflection points, extreme points, and stable intervals, and obtain a set of air pressure change feature points;

[0056] Perform spectrum analysis on the force data in the dataset and extract the force amplitudes of the first extrusion plate and the second extrusion plate to obtain the force time series characteristics of the extrusion plate;

[0057] Calculate the force imbalance degree based on the force-time sequence characteristics of the extrusion plate and draw the change curve of the force imbalance degree over time to obtain the force distribution characteristic map;

[0058] Perform time-synchronous superposition analysis on the set of air pressure change characteristic points and the force distribution characteristic map, and comprehensively quantify the material distribution uniformity index through a weighted fusion algorithm to obtain the brick material distribution characteristics.

[0059] Specifically, the collected air pressure data is denoised and filtered. Since the air pressure data is affected by factors such as equipment vibration, environmental fluctuations, and material flow during brick production, the original data is accompanied by high-frequency noise and abnormal fluctuations. In the data preprocessing stage, methods such as moving average filtering or Gaussian filtering are used to smooth the air pressure data. The mean value of the data is calculated through a moving window or the adjacent data is weighted averaged using Gaussian weights to remove high-frequency noise within a short period and retain the main trend of air pressure changes. At the same time, to avoid data distortion, by setting a reasonable filtering window length, the smoothed air pressure data can truly reflect the change process of the air pressure in the cavity, and a smoothed air pressure data sequence after noise suppression is obtained. The first derivative and the second derivative are calculated for the preprocessed air pressure data sequence by the five-point central difference method to obtain a derivative matrix representing the air pressure change rate and acceleration in the cavity. The five-point central difference method is a numerical differentiation method that calculates the derivative of data through five adjacent time points, thereby improving the accuracy of differential calculation. For the calculation of the first derivative, the five-point central difference formula is used to process the air pressure data sequence to accurately reflect the change rate of the air pressure in the cavity over time, while the calculation of the second derivative describes the acceleration characteristics of the air pressure change. The first derivative reflects the speed of the air pressure change in the cavity, that is, the air pressure change rate, revealing the transient fluctuations generated by the air pressure during the material feeding, vibration, and material flow processes, while the second derivative reflects the acceleration of the air pressure change, capable of identifying the inflection points, extreme points, and stable intervals in the air pressure change process, thereby constructing a derivative matrix. The feature point detection algorithm is executed on the derivative matrix to identify the inflection points, extreme points, and stable intervals, obtaining a set of air pressure change feature points. The core of feature point detection lies in analyzing the change trends of the first derivative and the second derivative. By detecting the points where the first derivative is zero and the sign of the second derivative changes, the extreme points in the air pressure change process are identified, and these extreme points reflect the local maximum and minimum values of the air pressure in the cavity during the material feeding and forming processes. By analyzing the zero-crossing points of the second derivative, the inflection points are identified. The inflection points correspond to the critical points where the air pressure change accelerates or decelerates, and these points are the turning points of the air pressure change inside the cavity. By identifying the time periods when the first derivative is close to zero, the stable intervals of the air pressure change are detected, and these stable intervals correspond to the time periods when the air pressure in the cavity reaches an equilibrium state, reflecting the process of the material distribution tending to be stable. Through the feature point detection of the derivative matrix, a set of air pressure change feature points is obtained. At the same time, spectrum analysis is performed on the force data in the dataset and the force amplitudes of the first extrusion plate and the second extrusion plate are extracted to obtain the force time series characteristics of the extrusion plates. During the extrusion process, the force states of the first extrusion plate and the second extrusion plate directly affect the material distribution and the uniformity of brick forming. By performing spectrum analysis on the force data, the main frequency components of the force signals on the extrusion plates are identified, and the corresponding amplitude characteristics are extracted.Spectrum analysis converts the time-domain signal to the frequency domain through Fourier transform, revealing the hidden periodic characteristics in the force signal and the force change amplitudes corresponding to different frequency components, thereby identifying the characteristic frequencies related to the pusher cycle, material fluctuations, and vibration signals. By extracting the force amplitudes of the first extrusion plate and the second extrusion plate and combining the time series information, the time-series characteristics of the extrusion plate force are generated. Based on the extracted force time-series characteristics, the force imbalance is calculated and the curve of the force imbalance changing with time is plotted to obtain the force distribution characteristic map. The force imbalance is calculated by comparing the force differences between the first extrusion plate and the second extrusion plate during the pusher process. The force imbalance reflects the non-uniformity of the force on the material during the pusher process and the distribution state of the pusher pressure. By analyzing the force differences at different time points and plotting the curve of the force imbalance changing with time, the force change trend of the material at different pusher stages is shown, thereby constructing the force distribution characteristic map. The set of air pressure change characteristic points and the force distribution characteristic map are subjected to time-synchronous superposition analysis. The air pressure change characteristic points and the force distribution characteristic map are aligned on the time axis. By adding a unified timestamp to different data types, it is ensured that various types of data are matched under the same time reference, realizing the time synchronization of the characteristic point set and the force characteristics. After completing the time synchronization, a weighted fusion algorithm is used to comprehensively quantify the air pressure characteristic points and the force characteristics. The weighted fusion algorithm dynamically adjusts the contribution degrees of the characteristic points and the characteristic map according to the influence weights of different characteristics on the material distribution uniformity, and comprehensively calculates the material distribution uniformity index by means of weighted summation of the characteristics. The material distribution uniformity index is a quantitative index that reflects the uniform distribution state of the material inside the cavity during the brick production process. The closer the index is to the ideal value, the more uniform the material distribution is, otherwise it indicates that there are abnormalities or non-uniformities in the material distribution. Through the above steps, the brick material distribution characteristics are finally obtained.

[0060] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0061] The brick surface defect characteristic frequency, the brick contour feature vector, and the quality anomaly index are combined into a unified feature matrix and subjected to normalization and standardization processing to obtain the network input data;

[0062] The network input data is input into the three-layer convolution structure of the convolutional neural network for feature extraction to obtain the deep mapping features;

[0063] The deep mapping features are input into the two max-pooling layers of the convolutional neural network for dimensionality reduction to obtain the brick feature representation vector;

[0064] Input the brick feature representation vector into the defect classifier of the convolutional neural network, and calculate the determination result of the brick defect type through the defect classifier corresponding to the predefined F types of brick defects and using the softmax activation function and the cross-entropy loss function;

[0065] Input the brick feature representation vector into the modified output layer of the convolutional neural network, and perform the root mean square error loss calculation through the regression network with the sigmoid activation function in the modified output layer to obtain the defect severity score.

[0066] Specifically, feature fusion is performed on the features obtained from different data sources, integrating the brick surface defect feature frequencies, brick contour feature vectors, and quality anomaly indicators into a unified feature matrix. The brick surface defect feature frequencies are frequency-domain features obtained by performing domain synchronous averaging and residual signal envelope spectrum calculation on vibration signals. These features reflect the periodic characteristics of polygon defects and contour errors on the brick surface. The brick contour feature vectors are shape features obtained by performing Canny edge detection, Fourier descriptor calculation, and Hausdorff distance measurement on the brick surface image. These features describe the changes in the edge contour of the brick and the deviation between the actual contour and the standard contour. At the same time, the quality anomaly indicator is an indicator obtained by weighted fusion of the brick material distribution feature, force imbalance degree, and air pressure change feature points, reflecting the quality change trend during the brick production process. Unified normalization and standardization processing are performed on these features. By scaling the data ranges of different features to the same numerical interval, the differences in eigenvalue scales are eliminated, and through standardization processing, the feature data is converted into a standard distribution with a mean of zero and a variance of one, thereby ensuring the consistency of different features in the same feature space and obtaining the network input data required by the convolutional neural network. The network input data is input into the three-layer convolutional structure of the convolutional neural network for feature extraction. The three-layer convolutional structure of the convolutional neural network uses convolutional kernels of different sizes and performs convolutional operations on the input feature matrix using 3×3, 5×5, and 7×7 convolutional kernels respectively to capture local features at different scales and extract multi-dimensional feature information of brick surface defects, contour shapes, and quality anomalies. The convolutional operation performs point-by-point convolution on the input data through a sliding window, multiplies the local area of the input data element by element with the convolutional kernel, and sums them to generate a feature map matrix. After each convolutional layer, the ReLU activation function is introduced for non-linear transformation to enhance the expression ability of the network, and the convolutional result is standardized by adding a batch normalization layer, thereby accelerating network training and preventing gradient disappearance. After three-layer convolutional processing, the obtained feature map matrix contains deep mapping features of brick surface defects, contour features, and quality anomalies. The deep mapping features are input into the two max-pooling layers of the convolutional neural network for dimensionality reduction to obtain the brick feature representation vector. Max-pooling is a dimensionality reduction operation that slides a fixed-size window on the feature map matrix and takes the maximum value within the window, thereby retaining the main information of the features and removing redundant data. The pooling operation can effectively reduce the feature dimension, maintain the sparsity of the feature space, and enhance the generalization ability of the network. In the convolutional neural network, the two max-pooling layers use 2×2 and 3×3 pooling windows respectively to perform dimensionality reduction processing on the deep mapping features, converting the high-dimensional feature map matrix into a low-dimensional feature vector to obtain the brick feature representation vector with a fixed dimension. The brick feature representation vector is input into the defect classifier of the convolutional neural network, and the defect classifier classifies the predefined F types of brick defect types.The defect classifier consists of two fully connected layers. The fully connected layers perform linear transformations by multiplying the feature vectors with the weight matrices and adding the bias vectors, generating a probability distribution for defect classification. The output layer uses the softmax activation function to normalize the classification results into a probability distribution, converting the original classification scores into a probability distribution, and calculates the classification error of the model through the cross-entropy loss function. The softmax activation function can normalize the output results into probability values, making the sum of the predicted probabilities for each defect type equal to 1, so as to determine the defect type of the brick based on the maximum probability. The cross-entropy loss function measures the performance of the classification model by comparing the difference between the probability distribution predicted by the model and the true labels, and adjusts the network weights through the backpropagation algorithm to continuously optimize the classification accuracy. Through the training and optimization of the defect classifier, the determination result of the brick defect type is obtained. The brick feature representation vector is input into the modified output layer of the convolutional neural network, and the defect severity is scored through the modified output layer. The modified output layer adopts a regression network structure and outputs a continuous severity score through a single neuron with a sigmoid activation function. The sigmoid activation function limits the output results between 0 and 1, mapping the defect severity score to a fixed numerical interval, which is convenient for quantitatively comparing the severity of different defects. The training objective of the regression network is to minimize the root mean square error loss function, and optimize the network parameters by calculating the mean square error between the predicted severity score and the true severity label, so as to improve the prediction accuracy of the model for the defect severity. The root mean square error can effectively measure the prediction error of the regression model, and propagates the error to each layer of the neural network through backpropagation for gradient update, finally forming a regression model for quantifying the defect severity. After the model training is completed, the defect severity score is automatically generated according to the input feature data.

[0067] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0068] Based on the defect type determination result and the defect severity score, perform an association analysis of the brick quality and process parameters to obtain a process parameter vector including the conveyor belt speed, the initial position of the extrusion plate, the spring pressure, the baffle height, the vibration bar height, and the electronic stop valve switching time;

[0069] Based on the process parameter vector, construct a process parameter optimization mathematical model, and calculate the contribution degree of each process parameter to the quality target according to the process parameter optimization mathematical model to obtain a parameter contribution degree matrix;

[0070] According to the parameter contribution degree matrix, divide the process parameters into a high contribution degree group and a low contribution degree group, and perform a differential evolution algorithm search in the subspace composed of the high contribution degree parameters to obtain the optimal process parameter configuration;

[0071] Based on the optimal process parameter configuration, corresponding compensation strategies are designed for different types of brick defects. For polygon defects, the conveyor belt speed and the height of the vibrating bar are adjusted; for contour errors, the positions of the extrusion plate and the partition plate are adjusted; for the problem of uneven material mixing, the switching time of the electronic cut-off valve is adjusted to obtain a defect-specific compensation scheme.

[0072] The defect-specific compensation scheme is converted into an equipment control strategy, which includes speed control instructions for the PID controller to adjust the motor output torque, position control instructions for the servo motor control signal, spring pressure control instructions for the pressure regulating mechanism, baffle height control instructions for the pneumatic system, and timing control instructions for the electronic cut-off valve.

[0073] Specifically, the data from the quality monitoring system is deeply correlated with the historical production process data. Through multivariate regression analysis or feature selection methods based on machine learning, the key process parameters affecting the quality of bricks are identified, and the influence degree of each parameter on the quality target is quantified. These parameters include the conveyor belt speed, the initial position of the extrusion plate, the spring pressure, the baffle height, the vibration bar height, and the switching time of the electronic shut-off valve. The conveyor belt speed affects the movement stability of the bricks during transportation and the uniformity of material distribution, while the initial position of the extrusion plate determines the stress distribution state during the material forming process; the magnitude of the spring pressure directly affects the thrust of the extrusion plate and the material compaction effect, and the baffle height and the vibration bar height affect the flow rate and accumulation form of the material, thus acting on the shape and density of the bricks, and the switching time of the electronic shut-off valve determines the amount and rhythm of material feeding, which has an important impact on the material mixing uniformity. These key process parameters are combined into a process parameter vector. Based on the process parameter vector, a mathematical model for optimizing process parameters is constructed, and a parameter contribution matrix is formed by calculating the contribution degree of each process parameter to the quality target. The core of the mathematical model for optimizing process parameters is to establish a mathematical expression with the brick quality target as the optimization target. The objective function uses a linear regression model to describe the linear relationship between the parameters and the quality target, or a non-linear model based on neural networks or support vector machines to fit the complex relationship between parameter changes and quality changes. During the solution process, by slightly perturbing the process parameters and observing their influence on the quality target, the contribution degree of each process parameter to the brick quality target is calculated, forming a parameter contribution matrix. Each element in the matrix represents the contribution value of a specific parameter under different quality targets, and this matrix reflects which parameters have a higher influence during the quality target optimization process. According to the results of the parameter contribution matrix, the process parameters are divided into a high contribution group and a low contribution group, and the differential evolution algorithm is executed for search in the subspace composed of high contribution parameters to obtain the optimal process parameter configuration. The differential evolution algorithm is a population-based global optimization algorithm, suitable for solving optimization problems in multi-dimensional complex parameter spaces. By initializing a population containing multiple process parameter configurations, the differential evolution algorithm performs mutation, crossover, and selection operations on the population in each iteration, thus gradually approaching the optimal solution. During the optimization process, the algorithm focuses on searching in the subspace composed of high contribution parameters and dynamically adjusts the weights of different parameters according to the parameter contribution matrix, thereby improving the optimization efficiency and the accuracy of the solution. After multiple iterations, the differential evolution algorithm finds an optimal process parameter configuration that meets the quality target. Based on the obtained optimal process parameter configuration, corresponding compensation strategies are designed for different types of brick defects to ensure specific correction of different types of defects during the production process.For polygon defects, the motion state during material conveyance is changed by adjusting the conveyor belt speed and the height of the vibrating bar, thereby optimizing the shape characteristics of the brick surface; for contour errors, the influence of contour deviation during the material pushing process is compensated by adjusting the initial positions of the first extrusion plate and the second extrusion plate and the position of the partition plate, thereby improving the contour accuracy of the brick; for the problem of uneven material mixing, the feeding rhythm of the material is changed by adjusting the opening and closing time of the electronic shut-off valve, thereby optimizing the mixing uniformity of the material in the cavity. These compensation strategies are specifically adjusted according to the characteristics of different defect types to form a defect-specific compensation scheme. The defect-specific compensation scheme is converted into an equipment control strategy to achieve the precise execution of various control instructions. The equipment control strategy includes a variety of specific control instructions. Among them, the speed control instruction adjusts the motor output torque by a PID controller to precisely control the speed of the conveyor belt; the position control instruction adjusts the positions of the first extrusion plate and the second extrusion plate through the control signal of the servo motor to achieve dynamic compensation for contour errors; the spring pressure control instruction of the pressure regulating mechanism controls the force on the material pushing assembly through feedback regulation to ensure the uniformity of the force during the material pushing process; the baffle height control instruction adjusts the air pressure inside the cavity based on the feedback signal of the pneumatic system, thereby changing the height of the baffle to ensure the stability during the material conveyance process; and the timing control instruction of the electronic shut-off valve adjusts the opening and closing state of the shut-off valve through an accurate time control algorithm to ensure the uniformity of material flow. Through the synergistic effect of these control strategies, closed-loop control during the brick production process is achieved, enabling the production process to be adaptively adjusted according to real-time monitoring data, thereby improving production efficiency and quality stability.

[0074] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0075] Decompose the equipment control strategy into equipment layer control, process layer control, and production management layer;

[0076] Decompose the control strategies in the equipment layer control and transmit them to each actuator through the fieldbus to obtain the equipment basic control network;

[0077] Based on the equipment basic control network, construct a three-phase current control model for motor control by applying a vector control algorithm, and implement servo positioning for the extrusion plate position control to obtain motor control signals and servo positioning instructions;

[0078] Based on the iterative contour error compensation algorithm in the process layer control, calculate the motion synchronization points of the material pushing assembly and the vibrating assembly, and design a brick contour error compensation function to obtain contour adjustment instructions;

[0079] Based on the dynamic production plan adjustment strategy in production management control, set the condition response mechanism of the self-optimization program triggered by the same defects in consecutive B bricks, the emergency adjustment program triggered by serious defects, and quality trend prediction, and obtain the optimized decision-making information of the management level.

[0080] Specifically, according to the different control requirements in the brick production process, the control strategies are refined and classified by level. The equipment layer control is responsible for the basic operation of the equipment and the real-time instructions of the actuators, including the precise control of key components such as motors, extrusion plates, vibration components, electronic stop valves, and baffles. The process layer control is responsible for the real-time compensation and dynamic adjustment of quality errors occurring in the production process, optimizing the production process through data feedback. The production management layer control focuses on the management and optimization of the overall production plan, adaptively adjusting the process parameters according to the production data trend and quality inspection results, so as to ensure the dynamic balance between the production plan and the quality target. After decomposing the equipment control strategy into different levels, the response speed and stability of the control system are effectively improved, realizing the full-range dynamic control from the bottom-layer equipment control to the high-layer production management. In the equipment layer control, the control strategy is further decomposed into instructions and transmitted to each actuator through the fieldbus, thus constructing the basic equipment control network. The fieldbus can achieve high-speed data communication between different actuators, ensuring that the control instructions can be accurately sent to each equipment node within milliseconds. Through the decomposition of instructions, the complex control strategy is disassembled into specific control commands, and different types of control signals are generated according to the functions and control requirements of different equipment, such as motor torque control signals, servo motor position control signals, pressure regulation signals, baffle height adjustment signals, and on-off control signals of electronic stop valves. These control signals are transmitted to the corresponding actuators through the fieldbus to form the basic equipment control network, ensuring that all equipment on the production line can work in coordination. Based on the basic equipment control network, a three-phase current control model is constructed for motor control by applying the vector control algorithm, and servo positioning is implemented for the extrusion plate position control to obtain the motor control signal and the servo positioning instruction. The vector control algorithm decouples the three-phase current of the motor, converting the three-phase current of the AC motor into DC components, thereby realizing the independent control of the motor torque and flux. By constructing the three-phase current control model, the speed and torque of the motor are controlled to make the conveyor belt operate within the optimal speed range, and at the same time, the speed change caused by load fluctuations is avoided. The servo positioning control precisely adjusts the position of the extrusion plate through closed-loop feedback control. After comparing the feedback signal of the servo motor with the target position, an error signal is generated and adjusted in real time by the PID controller, so as to ensure that the position accuracy of the extrusion plate during the material pushing process reaches the control accuracy of ±0.1 mm. The motor control signal and the servo positioning instruction are sent to the actuator in real time through the basic equipment control network, ensuring high-speed and stable equipment layer control during the production process. In the process layer control, based on the iterative contour error compensation algorithm, the motion synchronization points of the material pushing component and the vibration component are calculated, and a brick contour error compensation function is designed to generate the contour adjustment instruction.The iterative contour error compensation algorithm analyzes and processes the motion data of the pusher component and the vibration component, calculates the motion synchronization points of the two under the set delay time τ, and adjusts the compensation strategy according to the differences in the motion trajectories. The calculation of the motion synchronization points is based on the kinematic model, ensuring a high degree of coordination between the pusher action and the vibration bar movement in terms of time and space, thereby eliminating the contour errors generated during the pusher process. Based on the motion synchronization points, a brick contour error compensation function is designed. This compensation function generates a set of dynamic compensation signals by analyzing the variation law of the contour errors. These compensation signals are superimposed on the original control signals of the pusher component and the vibration component, thus achieving real-time compensation for the contour errors. The contour adjustment instructions are sent to the actuators through the device basic control network to ensure high-precision contour control during the brick forming process and improve the quality and consistency of the brick surface. In the production management control, based on the dynamic production plan adjustment strategy, conditions response mechanisms for triggering the parameter self-optimization program when the same defects appear in consecutive B bricks, triggering the emergency adjustment program for serious defects, and predicting the quality trend are set, thereby generating management optimization decision-making information. The dynamic production plan adjustment strategy analyzes historical production data, real-time quality inspection results, and production targets, combines machine learning algorithms to predict future production quality trends, and dynamically adjusts the production plan according to the prediction results. When it is detected that the same type of defects appear in consecutive B bricks, the system automatically triggers the parameter self-optimization program, recalculates the optimal process parameter configuration, and dynamically adjusts the device control strategy to eliminate persistent defects. For detected serious defects, such as large-area damage on the brick surface or contour errors exceeding the set threshold, the system triggers the emergency adjustment program, suspends the production line and checks the device status, and at the same time adjusts the key process parameters to prevent the production of a large number of unqualified products. By analyzing the quality trend prediction results, potential quality fluctuation trends are identified, and key process parameters are adjusted in advance, thereby achieving preventive control of the production process. The management optimization decision-making information is sent to each actuator through the device basic control network, and combined with the dynamic adjustment strategies of the device layer control and the process layer control, a closed-loop control of the entire process is realized. The motor control signal, servo positioning instruction, contour adjustment instruction, and management optimization decision-making information cooperate with each other to ensure precise control and dynamic adjustment of each link in the brick production process, thereby achieving continuous optimization of the brick quality and improvement of the production efficiency.

[0081] The above describes the brick quality monitoring method based on data analysis in the embodiments of the present application. Next, the brick quality monitoring system based on data analysis in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the brick quality monitoring system based on data analysis in the embodiments of the present application includes:

[0082] An acquisition module 201, configured to acquire vibration, pressure, displacement, image, and air pressure data on the brick production line to obtain a data set;

[0083] The detection module 202 is used to perform domain synchronous averaging and edge detection on the data set to obtain the brick surface defect characteristic frequency, the brick contour feature vector, and the quality anomaly index;

[0084] The analysis module 203 is used to input the brick surface defect characteristic frequency, the brick contour feature vector, and the quality anomaly index into a convolutional neural network for analysis to obtain the brick defect type determination result and the defect severity score;

[0085] The parameter optimization module 204 is used to perform process parameter optimization based on the defect type determination result and the defect severity score to obtain the equipment control strategy;

[0086] The transmission module 205 is used to transmit the equipment control strategy to each actuator to perform closed-loop feedback to obtain the motor control signal, the servo positioning instruction, the contour adjustment instruction, and the management layer optimization decision information.

[0087] Through the collaborative cooperation of the above-mentioned various components, the present invention configures a multi-sensor network on the brick production line to collect multi-dimensional data such as vibration, pressure, displacement, image, and air pressure, forming a complete production process data set. Compared with the traditional monitoring method that only relies on a single parameter, it realizes the comprehensive monitoring of the entire brick production process and effectively avoids misjudgment caused by a single parameter anomaly. The present invention applies advanced algorithms such as domain synchronous averaging to process vibration signals, Canny edge detection to analyze images, and Fourier descriptors to extract contour features, which can accurately identify the brick surface defect characteristic frequency, the contour feature vector, and the quality anomaly index, greatly improving the sensitivity and accuracy of brick defect detection. The present invention inputs the extracted features into a convolutional neural network for in-depth analysis. Through multi-layer feature extraction and classification regression, it can not only accurately determine various different types of brick defects but also give the defect severity score, providing an accurate basis for subsequent process parameter optimization. Based on the defect analysis results, the present invention constructs an association model between brick quality and process parameters, searches in the high contribution degree parameter subspace through the differential evolution algorithm, and designs specific compensation strategies for different types of defects, realizing the precise dynamic adjustment of process parameters during the brick production process. The present invention adopts a three-level control architecture of the equipment layer, the process layer, and the production management layer to achieve hierarchical collaboration from basic control to advanced decision-making. Through the iterative contour error compensation algorithm and the dynamic production plan adjustment strategy, it ensures the accurate execution of control instructions. At the same time, a three-level alarm mechanism of early warning, warning, and emergency warning is established to improve the stability and security of the system. Through the calculation of control performance indicators and the dynamic adjustment of parameters, the present invention forms a complete closed-loop of data collection, analysis and processing, defect diagnosis, parameter optimization, and control execution, enabling the brick production system to continuously self-optimize, effectively cope with internal changes and external disturbances, and significantly improve the quality stability and production efficiency of brick production.

[0088] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions run on a computer, the computer is caused to execute the steps of the brick quality monitoring method based on data analysis.

[0089] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system, system, and unit can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0090] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions to cause a brick quality monitoring device based on data analysis (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0091] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of various embodiments of the present invention.

Claims

1. A method for monitoring the quality of bricks based on data analysis, characterized in that, Including: Collect vibration, pressure, displacement, image and air pressure data on the brick production line to obtain a data set; Perform domain synchronous averaging and edge detection on the data set to obtain the brick surface defect characteristic frequency, the brick contour feature vector and the quality anomaly index; Input the brick surface defect characteristic frequency, the brick contour feature vector and the quality anomaly index into a convolutional neural network for analysis to obtain the brick defect type determination result and the defect severity score; Execute process parameter optimization based on the defect type determination result and the defect severity score to obtain an equipment control strategy; Transmit the equipment control strategy to each actuator to execute closed-loop feedback to obtain motor control signals, servo positioning instructions, contour adjustment instructions and management optimization decision-making information.

2. The brick quality monitoring method based on data analysis according to claim 1, characterized in that The collecting of vibration, pressure, displacement, image and air pressure data on the brick production line to obtain a data set includes: Uniformly install n vibration sensors on the surface of the conveyor belt on the brick production line, and collect vibration signals characterizing the dynamic characteristics of the conveyor belt through the vibration sensors; Respectively install m pairs of pressure sensors on the first extrusion plate and the second extrusion plate on the brick production line, and collect force data characterizing the force state of the extrusion plate through the pressure sensors; Fix the displacement sensor at the bottom of the vibrating bar, and collect displacement data characterizing the motion state of the vibrating bar through the displacement sensor; Collect multi-angle brick surface images through two high-speed industrial cameras installed at the end of the conveyor belt, and use air pressure sensors installed inside the cavity to collect air pressure data characterizing the dynamic change of the internal pressure of the cavity; Perform time synchronization and storage on the vibration signal, the force data, the displacement data, the brick surface image and the air pressure data to obtain a data set.

3. The method for monitoring the quality of bricks based on data analysis according to claim 1, wherein The performing of domain synchronous averaging and edge detection on the data set to obtain the brick surface defect characteristic frequency, the brick contour feature vector and the quality anomaly index includes: Segment the vibration signals in the data set according to the conveyor belt operation cycle and calculate the synchronous average signal of N cycles to obtain the domain synchronous average signal characterizing the periodic defects on the brick surface; Extract the residual signal from the domain synchronous average signal and calculate its envelope spectrum to obtain the brick surface defect characteristic frequency; Perform gray conversion, histogram equalization and image enhancement on the brick surface images in the data set to obtain the target surface image, and perform Canny edge detection on the target surface image to obtain the brick contour; Use Fourier descriptors to calculate the shape features of the 1st to 20th orders of the brick contour and calculate the Hausdorff distance between the actual contour and the standard contour to obtain the brick contour feature vector; Calculate the first derivative and the second derivative according to the air pressure data in the data set and extract the inflection points, extreme points and stable intervals. At the same time, perform moment analysis through the force data in the data set and calculate the force imbalance degree to obtain the brick material distribution characteristics; Normalize and fuse the surface defect characteristic frequencies, contour feature vectors, and material distribution characteristics of the bricks to obtain a target feature vector, and detect the target feature vector by setting an anomaly score threshold using the Isolation Forest algorithm to obtain a quality anomaly index.

4. The method for monitoring the quality of bricks based on data analysis according to claim 3, characterized in that Calculate the first derivative and second derivative of the air pressure data in the dataset and extract inflection points, extreme points, and stable intervals. At the same time, perform torque analysis on the force data in the dataset and calculate the force imbalance degree to obtain the brick material distribution characteristics, including: Suppress noise and smooth the air pressure data in the dataset to obtain a preprocessed air pressure data sequence; Calculate the first derivative and second derivative of the preprocessed air pressure data sequence using the five-point central difference method to obtain a derivative matrix representing the air pressure change rate and acceleration in the cavity; Perform a feature point detection algorithm on the derivative matrix to identify inflection points, extreme points, and stable intervals to obtain a set of air pressure change feature points; Perform spectrum analysis on the force data in the dataset and extract the force amplitudes of the first and second pressing plates to obtain the force time series characteristics of the pressing plates; Calculate the force imbalance degree based on the force time series characteristics of the pressing plates and draw a curve of the force imbalance degree changing with time to obtain a force distribution characteristic map; Perform time synchronization superposition analysis on the set of air pressure change feature points and the force distribution characteristic map and comprehensively quantify the material distribution uniformity index through a weighted fusion algorithm to obtain the brick material distribution characteristics.

5. The method for monitoring the quality of bricks based on data analysis according to claim 1, characterized in that, Input the surface defect characteristic frequencies, contour feature vectors, and quality anomaly index of the bricks into a convolutional neural network for analysis to obtain the brick defect type determination result and defect severity score, including: Combine the surface defect characteristic frequencies, contour feature vectors, and quality anomaly index of the bricks into a unified feature matrix and perform normalization and standardization processing to obtain network input data; Input the network input data into the three-layer convolutional structure of the convolutional neural network for feature extraction to obtain deep mapping features; Input the deep mapping features into the two max-pooling layers of the convolutional neural network for dimensionality reduction to obtain a brick feature representation vector; Input the brick feature representation vector into the defect classifier of the convolutional neural network, corresponding to the predefined F types of brick defects, and use the softmax activation function and cross-entropy loss function to calculate the brick defect type determination result; Input the brick feature representation vector into the modified output layer of the convolutional neural network, and perform root mean square error loss calculation through the regression network with the sigmoid activation function in the modified output layer to obtain the defect severity score.

6. The method for monitoring the quality of bricks based on data analysis according to claim 1, characterized in that, Optimize the process parameters based on the defect type determination result and the defect severity score to obtain an equipment control strategy, including: Perform the correlation analysis of brick quality and process parameters based on the defect type determination result and the defect severity score, and obtain a process parameter vector including conveyor belt speed, initial position of the extrusion plate, spring pressure, baffle height, vibration bar height, and electronic stop valve switching time; Construct a process parameter optimization mathematical model based on the process parameter vector, and calculate the contribution degree of each process parameter to the quality target according to the process parameter optimization mathematical model to obtain a parameter contribution degree matrix; Divide the process parameters into a high contribution degree group and a low contribution degree group according to the parameter contribution degree matrix, and perform a differential evolution algorithm search in the subspace composed of high contribution degree parameters to obtain an optimal process parameter configuration; Based on the optimal process parameter configuration, design corresponding compensation strategies for different types of brick defects, adjust the conveyor belt speed and vibration bar height for polygon defects, adjust the extrusion plate position and partition position for contour errors, and adjust the electronic stop valve switching time for uneven material mixing problems to obtain a defect-specific compensation plan; Convert the defect-specific compensation plan into a device control strategy, and the device control strategy includes a speed control instruction for adjusting the motor output torque by a PID controller, a position control instruction for a servo motor control signal, a spring pressure control instruction for a pressure regulating mechanism, a baffle height control instruction for a pneumatic system, and a timing control instruction for an electronic stop valve.

7. The method for monitoring the quality of bricks based on data analysis according to claim 1, characterized in that, Transmit the device control strategy to each actuator to perform closed-loop feedback, and obtain a motor control signal, a servo positioning instruction, a contour adjustment instruction, and management layer optimization decision information, including: Decompose the device control strategy into device layer control, process layer control, and production management layer; Decompose the control strategy in the device layer control and transmit it to each actuator through a fieldbus to obtain a device basic control network; Based on the device basic control network, construct a three-phase current control model for motor control by applying a vector control algorithm, and implement servo positioning for the extrusion plate position control to obtain a motor control signal and a servo positioning instruction; Based on the iterative contour error compensation algorithm in the process layer control, calculate the motion synchronization points of the pusher component and the vibration component, and design a brick contour error compensation function to obtain a contour adjustment instruction; Based on the dynamic production plan adjustment strategy in the production management layer control, set the condition response mechanism for triggering the parameter self-optimization program when the same defect appears in continuously B bricks, triggering the emergency adjustment program for serious defects, and predicting the quality trend to obtain the management layer optimization decision information.

8. A brick quality monitoring system based on data analysis, characterized in that, For implementing the data analysis-based brick quality monitoring method according to any one of claims 1-7, the data analysis-based brick quality monitoring system includes: An acquisition module for acquiring vibration, pressure, displacement, image, and pneumatic data on a brick production line to obtain a data set; A detection module for performing domain synchronous averaging and edge detection on the data set to obtain the surface defect characteristic frequency of the brick, the brick contour characteristic vector, and the quality anomaly index; An analysis module, configured to input the surface defect feature frequency of the brick, the brick profile feature vector, and the quality anomaly index into a convolutional neural network for analysis, so as to obtain a determination result of the brick defect type and a defect severity score; A parameter optimization module, configured to perform process parameter optimization based on the defect type determination result and the defect severity score, so as to obtain a device control strategy; A transmission module, configured to transmit the device control strategy to each actuator to perform closed-loop feedback, so as to obtain a motor control signal, a servo positioning instruction, a profile adjustment instruction, and management layer optimization decision information.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by a processor, the processor is caused to execute the data analysis-based brick quality monitoring method according to any one of claims 1 to 7.

Images