Intelligent control system and method for stone decorative plate production

Abstract

The invention relates to an intelligent control system and method for stone decorative plate production, and belongs to the field of stone machining intelligent manufacturing. The system comprises a data scanning unit for generating a stone digital twin model through hyperspectral imaging, three-dimensional laser scanning and ultrasonic flaw detection multi-sensor fusion; the intelligent analysis and decision-making unit is used for extracting texture features through deep learning based on a model, calculating a crack risk index to perform defect grading, and optimizing a cutting scheme by taking maximization of an economic value as a target; the self-adaptive processing control unit calls a process parameter library according to stone characteristics, dynamically adjusts processing parameters based on real-time texture feedback, and cooperates with multiple devices to operate through an OPCUA protocol; the quality tracing and optimizing unit is used for endowing a unique two-dimensional code to the plate, associating whole-process data and iteratively optimizing process parameters by utilizing machine learning; the full-process closed-loop intelligent control of stone from sensing, planning, processing to optimization is realized, and the outturn percentage, the product quality and the production efficiency are further improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology for stone processing, and more specifically, to an intelligent control system and method for the production of stone decorative panels. Background Technology

[0002] The production of stone decorative panels is a complex process, encompassing multiple steps such as raw material evaluation, sawing, grinding, polishing, and cutting. Traditional production methods rely heavily on manual experience, which has many insurmountable drawbacks. During the production process, judging the grain direction, identifying defects, and grading are highly dependent on the experience of master craftsmen, resulting in large fluctuations in the quality of panels produced by different batches and different operators. The rate of superior products is generally low. At the same time, the material layout plan is mostly estimated by visual inspection, making it difficult to optimize globally and effectively avoid internal defects and follow the grain. This results in a comprehensive yield of only about 70% of high-quality raw materials, leading to significant waste. Furthermore, the parameters of processing equipment, such as sawing machines and polishing machines, are usually set to fixed values ​​or simple grading, which cannot be dynamically adjusted according to the real-time hardness, grain density, and other characteristics of the stone, resulting in high energy consumption, rapid tool wear, and low processing efficiency. While some existing technologies have attempted to introduce machine vision or automated control, such as the cutting method of a smart vision-based contour-following stone cutting robot (patent publication number CN111086113A), its functions mainly focus on the programmed start / stop and speed control of the equipment, without involving intelligent perception and decision-making regarding the complex characteristics of the stone itself, thus still having significant limitations. Therefore, based on the existing problems, this paper designs a system and method that can run through the entire production process of stone decorative panels and realize closed-loop intelligent control from perception to optimization, in order to systematically solve the above problems. Summary of the Invention

[0003] In view of the shortcomings of the existing technology, the purpose of this invention is to provide an intelligent control system and method for the production of stone decorative panels, so as to solve the problems existing in the background technology.

[0004] The above-mentioned technical objective of the present invention is achieved through the following technical solution: an intelligent control system for the production of stone decorative panels, comprising: The data scanning unit is used to acquire physical characteristic data of stone raw materials and generate a digital twin model of the stone. The intelligent analysis and decision-making unit is communicatively connected to the data scanning unit and is used for texture recognition, defect assessment, and production planning based on the digital twin model. An adaptive processing control unit, connected to the intelligent analysis and decision-making unit and the production line equipment, is used to execute and control the processing process in real time, including a process parameter library, a real-time parameter adjustment module and a multi-equipment collaborative controller; The quality traceability and optimization unit is used to assign a unique identifier to the produced board material, associate it with data from the entire process, and iteratively optimize the process parameter library based on historical production data.

[0005] Optionally, the data scanning unit includes: The hyperspectral imaging module, operating in the 380nm to 2500nm band, is used to collect spectral reflectance data of the stone surface to identify the distribution of mineral components. The 3D laser scanning module uses line laser scanning to acquire 3D point cloud data of the stone surface and then reconstruct the geometric shape of the stone. The ultrasonic flaw detection module uses an ultrasonic array sensor with a frequency of 5-10MHz to detect cracks and defects with a depth of ≥2mm inside the stone and record their three-dimensional spatial coordinates, length and direction. A multi-sensor fusion processor is used to perform spatiotemporal registration and feature layer fusion on data from the hyperspectral imaging module, the three-dimensional laser scanning module, and the ultrasonic flaw detection module, thereby generating the digital twin model of the stone.

[0006] Optionally, the intelligent analysis and decision-making unit includes: The texture feature extraction module has a built-in texture recognition model trained based on deep learning, which is used to segment the texture region from the digital twin model of the stone and quantify and output the main orientation angle, line density and contrast of the texture. The defect assessment and classification module is used to calculate the risk index R of internal cracks and classify defects accordingly. The risk index R is calculated as follows: Where d represents the normalized fracture depth, The normalized fracture length The angle between the fracture orientation and the nearest dominant texture orientation; The adaptive cutting planning module, with the objective function of maximizing economic value, employs an improved genetic algorithm for sorting optimization. The objective function is... ,in, Let i be the value coefficient of the i-th planned board. Let be the area of ​​the i-th planned board. This represents the estimated total value of the planned board material.

[0007] Optionally, the texture recognition model in the texture feature extraction module is a U-Net network structure with an attention mechanism, the output texture principal orientation angle accuracy is ±0.5°, the texture line density unit is lines / cm, and the texture contrast is a normalized value.

[0008] Optionally, the defect assessment and classification module classifies defects into three levels based on the calculated risk index R: Level I, represented by R<3, is a minor defect; Level II, represented by 3≤R<10, is a moderate defect; and Level III, represented by R≥10, is a severe defect. The optimization constraints of the adaptive cutting planning module include: for superior grade boards, the angle between the main grain direction and the preset cutting baseline must be ≤10°; the planned path must completely avoid all Class III defect areas and be at least 5mm away from their edges.

[0009] Optionally, in the adaptive machining control unit: The process parameter library stores the physical property parameters of different types of stone and the corresponding set of recommended processing parameters; The real-time parameter adjustment module is used to dynamically adjust the spindle speed during processing based on real-time feedback of texture density changes. The adjustment formula is as follows: ,in Indicated as the base speed, This is expressed as a sensitivity coefficient related to stone materials. This represents the real-time texture density at the current processing point. The average texture density of the processed stone variety; The multi-device collaborative controller is based on the OPC UA communication protocol and is used to synchronously control the operating cycle of the bridge saw, continuous grinding and polishing line, and CNC cutting center.

[0010] Optionally, the real-time parameter adjustment module is also connected to a vibration acceleration sensor and an acoustic emission sensor, and has a processing state classifier based on a support vector machine, which is used to determine the processing state based on the characteristics of the vibration and acoustic emission signals monitored in real time, and then automatically adjust the equipment parameters or trigger protection actions based on the judgment results.

[0011] Optionally, the quality traceability and optimization unit includes: The unique identifier assignment module is used to assign a unique QR code identifier to each cut piece of raw board. The end-to-end data association module is used to associate and store the unique QR code with data from the full-dimensional scanning unit, intelligent analysis and decision-making unit, adaptive processing control unit, and quality inspection data of the final product. The process parameter iterative optimization module uses a random forest regression model to periodically retrain the model with actual processing parameters and corresponding finished product quality data from historical production data, in order to update the process parameter library.

[0012] An intelligent control method for the production of stone decorative panels, based on the above-mentioned intelligent control system for stone decorative panel production, includes: Step S1, Full-Dimensional Data Acquisition and Modeling: The data scanning unit collects data from the stone and then generates a digital twin model of the stone; Step S2, Intelligent Analysis and Production Planning: Based on the generated digital twin model of the stone, the intelligent analysis and decision-making unit performs texture feature extraction, defect assessment and grading, and calculates and generates the optimal cutting scheme and CNC machining code; Step S3, Adaptive Collaborative Machining: The adaptive machining control unit executes the instructions of the intelligent analysis and decision-making unit, calls the process parameters, collaboratively controls the equipment to perform machining, and dynamically adjusts the machining parameters based on real-time feedback; Step S4, Data Traceability and Closed-Loop Optimization: The quality traceability and optimization unit binds and identifies the processed boards and associates the data, and then iterates and optimizes the process parameters based on the analysis of the associated data.

[0013] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the aforementioned intelligent control method for the production of stone decorative panels.

[0014] In summary, the present invention has the following beneficial effects: 1. By fusing hyperspectral, 3D laser, and ultrasonic flaw detection technologies in the data scanning unit, objective, accurate, and digital perception of the internal and external characteristics of the stone is achieved. Through texture feature extraction, defect assessment, and grading by the intelligent analysis and decision-making unit, and by employing quantitative models, such as the risk index R, and AI algorithms to replace subjective judgment, the texture direction and defect level are scientifically and consistently evaluated, reducing fluctuations caused by human factors and ensuring unified and reproducible quality judgment standards. During planning, the texture direction of superior-grade slabs is forcibly constrained, and serious defects are intelligently avoided, ensuring the inherent quality of the slabs from the source. This leads to a stable increase in the rate of superior-grade products, with minimal quality differences between batches, achieving large-scale and stable production of high-quality products.

[0015] 2. The adaptive cutting planning module in the intelligent analysis and decision-making unit aims to maximize economic value. It uses an improved genetic algorithm for global layout, considering not only area but also innovatively introducing a value coefficient Vi determined by both texture aesthetics and defect location. Differentiated treatment strategies, such as repair, guided utilization, and forced avoidance, are implemented for different defect levels, such as I, II, and III. This achieves a transformation from simple area maximization to comprehensive value maximization. The system can make the best use of resources, planning high-quality parts as high-value slabs, cleverly utilizing flawed but usable parts, and completely discarding only severely defective areas that are unusable, thereby improving the comprehensive utilization rate of stone raw materials.

[0016] 3. The adaptive machining control unit includes a real-time parameter adjustment module and connected vibration / acoustic emission sensors. Based on real-time texture features, such as density, obtained from a digital twin model, the system dynamically adjusts equipment parameters using preset adjustment models, such as spindle speed formulas. Simultaneously, a machining status classifier monitors equipment load in real time, enabling preventative adjustments or alarms. This achieves flexible machining tailored to individual material conditions, automatically enhancing cutting capabilities to ensure quality in areas with dense textures and automatically reducing energy consumption in areas with loose textures. Timely intervention before equipment malfunctions prevents destructive machining. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the system implementation process of the present invention; Figure 2 This is a schematic diagram of the method steps of the present invention. Detailed Implementation

[0018] To make the objectives, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Several embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein.

[0019] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances. The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature.

[0020] In this invention, unless otherwise expressly specified and limited, "above" or "below" a second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of a second feature includes the first feature being directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" of a second feature includes the first feature being directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature. The terms "vertical," "horizontal," "left," "right," "above," "below," and similar expressions are for illustrative purposes only and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed or operated in a specific orientation, and therefore should not be construed as limiting the invention.

[0021] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] This invention provides an intelligent control system and method for the production of stone decorative panels, such as... Figure 1 As shown, it includes: The data scanning unit is used to acquire physical characteristic data of stone raw materials and generate a digital twin model of the stone. The intelligent analysis and decision-making unit is communicatively connected to the data scanning unit and is used for texture recognition, defect assessment, and production planning based on the digital twin model. An adaptive processing control unit, connected to the intelligent analysis and decision-making unit and the production line equipment, is used to execute and control the processing process in real time, including a process parameter library, a real-time parameter adjustment module and a multi-equipment collaborative controller; The quality traceability and optimization unit is used to assign a unique identifier to the produced board material, associate it with data from the entire process, and iteratively optimize the process parameter library based on historical production data.

[0023] Furthermore, the data scanning unit includes: The hyperspectral imaging module, operating in the 380nm to 2500nm band, is used to collect spectral reflectance data of the stone surface to identify the distribution of mineral components. The 3D laser scanning module uses line laser scanning to acquire 3D point cloud data of the stone surface and then reconstruct the geometric shape of the stone. The ultrasonic flaw detection module uses an ultrasonic array sensor with a frequency of 5-10MHz to detect cracks and defects with a depth of ≥2mm inside the stone and record their three-dimensional spatial coordinates, length and direction. A multi-sensor fusion processor is used to perform spatiotemporal registration and feature layer fusion on data from the hyperspectral imaging module, the three-dimensional laser scanning module, and the ultrasonic flaw detection module, thereby generating the digital twin model of the stone.

[0024] Specifically, the data scanning unit is a physical detection system integrating multiple non-contact sensing technologies. Its core task is to collect comprehensive, multi-dimensional, and high-precision data from stone blocks. This unit consists of the following four hardware modules, which are integrated through a unified mechanical structure and control system: The hyperspectral imaging module mainly consists of a pushbroom hyperspectral imaging camera, a matching linear guide rail, a uniform light source system, and a control box. The camera's operating spectral range covers 380nm to 2500nm, including the visible light (VIS) and short-wave infrared (SWIR) bands. Its spatial resolution is designed to be ≤0.1mm, ensuring the capture of subtle differences in mineral distribution for acquiring spectral reflectance data of stone surfaces. Each pixel contains a continuous spectral curve, which serves as a fingerprint for identifying surface mineral composition, such as calcite, quartz, and iron oxides. The 3D laser scanning module employs the principle of line laser triangulation and includes a high-precision line laser generator, a high-speed CMOS camera, and a control unit. The laser line forms a deformation light stripe on the surface of the stone being measured. The camera captures this light stripe and generates 3D points through triangulation calculations. The system is designed with a measurement accuracy of 0.01mm to meet the needs of fine reconstruction of the geometric features of the stone surface, such as unevenness and warping. It is used to quickly acquire complete and dense 3D point cloud data of the stone surface, establish an accurate geometric model, and provide a spatial reference framework for all other data. The ultrasonic flaw detection module's core consists of a multi-channel, 5-10MHz ultrasonic phased array probe array, a pulse transmitter / receiver, and a data acquisition card. The probe array is arranged in a specific pattern, enabling electronic scanning and focusing of the sound beam to detect defects of varying depths and orientations. It is designed to penetrate stone and detect internal cracks, pores, and other defects. Its design sensitivity is for defects with a depth ≥2mm, and it can calculate the three-dimensional spatial coordinates, length, and orientation of defects using Time-of-Flight Diffraction (TOFD) or Synthetic Aperture Focusing Technique (SAFT). Multi-sensor fusion processor, a high-performance industrial computer or embedded processing platform, equipped with multi-channel data acquisition cards and powerful GPUs, is used to process data from various sources in real time. It is responsible for receiving, synchronizing, registering and fusing heterogeneous data streams from the three independent modules mentioned above, and integrating spectral, geometric and internal structure information in a unified spatiotemporal coordinate system through algorithms. In the specific implementation process, the stone raw materials are positioned and fixed at the scanning station by the conveying system on the production line. The system performs a power-on self-test and sensor calibration. The hyperspectral module performs whiteboard calibration to correct the illumination; the three-dimensional laser module performs planar target calibration to determine internal and external parameters; and the ultrasonic module performs sound velocity calibration on a standard test block. Multiple sensors simultaneously scan and collect data. Three modules begin scanning the raw stone along a preset path. The hyperspectral imaging module moves along the guide rail, collecting data line by line using a push-scan method, generating a three-dimensional data cube H(x,y,λ), where (x,y) is the spatial pixel position and λ is the wavelength. Simultaneously, the three-dimensional laser scanning module moves along another dimension, scanning the stone surface with a laser line and generating a point cloud set in real time, represented as: The ultrasonic flaw detection module's probe array, based on a preset scanning grid, sequentially excites and receives ultrasonic signals at coupling points on the stone surface, forming a raw radio frequency waveform data set U(t,pos), where t is the echo time and pos is the probe position coordinates. The collected raw data is transmitted in real time to the multi-sensor fusion processor, which performs data preprocessing: radiometric calibration and denoising of the hyperspectral data; filtering and simplification of the point cloud data; gain compensation and filtering of the ultrasonic data; and spatiotemporal registration using the preprocessed data as the basis. Spatial reference establishment: A high-precision 3D laser point cloud P is used as the spatial geometric reference for the entire system; Hyperspectral data mapping: Each pixel is mapped using pre-calibrated intrinsic and extrinsic parameters of the hyperspectral camera. The mapping relationship between the corresponding surface points in the 3D point cloud P and the point cloud P is expressed as follows: Then, the two-dimensional spectral information H is attached to the surface of the three-dimensional model; Ultrasonic data mapping: Similarly, by calibrating the parameters through the installation position of the ultrasonic probe, each detection position pos and the detected defect depth d are represented by coordinate transformation as follows. Transform the endpoint coordinates of the internal defects to the same world coordinate system as the point cloud P; After spatiotemporal registration is completed, the processor executes a fusion algorithm to construct a triangular mesh surface from the 3D point cloud P, serving as the geometric skeleton of the model. The registered hyperspectral data is then used as texture maps or higher-level material properties, associated with each vertex or facet of the mesh, recording its spectral characteristics. The registered ultrasonic defect data is used to generate independent 3D objects representing the defect geometry within the 3D mesh model, such as slender ellipsoids or planes, and these objects are assigned defect attributes, recording the calculated length l and orientation θ. Finally, a structured data file, namely the stone digital twin model, is output. The data includes geometric attributes: precise three-dimensional surface shape; spectral / material attributes: mineral composition information at every point on the surface (derived from hyperspectral data); structural attributes: three-dimensional spatial distribution, size, and orientation of internal defects; and the generated digital twin model of the stone. Through the internal network, the data is fully transmitted to the intelligent analysis and decision-making unit, serving as the sole data basis for all its analysis, planning, and decision-making.

[0025] Furthermore, the intelligent analysis and decision-making unit includes: The texture feature extraction module has a built-in texture recognition model trained based on deep learning, which is used to segment the texture region from the digital twin model of the stone and quantify and output the main orientation angle, line density and contrast of the texture. The defect assessment and classification module is used to calculate the risk index R of internal cracks and classify defects accordingly. The risk index R is calculated as follows: Where d represents the normalized fracture depth, The normalized fracture length The angle between the fracture orientation and the nearest dominant texture orientation; The adaptive cutting planning module, with the objective function of maximizing economic value, employs an improved genetic algorithm for sorting optimization. The objective function is... ,in, Let i be the value coefficient of the i-th planned board. Let be the area of ​​the i-th planned board. This represents the estimated total value of the planned board material.

[0026] Furthermore, the texture recognition model in the texture feature extraction module is a U-Net network structure with an attention mechanism, the output texture principal orientation angle accuracy is ±0.5°, the texture line density unit is lines / cm, and the texture contrast is a normalized value.

[0027] Furthermore, the defect assessment and grading module classifies defects into three levels based on the calculated risk index R: Level I, represented by R<3, is a minor defect; Level II, represented by 3≤R<10, is a moderate defect; and Level III, represented by R≥10, is a severe defect. The optimization constraints of the adaptive cutting planning module include: for superior grade boards, the angle between the main grain direction and the preset cutting baseline must be ≤10°; the planned path must completely avoid all Class III defect areas and be at least 5mm away from their edges.

[0028] Specifically, the intelligent analysis and decision-making unit is a data processing and analysis center based on software algorithms and computational models. Its physical carrier is a high-performance server or industrial computing platform, and it does not contain new sensing hardware. Its functions are entirely implemented by executing pre-installed specialized software modules. This unit consists of three core algorithm modules that work in sequence to transform the input digital twin model into executable production planning instructions, including: The texture feature extraction module is based on a pre-trained deep learning model. The model uses a U-Net network with an attention mechanism as its basic architecture. The training data consists of a large number of labeled stone surface images and their corresponding texture directions and region ground truth values. After the module loads this model, it runs as an image processing function. The input is an image, and the output is structured data. It is used to automatically identify, segment, and quantify the texture visual features of stone from two-dimensional projection images rendered by digital twin models or directly processed three-dimensional texture data, providing key parameters for subsequent planning and processing. The defect assessment and grading module is an expert system based on deterministic mathematical formulas and logical rules. Its core is a calculation model for the defined crack risk index R, expressed as: And a fixed grading threshold based on the R value: Level I: R<3; Level II: 3≤R<10; Level III: R≥10. The module does not contain a machine learning training process, but only performs calculations and logical judgments based on the input parameters. It is used to perform quantitative risk assessment and severity grading of each internal defect found by the upstream scanning unit, transforming the defect from a simple geometric description into a grade label with production guidance significance. The adaptive cutting planning module is an optimization solver that encapsulates an improved genetic algorithm. The algorithm uses a hybrid encoding: real numbers encode the cutting line position, and integers encode the board material grade to represent the solution. Its core is the defined objective function, expressed as: In addition to hard constraints, such as texture angle ≤10° and avoidance of Class III defects, the algorithm searches for the optimal solution through selection, crossover, and mutation iteration. The module integrates texture features and defect distribution, and under multiple constraints, automatically calculates the optimal cutting and layout scheme that maximizes the economic value of the raw blocks, and generates the CNC code for driving the equipment. In the specific work process, the intelligent analysis and decision-making unit receives the complete digital twin model of the stone. The data format is parsed, and the surface geometry, texture images and internal defect list in the model are loaded into memory to prepare standardized input interfaces for each module. Texture feature extraction, with data input from a digital twin model of the stone. Extract or render in real time to generate a high-resolution two-dimensional orthographic image of the stone surface. The texture feature extraction module loads its internal deep learning model to extract the image. Input the model for forward inference; the model outputs the following quantization results, covering the entire stone surface: texture dominance angle (φ): an angle value representing the dominant direction of the texture within a specific area; the system ensures its output accuracy is ±0.5°; texture line density (ρ): a scalar value representing the number of texture lines per unit length (cm), in lines / cm; texture contrast (C): a value normalized to the [0,1] interval, representing the degree of distinction between the texture and the background; the features are used to generate a feature parameter mapping map corresponding to the stone surface; Defect assessment and grading, with data input from the stone digital twin model. Read the original list of all internal defects, each defect containing its three-dimensional spatial coordinates, measured original depth, and original length. and spatial orientation vector; The defect assessment and grading module performs the following calculations for each defect in the list: parameter normalization using system-preset baseline values. and The calculation is performed and expressed as follows: ; Angle calculation: Calculate the absolute value of the angle |θ| (in radians) between the defect's direction vector and the nearest main texture direction φ provided by the texture feature extraction module; Risk index calculation: Calculated as follows: The risk index R of the defect is then obtained. Logical classification is performed to classify the risk index R of each defect. If R < 3, it is marked as Level I (minor). If 3 ≤ R < 10, it is marked as Level II (moderate). If R ≥ 10, it is marked as Level III (severe). An updated defect list is then output, in which each defect has its risk index R value and classification label added to its original geometric information. The adaptive cutting planning module gathers all the outputs from the first two steps—a texture feature parameter mapping map and a list of defects with hierarchical labels—while also receiving production order requirements, such as the target board size and the requirements for each grade of board. The adaptive cutting planning module initiates its improved genetic algorithm solver to perform the following optimizations: Within the three-dimensional space of the stone, it randomly generates a set (a population) of initial cutting schemes (chromosomes); for each scheme, it calculates its economic value f(Plan) based on the objective function; during calculation, the value coefficient of each slab in the scheme is used. The grade of the material is determined by its size, the texture level of its location (queried from the feature map), and whether it contains or is adjacent to a Class II defect (queried from the defect list). Simultaneously, the plan is checked to ensure it does not violate the following hard constraints: for boards planned as premium grade, the average main grain direction within its area... With respect to the cutting direction baseline set by the system The included angle must satisfy No cutting path may cross any defect area marked as Level III, and must be at a preset distance from its edge. The sheet material dimensions must meet the standard. Solutions that violate these constraints will be subject to a severe penalty, significantly reducing their fitness. The algorithm generates a new generation of population through selection (retaining high-fitness solutions), crossover (mixing features of high-quality solutions, using constraint-preserving arithmetic crossover), and mutation (fine-tuning some solutions, introducing local search), and then repeatedly evaluates them. This process iterates until convergence conditions are met (such as reaching the maximum number of iterations or fitness stabilization). Finally, the optimal cutting solution is output, including a detailed data file describing the location and sequence of all cutting paths, as well as the size, grade, and texture characteristics of each produced sheet material; and CNC machining code, which is automatically generated according to the plan and can be directly recognized and executed by downstream sawing equipment using G-code or a dedicated instruction set.

[0029] Furthermore, in the adaptive machining control unit: The process parameter library stores the physical property parameters of different types of stone and the corresponding set of recommended processing parameters; The real-time parameter adjustment module is used to dynamically adjust the spindle speed during processing based on real-time feedback of texture density changes. The adjustment formula is as follows: ,in Indicated as the base speed, This is expressed as a sensitivity coefficient related to stone materials. This represents the real-time texture density at the current processing point. The average texture density of the processed stone variety; The multi-device collaborative controller is based on the OPC UA communication protocol and is used to synchronously control the operating cycle of the bridge saw, continuous grinding and polishing line, and CNC cutting center.

[0030] Furthermore, the real-time parameter adjustment module is also connected to a vibration acceleration sensor and an acoustic emission sensor, and has a built-in processing state classifier based on a support vector machine, which is used to determine the processing state based on the characteristics of the vibration and acoustic emission signals monitored in real time, and then automatically adjust the equipment parameters or trigger protection actions based on the judgment results.

[0031] Specifically, the adaptive machining control unit is the execution and control center connecting the decision-making system and the physical production line. It consists of software control modules, a database, and necessary industrial sensor interfaces. Its core function is to execute machining instructions accurately, flexibly, and collaboratively, including: The process parameter library is a structured relational database or knowledge base. Its core data tables store the mapping relationship between different stone categories and their recommended processing parameter sets. Each category is associated with a set of parameters, including but not limited to: sawing line speed, feed speed, grinding head pressure, polishing speed, and coolant flow rate. These parameters serve as the basis for subsequent adaptive adjustments. The real-time parameter adjustment module is a real-time control software module that embeds a dynamic spindle speed adjustment model, represented as follows: ,in, This is represented by the calculated adjusted spindle speed command value. This represents the base rotational speed value obtained from the process parameter database. This is expressed as a material sensitivity coefficient, an empirical constant preset based on the type of stone and processing stage, with a value ranging from -0.15 to 0.25. This is represented as the real-time texture line density feedback value at the current processing point. This is expressed as the average linear density value of the texture of this stone variety; the module connects to physical sensors from the equipment layer, including but not limited to: vibration acceleration sensors: mounted on the machining spindle or cutter head, used to collect vibration signals. Acoustic emission sensor: Installed on the equipment structure, used to collect high-frequency stress wave signals generated by material deformation and fracture. Simultaneously, the module also runs a pre-trained classification model based on Support Vector Machine (SVM), whose input is... and The time-domain and frequency-domain features (such as root mean square, peak value, and spectral centroid) are extracted in real time, and the output is a predefined machining state category: {smooth cutting, slight vibration, overload risk}. The multi-device collaborative controller is a software controller developed based on the industry standard OPC UA (Unified Architecture) communication protocol. It is used to build a server containing equipment information models such as bridge saws, continuous grinding and polishing lines, and CNC cutting centers, and can conduct bidirectional and secure data exchange and command issuance with the clients of these devices. In the specific implementation process, the unit receives the CNC machining code and related stone category identifier from the adaptive machining control unit. Based on the stone category identifier, the process parameter library is queried to obtain the corresponding recommended processing parameter set, and the values ​​in it are set as the initial set value or basic value of each equipment control loop. The real-time parameter adjustment module obtains the average texture line density of the stone from upstream and loads the preset material sensitivity coefficient. The multi-device collaborative controller parses the process logic in the CNC machining code and sends equipment enable, start, and initial parameter setting commands to the bridge sawing machine, continuous grinding and polishing line, CNC cutting center, etc., according to the preset cycle time through the OPC UA interface. The controller monitors the readiness and running status of each device and ensures that the previous process is completed and the result is normal before triggering the start of the next process, thereby achieving synchronization of the running cycle time. Online monitoring and adaptive parameter adjustment loop, which runs continuously during the processing, taking the sawing process as an example: Texture Feedback and Spindle Speed ​​Adjustment: The module maps the extracted texture feature data to obtain the real-time texture line density at the current saw head position. The real-time parameter adjustment module substitutes the corresponding data and the data retrieved from the process parameter library into the adjustment formula to calculate the speed. As a new speed command, it is sent to the sawing machine spindle drive in real time through the controller to complete a dynamic adjustment based on material properties; Condition monitoring and protective intervention: Vibration acceleration sensors and acoustic emission sensors continuously collect signals, which are sent to the SVM machining state classifier in the real-time parameter adjustment module for real-time analysis. Based on the learned feature patterns, the classifier outputs the current state: if the state is "smooth cutting", the system maintains the current parameters and continues machining; if the state is "slight vibration", the module automatically generates control commands to reduce the feed rate f of the equipment by 10%-20% to stabilize the machining process; if the state is "overload risk", the module immediately generates the highest priority interrupt command, triggering the protective suspension of the equipment (such as spindle stop, feed stop), and at the same time sends an audible and visual alarm signal to the operation interface to prompt manual intervention. It also features dynamic cycle time adjustment. When a process takes longer than the original cycle time due to parameter adjustments, such as reducing the feed speed to handle "slight vibrations," the multi-device collaborative controller will detect this deviation. The controller dynamically calculates and adjusts the rhythm of the preceding and following processes based on the equipment status and queue situation in the OPC UA information model. For example, it can automatically extend the buffer time between the previous sawing process and the current grinding and polishing process, or fine-tune the start-up waiting logic of the subsequent cutting process to minimize the overall production line waiting time and maintain smooth operation. Throughout the entire processing, all actual processing parameters, equipment status, and any triggered adjustments or alarm events are recorded in real time; the actual operating data is synchronously sent to the quality traceability and optimization unit via the internal network for data correlation and subsequent analysis.

[0032] Furthermore, the quality traceability and optimization unit includes: The unique identifier assignment module is used to assign a unique QR code identifier to each cut piece of raw board. The end-to-end data association module is used to associate and store the unique QR code with data from the full-dimensional scanning unit, intelligent analysis and decision-making unit, adaptive processing control unit, and quality inspection data of the final product. The process parameter iterative optimization module uses a random forest regression model to periodically retrain the model with actual processing parameters and corresponding finished product quality data from historical production data, in order to update the process parameter library.

[0033] Specifically, the quality traceability and optimization unit is the system's data brain and learning center. Its core function is to realize the full lifecycle management of production data and drive the continuous updating of process knowledge. This unit consists entirely of software modules and a database, and includes the following three core parts: The unique identifier assignment module is a software service bound to the production line control logic. Its core is an industry-standard QR code generation algorithm and a communication interface with marking equipment (such as a laser marking machine or inkjet printer). This module does not store business data; it is only responsible for generating and issuing identification instructions at specific points in time. It establishes a unique and non-repeatable association between the physical material and the data records in the digital world, ensuring that each produced material can be uniquely identified throughout its entire lifecycle. The end-to-end data association module is a relational database or data lake based on time series and event-driven principles. Its core design revolves around the QR code identifier as the primary key, building multiple data tables to store and index structured and unstructured data from upstream units. It does not generate new raw data, but only performs data ingestion, association, and persistent storage. It is used to achieve automatic aggregation and precise association of data across the entire production chain, establishing a complete data identifier for each piece of board. The process parameter iterative optimization module provides offline machine learning training and optimization services. Its core encapsulates the implementation of the random forest regression algorithm. This module periodically extracts historical data from the full-process data association module as a training set, performs model training, and outputs the training results, namely the optimized parameter mapping relationship. It is used to discover potential optimization relationships between process parameters and product quality by analyzing historical production big data and automatically update the process parameter library. In the specific implementation process, the event triggering and unique identifier assignment are as follows: When the adaptive processing control unit controls the bridge saw to complete the cutting and separation of a rough board, and the rough board is moved to the marking station, the position sensor will trigger a board generation event; the unique identifier assignment module is awakened by this event and performs the following operations: calling the QR code generation algorithm to generate a unique string code, and generating QR code graphic data accordingly; controlling the laser marking machine to engrave this QR code on the non-decorative side (usually the back) of the rough board through the communication interface; using this QR code string as the primary key of the board, publishing it within the system, and notifying the full-process data association module to prepare to receive all data related to this QR code string; The end-to-end data association module continuously monitors the data bus of each unit in the system; once it detects a data packet related to the stone raw material being processed or the newly identifiable slab, it captures it; the module collects and associates the captured data with ID_j as the primary key and stores it in the database. The data record established for each slab ID_j contains at least the following chains: Source data: A subset from the digital twin model of the data scanning unit, i.e., the data of the stone raw material that generated the slab; Planning data: Detailed planning information about the slab from the cutting plan of the intelligent analysis and decision-making unit, such as size, grade, planned texture direction, etc.; Execution data: Actual processing parameter records from the adaptive processing control unit, such as actual spindle rotation, actual feed speed, vibration status log, any alarms or adjustment events, etc.; Result data: Finished product inspection report from the quality inspection equipment at the end of the production line, including gloss (unit: GU), flatness (unit: mm), dimensional deviation ΔD, and the final grade (superior, first-class, etc.) determined manually or visually; When the slab completes all processes and is inspected and put into storage, its associated data chain is marked as complete and sealed, forming an unalterable production file; The process parameter iterative optimization module starts automatically at a preset cycle, such as every N cubic meters of raw material produced or every half month, or is manually triggered by the administrator. The module extracts complete production records from the database of the full-process data association module, covering all data chains since the previous optimization cycle. For each record, it extracts: Input features Stone characteristics (type, average texture density, mean defect index) and a set of key processing parameters for actual use. (e.g., actual average sawing speed, grinding head pressure); Output target The overall quality score of the board is calculated by weighting multiple indicators (such as gloss and flatness) in the finished product inspection report, or by directly using its commercial grade (e.g., superior grade = 1.0, first grade = 0.8) as the quantitative value; The random forest regression model is trained using this as the training set. The learning objective of this model is to find a function. This allows us to predict, for a given X, the potential for higher [achievements]. Recommended values ​​for optimization parameters ; After training, the module evaluates the learning results (e.g., cross-validation). If the new model's predictive performance on the validation set is significantly better than the historical performance of the old model, the system will automatically generate a suggestion to update the process parameter library. After this suggestion is reviewed and confirmed by engineers or managers, the recommended processing parameter set for the corresponding stone category in the process parameter library will be updated or new parameter combination options will be added. Thus, a data-driven knowledge iteration is completed. The workflow of the quality traceability and optimization unit begins with the identification of the slab at birth, builds a transparent and traceable data chain through full-process data association, and finally uses machine learning models to mine massive amounts of historical data to achieve closed-loop iterative optimization of process knowledge.

[0034] An intelligent control method for the production of stone decorative panels, based on the above-mentioned intelligent control system for stone decorative panel production, includes: Step S1, Full-Dimensional Data Acquisition and Modeling: The data scanning unit collects data from the stone and then generates a digital twin model of the stone; Step S2, Intelligent Analysis and Production Planning: Based on the generated digital twin model of the stone, the intelligent analysis and decision-making unit performs texture feature extraction, defect assessment and grading, and calculates and generates the optimal cutting scheme and CNC machining code; Step S3, Adaptive Collaborative Machining: The adaptive machining control unit executes the instructions of the intelligent analysis and decision-making unit, calls the process parameters, collaboratively controls the equipment to perform machining, and dynamically adjusts the machining parameters based on real-time feedback; Step S4, Data Traceability and Closed-Loop Optimization: The quality traceability and optimization unit binds and identifies the processed boards and associates the data, and then iterates and optimizes the process parameters based on the analysis of the associated data.

[0035] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the aforementioned intelligent control method for the production of stone decorative panels.

[0036] This invention discloses an intelligent control system and method for the production of stone decorative panels. Through the fusion of hyperspectral, 3D laser, and ultrasonic flaw detection by a data scanning unit, it achieves objective, accurate, and digital perception of the internal and external characteristics of the stone. The intelligent analysis and decision-making unit extracts texture features, assesses and grades defects, and employs quantitative models, such as the risk index R, along with AI algorithms to replace subjective judgment. This scientifically and consistently evaluates texture direction and defect levels, reducing fluctuations caused by human factors and ensuring unified and reproducible quality judgment standards. During planning, the system forcibly constrains the texture direction of superior-grade panels and intelligently avoids serious defects, ensuring the inherent quality of the panels from the source. This leads to a stable increase in the rate of superior-grade products, with minimal batch-to-batch quality differences, achieving large-scale and stable production of high-quality products. The adaptive cutting planning module in the intelligent analysis and decision-making unit, with the optimization goal of maximizing economic value, uses an improved genetic algorithm for global layout, considering not only area but also innovatively introducing factors such as texture aesthetics and... The value coefficient Vi is determined by the location of defects, and differentiated treatment strategies such as repair, guided utilization, and forced avoidance are implemented for defects of different levels, such as I, II, and III. This achieves a transformation from simple area maximization to comprehensive value maximization. The system can make the most of resources, planning high-quality parts as high-value slabs, cleverly utilizing flawed but usable parts, and completely discarding only severely defective areas that are unusable, thereby improving the comprehensive utilization rate of stone raw materials. The adaptive processing control unit has a real-time parameter adjustment module and connected vibration / acoustic emission sensors. The system dynamically adjusts equipment parameters based on real-time texture features obtained from the digital twin model, such as density, through preset adjustment models, such as spindle speed formulas. At the same time, the processing status classifier monitors the equipment load in real time, making preventive adjustments or alarms. It realizes flexible processing tailored to the material, automatically enhancing cutting ability to ensure quality in dense textures and automatically reducing energy consumption in loose textures. It intervenes in time before the equipment status becomes abnormal to avoid destructive processing.

[0037] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. An intelligent control system for the production of stone decorative panels, characterized in that, include: The data scanning unit is used to acquire physical characteristic data of stone raw materials and generate a digital twin model of the stone. The intelligent analysis and decision-making unit is communicatively connected to the data scanning unit and is used for texture recognition, defect assessment, and production planning based on the digital twin model. An adaptive processing control unit, connected to the intelligent analysis and decision-making unit and the production line equipment, is used to execute and control the processing process in real time, including a process parameter library, a real-time parameter adjustment module and a multi-equipment collaborative controller; The quality traceability and optimization unit is used to assign a unique identifier to the produced board material, associate it with data from the entire process, and iteratively optimize the process parameter library based on historical production data.

2. The intelligent control system for the production of stone decorative panels according to claim 1, characterized in that, The data scanning unit includes: The hyperspectral imaging module, operating in the 380nm to 2500nm band, is used to collect spectral reflectance data of the stone surface to identify the distribution of mineral components. The 3D laser scanning module uses line laser scanning to acquire 3D point cloud data of the stone surface and then reconstruct the geometric shape of the stone. The ultrasonic flaw detection module uses an ultrasonic array sensor with a frequency of 5-10MHz to detect cracks and defects with a depth of ≥2mm inside the stone and record their three-dimensional spatial coordinates, length and direction. The multi-sensor fusion processor is used to perform spatiotemporal registration and feature layer fusion on data from the hyperspectral imaging module, the three-dimensional laser scanning module and the ultrasonic flaw detection module, thereby generating the digital twin model of the stone.

3. The intelligent control system for the production of stone decorative panels according to claim 1, characterized in that, The intelligent analysis and decision-making unit includes: The texture feature extraction module has a built-in texture recognition model trained based on deep learning, which is used to segment the texture region from the digital twin model of the stone and quantify and output the main orientation angle, line density and contrast of the texture. The defect assessment and classification module is used to calculate the risk index R of internal cracks and classify defects accordingly. The risk index R is calculated as follows: Where d represents the normalized fracture depth, The normalized fracture length The angle between the fracture orientation and the nearest dominant texture orientation; The adaptive cutting planning module, with the objective function of maximizing economic value, employs an improved genetic algorithm for sorting optimization. The objective function is... ,in, Let i be the value coefficient of the i-th planned board. Let be the area of ​​the i-th planned board. This represents the estimated total value of the planned board material.

4. The intelligent control system for the production of stone decorative panels according to claim 3, characterized in that, The texture recognition model in the texture feature extraction module is a U-Net network structure with an attention mechanism. The output texture principal orientation angle accuracy is ±0.5°, the texture line density is in lines / cm, and the texture contrast is a normalized value.

5. The intelligent control system for the production of stone decorative panels according to claim 3, characterized in that, The defect assessment and classification module classifies defects into three levels based on the calculated risk index R: Level I, represented by R<3, is a minor defect; Level II, represented by 3≤R<10, is a moderate defect; and Level III, represented by R≥10, is a severe defect. The optimization constraints of the adaptive cutting planning module include: for superior grade boards, the angle between the main grain direction and the preset cutting baseline must be ≤10°; the planned path must completely avoid all Class III defect areas and be at least 5mm away from their edges.

6. The intelligent control system for the production of stone decorative panels according to claim 1, characterized in that, In the adaptive machining control unit: The process parameter library stores the physical property parameters of different types of stone and the corresponding set of recommended processing parameters; The real-time parameter adjustment module is used to dynamically adjust the spindle speed during processing based on real-time feedback of texture density changes. The adjustment formula is as follows: ,in Indicated as the base speed, This is expressed as a sensitivity coefficient related to stone materials. This represents the real-time texture density at the current processing point. The average texture density of the processed stone variety; The multi-device collaborative controller is based on the OPC UA communication protocol and is used to synchronously control the operating cycle of the bridge saw, continuous grinding and polishing line, and CNC cutting center.

7. The intelligent control system for the production of stone decorative panels according to claim 6, characterized in that, The real-time parameter adjustment module is also connected to a vibration acceleration sensor and an acoustic emission sensor, and has a processing state classifier based on a support vector machine. This classifier is used to determine the processing state based on the characteristics of the vibration and acoustic emission signals monitored in real time, and then automatically adjust the equipment parameters or trigger protection actions based on the determination results.

8. The intelligent control system for the production of stone decorative panels according to claim 1, characterized in that, The quality traceability and optimization unit includes: The unique identifier assignment module is used to assign a unique QR code identifier to each cut piece of raw board. The end-to-end data association module is used to associate and store the unique QR code with data from the full-dimensional scanning unit, intelligent analysis and decision-making unit, adaptive processing control unit, and quality inspection data of the final product. The process parameter iterative optimization module uses a random forest regression model to periodically retrain the model with actual processing parameters and corresponding finished product quality data from historical production data, in order to update the process parameter library.

9. A method for intelligent control of stone decorative panel production based on an intelligent control system for stone decorative panel production according to any one of claims 1-8, characterized in that, include: Step S1, Full-Dimensional Data Acquisition and Modeling: The data scanning unit collects data from the stone and then generates a digital twin model of the stone; Step S2, Intelligent Analysis and Production Planning: Based on the generated digital twin model of the stone, the intelligent analysis and decision-making unit performs texture feature extraction, defect assessment and grading, and calculates and generates the optimal cutting scheme and CNC machining code; Step S3, Adaptive Collaborative Machining: The adaptive machining control unit executes the instructions of the intelligent analysis and decision-making unit, calls up process parameters, collaboratively controls the equipment to perform machining, and dynamically adjusts the machining parameters based on real-time feedback; Step S4, Data Traceability and Closed-Loop Optimization: The quality traceability and optimization unit binds and identifies the processed boards and associates the data, and then iterates and optimizes the process parameters based on the analysis of the associated data.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements each step of the intelligent control method for the production of stone decorative panels as described in claim 9.

Images