A flat grinder grinding process intelligent control system and method

Abstract

The present application belongs to the field of intelligent control of machine tool grinding processing, and relates to a kind of surface grinder grinding process intelligent control system and method.The system includes surface grinder and intelligent grinding process module, and the module adopts double-ring processing architecture: slow ring decision and self-learning layer, with single workpiece processing cycle as processing cycle, for process parameter offline derivation and model iteration update;Fast ring real-time monitoring and intervention layer, with 10ms~100ms as processing cycle, for real-time suppression of load surge to prevent burn.In the system, quality closed-loop feedback is realized through cross-ring data interaction mechanism, and successful samples meeting the conditions are written back to the process database to update the parameter mapping model after processing, supplemented by human-computer interaction and traceability function.The present application realizes adaptive optimization and real-time closed-loop control of grinding process, effectively improves the consistency of processing quality and the level of system intelligence under complex working conditions.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control of machine tool grinding, and relates to an intelligent control system and method for surface grinding process. Background Technology

[0002] Surface grinders are widely used in aerospace, energy, and rail transportation industries, and are used for machining high-hardness materials such as forgings and alloy steels. These materials are highly heat-sensitive, and if local overheating occurs during grinding, defects such as grinding burns and surface microcracks can easily occur, affecting the quality of the parts.

[0003] In traditional surface grinding processes, grinding parameters (such as feed rate, feed depth, and wheel speed) are mainly determined based on the operator's experience. When encountering fluctuations in operating conditions such as uneven material allowance, wheel wear, and changes in machine tool condition, operators find it difficult to adjust the parameters in a timely manner, affecting processing efficiency and product quality stability.

[0004] In existing technologies that introduce online spindle load monitoring to monitor the grinding process, once the process parameters are set, they remain fixed during the machining process. The parameter prediction model and the real-time monitoring system operate independently, without establishing an effective linkage intervention mechanism. Furthermore, the effective monitoring data and intervention records generated during the machining process are difficult to feed back to the process parameter decision system and cannot be used for parameter optimization in subsequent machining.

[0005] Furthermore, existing technologies mostly exist as independent monitoring modules or post-processing alarm systems, and have not yet formed a process software package that is deeply integrated with CNC systems and can be directly deployed. This makes it difficult to systematically accumulate and reuse machining experience. Therefore, current grinding control technologies are still unable to achieve automatic updating of process parameters and closed-loop control of the machining process under complex working conditions. Summary of the Invention

[0006] The purpose of this invention is to address the aforementioned problems in the prior art by providing an intelligent control system and method for surface grinding processes, thereby solving these problems.

[0007] The objective of this invention can be achieved through the following technical solution: an intelligent control system for grinding process of a surface grinder, characterized in that it includes a surface grinder and an intelligent grinding process module, wherein the surface grinder has a linkage mechanism for executing X, Y, and Z linear axes, and also includes a grinding spindle and a grinding wheel; The intelligent grinding process module adopts a dual-loop processing architecture, including a slow-loop decision-making and self-learning layer and a fast-loop real-time monitoring and intervention layer. The slow-loop decision-making and self-learning layer, with a processing cycle on the order of a single workpiece machining cycle, is used to perform offline derivation of process parameters and iterative model updates, and includes: The process input and workpiece modeling unit is used to receive workpiece material and processing condition information and generate corresponding workpiece processing condition vectors. The intelligent process decision unit is communicatively connected to the process input and workpiece modeling unit. Based on the process database, it uses a parameter mapping model to derive and generate a set of grinding parameters and dynamic thresholds corresponding to each stage of rough grinding, semi-fine grinding, fine grinding, and sparkless grinding. An automatic cyclic grinding and wheel dressing unit is used to process the grinding parameter set to generate a machining program and dressing subroutine that can be executed by the machine tool, and then download it to the CNC system. The fast-loop real-time monitoring and intervention layer, with a processing cycle of 10ms to 100ms, is used to suppress sudden load increases during processing to prevent burns. It includes: The real-time signal acquisition and feature extraction unit is used to acquire the machine tool spindle power signal and extract real-time load features during the machining process; The process identification and dynamic threshold matching unit is used to identify the current processing stage and call the corresponding dynamic threshold. The dynamic burn prevention adaptive control unit is used to output graded intervention commands to the feed rate adjustment interface, cooling control interface and program jump interface of the CNC system based on the comparison results of real-time load characteristics and dynamic thresholds. The intelligent grinding process module also includes a cross-ring data interaction mechanism: The quality closed-loop and process database reverse writing unit is set at the end of the slow loop decision and self-learning layer. It is used to perform quality inspection after processing and to reverse write successful samples to the process database to update the parameter mapping model when the preset update conditions are met. The human-computer interaction and traceability unit is used to provide data visualization and product traceability.

[0008] In the aforementioned intelligent control system for surface grinding processes, the dynamic threshold includes the maximum allowable load value for the current machining stage. and the upper limit of load rise rate , where k represents the rough grinding, semi-fine grinding, fine grinding or sparkless grinding stage; The threshold for the sparkless stage is set to 1.05-1.15 times the baseline of the machine tool's no-load power.

[0009] In the aforementioned intelligent control system for surface grinding process, the real-time load features extracted by the real-time signal acquisition and feature extraction unit include: Normalized Load Index This is used to characterize the relative level of the current grinding load; Load rise rate This is used to reflect the changing trend of grinding load; The auxiliary signals include at least one of the following: high-frequency acoustic emission signal, spindle vibration signal, and servo current signal, which are used to reflect transient changes in abrasive cutting, and are used to assist in verifying the effectiveness of the load characteristics.

[0010] In the aforementioned intelligent control system for surface grinding processes, the process identification and dynamic threshold matching unit uses a joint discrimination mechanism to identify the current processing stage, specifically: Read the program segment number currently being executed by the CNC system and determine whether it falls within the preset range of rough grinding, semi-fine grinding, fine grinding, or sparkless grinding program segments; Monitor the real-time position coordinates and rate of change of the grinding feed axis. When the grinding feed axis coordinates have not reached the end position and the feed speed is not zero, the feed grinding stage is determined by combining the stable fluctuation range of the load characteristics. The feed grinding stage includes rough grinding, semi-finish grinding or finish grinding stages. When the grinding feed axis reaches the end position and the load characteristics show a decreasing trend to approach the no-load value, it is determined to be the sparkless grinding stage. The dynamic threshold corresponding to this stage is automatically matched based on the judgment result. and .

[0011] In the aforementioned intelligent control system for surface grinding processes, the dynamic anti-burn adaptive control unit executes a graded intervention strategy: Level 1 intervention, when Slightly exceeding the limit or When the trend is abnormal, the feed rate is automatically reduced to decrease the material removal rate and heat accumulation per unit time; Level 2 intervention: If the load continues to exceed the limit after Level 1 intervention is executed, output a pause feed command and extend the sparkless grinding time; Level 3 intervention: When the preset dressing trigger conditions are met, an automatic grinding wheel dressing subroutine instruction is output, and the cooling interface is activated for forced cooling. The adjustment triggering condition adopts a judgment logic that combines the energy integration method and the trend method. Adjustment is triggered when any of the following conditions are met: Continuous overtime ; Cumulative excess energy ; Load increase rate within the preset time window .

[0012] In the aforementioned intelligent control system for surface grinding processes, the preset update conditions in the quality closed-loop and process database reverse writing unit are logical combination conditions: Condition A: The graded intervention command of the dynamic anti-burn adaptive control unit was triggered during the processing, and the final quality inspection results were all qualified. The quality inspection included dimensional accuracy, flatness, surface roughness and burn detection. Condition B: No graded intervention instructions were triggered during the processing, and all final quality inspection results were qualified; When either condition A or condition B is met, the system determines it as a successful sample and writes the current operating condition vector, the final set of execution parameters, and the quality inspection data back to the process database. If neither condition A nor condition B is met, it is determined as a failed sample, and abnormal data is recorded for optimizing the process parameter boundaries, but the parameter mapping model is not updated.

[0013] In the aforementioned intelligent control system for surface grinding processes, the parameter mapping model is updated using an exponentially weighted moving average algorithm with a forgetting factor. The updated formula is ,in, Updated weight values, The model weights before the update. The gradient weights are calculated based on the mapping relationship of the current successful samples. This is the forgetting coefficient, with a value ranging from 0.9 to 0.95.

[0014] In the above-mentioned intelligent control system for surface grinding process, the human-machine interaction and traceability unit supports the traceability of the workpiece processing process and provides functions such as load monitoring, threshold monitoring, intervention log recording and process version management.

[0015] A method for intelligent control of grinding processes, characterized in that the method is executed collaboratively by the slow-loop decision-making and self-learning layer and the fast-loop real-time monitoring and intervention layer, and includes the following steps: Slow loop processing phase: S1: Input the workpiece material, geometry, grinding allowance, target surface roughness, and grinding wheel material to establish the initial working condition vector. ; S2: Based on the process database, a set of staged grinding and dressing parameters is derived through a parameter mapping model. The system calculates dynamic thresholds for burn risk at each stage, including a maximum load limit. and the upper limit of load rise rate ; S3: The parameter set is post-processed to generate an automatic cyclic grinding program and a grinding wheel dressing program, and then downloaded to the CNC system for execution; The fast-loop processing phase executes cyclically with a period of 10ms to 100ms. S4: During the grinding cycle, the spindle power signal and related auxiliary signals are acquired in real time, and the normalized load index is extracted. and load increase rate ; S5: Based on the current program segment number, grinding feed axis position coordinates, and feed status, identify the current process stage and match the corresponding dynamic threshold. and ; S6: If Slightly exceeding the limit or If the trend is abnormal, a level one intervention is triggered, and an instruction to automatically reduce the feed rate is output. If the load continues to exceed the limit after the first-level intervention, the second-level intervention is triggered, outputting a pause feed command and extending the sparkless grinding time; if the dressing trigger condition is met, the third-level intervention is triggered, outputting an automatic insertion of the grinding wheel dressing subroutine command and linking to open the cooling interface for forced cooling. After intervention, return to step S4 to continue monitoring; if no limits are exceeded, proceed normally until processing is complete; Slow-loop feedback phase: S7: After processing, perform quality inspections on dimensions, flatness, surface roughness, and burn condition; S8: Determine whether the preset update conditions are met, wherein the preset update conditions are "intervention has occurred and the quality is qualified" or "no intervention has occurred and the quality is qualified"; If satisfied, then , Quality data is written back to the process database, and the parameter mapping model is updated using an exponentially weighted moving average algorithm with a forgetting factor. If the conditions are not met, record the abnormal sample and terminate the current learning process.

[0016] Compared with existing technologies, this intelligent control system and method for surface grinding processes adopts a fast-slow dual-loop collaborative architecture. The slow loop is responsible for offline derivation of process parameters and iterative updates of the model, while the fast loop performs real-time signal acquisition and hierarchical intervention with a cycle of 10ms to 100ms. The two loops are connected through a data interface to achieve parameter transmission and feedback, fundamentally overcoming the defect of the decision-making system and monitoring system being isolated from each other in existing technologies. Addressing the thermal sensitivity differences in rough grinding, semi-finish grinding, finish grinding, and sparkless grinding stages, a multi-stage dynamic threshold strategy is adopted. Based on a joint discrimination mechanism of program segment number, grinding feed axis position, and feed state, the upper limit of load and the upper limit of load rise rate for each stage are matched in real time, achieving a differentiated optimal balance between processing efficiency and burn prevention. A three-level progressive intervention strategy is constructed, from reducing the feed rate and pausing the feed to extend the finishing grinding to inserting the grinding wheel for dressing and linkage cooling. Combined with the triple dressing trigger criteria of time accumulation, energy accumulation, and trend change, this approach balances the dual needs of continuous processing and timely loss prevention. Through an exponential mechanism with a forgetting factor... The weighted moving average algorithm iteratively updates the parameter mapping model, writing only successful samples that have undergone intervention and passed or have not undergone intervention and passed back to the database. This ensures that the experience gained from each successful machining operation is transformed into the driving force for model optimization, effectively solidifying process experience and enabling continuous system evolution. Simultaneously, it integrates multi-source signals such as spindle power, acoustic emission, vibration, and servo current, using normalized load index and rate of rise as core monitoring features. Through cross-validation with auxiliary signals, it effectively improves the robustness and anti-interference capability of monitoring under complex working conditions. The model inference and real-time control are decoupled, with the fast loop executing only deterministic logic operations, avoiding the risk of intervention lag caused by model calculation delays and ensuring the determinism of real-time control. The system can be directly deployed on CNC systems, generating G-code and M-code through post-processing and calling the machine tool's standard interface to execute interventions without hardware modification, exhibiting good compatibility and portability. Coupled with full-process traceability, load threshold monitoring, intervention logs, and process version management functions, it achieves transparent control of the machining process. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of an intelligent control system for the grinding process of a surface grinder.

[0018] Figure 2 This is a schematic diagram of an intelligent grinding process module.

[0019] Figure 3 This is a schematic diagram of the functional units and data flow of the intelligent grinding process module.

[0020] Figure 4 Flowchart of intelligent control method for grinding process.

[0021] In the diagram, 1. Intelligent grinding process module; 2. Surface grinder; 3. Grinding spindle; 4. Grinding wheel. Detailed Implementation

[0022] The following are specific embodiments of the present invention, which are described in conjunction with the accompanying drawings. However, the present invention is not limited to these embodiments.

[0023] like Figure 1 The diagram shows a schematic of an intelligent control system for the grinding process of a surface grinder. This system mainly consists of a surface grinder and an intelligent grinding process module. The surface grinder has a linkage mechanism for executing the X, Y, and Z linear axes, and also includes a grinding spindle and grinding wheels.

[0024] like Figure 2 As shown, the intelligent grinding process module adopts a dual-loop processing architecture, including a slow-loop decision-making and self-learning layer and a fast-loop real-time monitoring and intervention layer. The slow-loop decision-making and self-learning layer has a processing cycle corresponding to the single-workpiece machining cycle, responsible for offline derivation of process parameters and iterative model updates. The fast-loop real-time monitoring and intervention layer has a processing cycle of 10ms to 100ms, responsible for real-time suppression of load surges during machining to prevent burns. The two loops communicate via a data interface to transmit parameters and record intervention feedback. The intelligent grinding process module mainly consists of a process input and workpiece modeling unit, an intelligent process decision-making unit, an automatic cyclic grinding and wheel dressing unit, a real-time signal acquisition and feature extraction unit, a process identification and dynamic threshold matching unit, a dynamic burn prevention adaptive control unit, a quality closed-loop and process database reverse writing unit, and a human-machine interaction and traceability unit.

[0025] (1) Process input and workpiece modeling unit: used to receive process parameters including workpiece material, geometric dimensions, grinding allowance, target surface roughness, target flatness and grinding wheel material, and generate a vector of the current machining condition; (2) Intelligent process decision unit: Based on the process database, the parameter mapping model is used to derive and calculate the working condition vector, and output the grinding and wheel dressing parameter set for rough grinding, semi-finish grinding, fine grinding and sparkless grinding stages. At the same time, dynamic thresholds with burn risk for each stage are generated, including the upper limit of load. and the upper limit of load rise rate ; (3) Automatic cyclic grinding and grinding wheel dressing unit, which generates machine tool executable G code and M code by post-processing the above parameter set, and realizes the setting of process segment switching, tool compensation and dressing time interval; (4) Real-time signal acquisition and feature extraction unit: During the machining process, the spindle power signal and related auxiliary signals are acquired, and after noise reduction and filtering, the effective value, mean and rate of change and other feature quantities are calculated, and the normalized load index is extracted. and load increase rate To construct dynamic characteristics that reflect grinding load; (5) The process identification and dynamic threshold matching unit identifies the rough grinding, semi-finish grinding, fine grinding and sparkless grinding stages based on the current program segment, grinding feed axis position and feed status, and automatically calls the dynamic threshold corresponding to the stage. and ; (6) The dynamic anti-burn adaptive control unit compares the extracted load characteristics with the dynamic threshold. When it detects that the limit is exceeded or the trend is exceeding the limit, it outputs graded intervention commands, including adjusting the feed rate, extending the sparkless grinding time, triggering wheel dressing and linkage to enhance cooling. (7) The quality closed-loop and process database reverse writing unit performs quality inspections such as dimensions, flatness, surface roughness, and burn condition after processing. It records successful samples after intervention and also includes successful samples without intervention in the update range. If the update conditions are met, the system writes the parameter set, working condition vector, and intervention sequence into the process database and updates the parameter mapping model to achieve model optimization; (8) The human-machine interaction and traceability unit supports traceability by workpiece and provides relevant load, threshold monitoring, intervention log recording and process version management functions.

[0026] like Figure 3 The functional units and data flow diagram of the intelligent grinding process module are shown. This system adopts a fast-loop and slow-loop architecture. The slow-loop decision-making and self-learning layer has a longer processing cycle and is responsible for optimization from process input to parameter generation. Its process includes working condition modeling, process decision derivation, generation of initial parameters and thresholds, and generation of CNC programs. After machining is completed, if the quality inspection passes and the conditions are met, the successful sample will be written back to the process database for iterative optimization of the working condition parameters for the next cycle. The processing cycle of the fast-loop real-time monitoring and intervention layer is preferably 10ms to 100ms. It is responsible for real-time control during product processing, with the goal of suppressing load surges and preventing burns within milliseconds. Its process includes: signal acquisition, feature fusion, stage identification and threshold matching, over-limit graded intervention, and cyclic recovery.

[0027] like Figure 4 As shown, the intelligent control method for the grinding process of this surface grinder includes the following steps: S1: Working Condition Modeling: Input material, geometry, grinding allowance, target surface roughness, and grinding wheel material to generate an initial working condition vector. ; S2: Parameter Derivation: Based on Initial Working Condition Vector Within the intelligent process decision-making unit, based on the process database, the current parameter mapping model weights are used. Reasoning is performed, and the intelligent process decision-making unit outputs the parameter sets for each stage. and dynamic threshold ; S3: Program Generation: Post-processing generates the main program for the grinding cycle and the dressing subroutine; S4: Real-time monitoring: The CNC system executes the machining program, collects the spindle load signal and related auxiliary signals in real time, and extracts the normalized load index. and its rate of increase ; S5: Stage Matching: Identifies the process stage based on the current program segment, axis position, and feed status, and calls the corresponding dynamic threshold. The system employs a multi-stage dynamic threshold strategy. In the rough grinding stage, a higher load level and a wider threshold range are allowed to maximize efficiency. In the semi-finish grinding, finish grinding, and sparkless grinding stages, the system is highly sensitive to heat, and the threshold settings are strict. Especially in the sparkless grinding stage, the theoretical depth of cut approaches zero, and any abnormal load is considered high-risk. S6: Tiered Intervention: Assessment or No more If the limit is not exceeded, continue execution; If the limit is exceeded, Level 1, Level 2, and Level 3 intervention strategies will be triggered based on the degree of exceedance. The logic of tiered intervention is as follows: 1. Primary intervention: When Slightly exceeding the limit or When the trend is abnormal, the feed rate is automatically reduced to decrease the material removal rate and heat accumulation per unit time; 2. Secondary intervention: If the load continues to exceed the limit after the primary intervention, the system will pause the feed and extend the sparkless grinding time to release elastic deformation and reduce local heat accumulation by idling. 3. Level 3 intervention: If Level 2 intervention is ineffective and exceeds the time limit. Exceeding the set threshold The system automatically inserts the grinding wheel dressing subroutine and activates the cooling interface accordingly. The conditions for triggering the repair are as follows: The judgment logic combines energy integration and trend analysis. Adjustment is triggered when any of the following conditions are met: 1. Accumulated Time: Continuous Overtime ; 2. Energy Accumulation: Accumulated excess energy ,in This represents the load limit for the k-th stage. 3. Trend abrupt change: Within a preset time window, the rate of increase... ; S7 Quality Assessment: After processing, inspect dimensional accuracy, flatness, surface roughness, and surface burn condition.

[0028] S8 self-learning update: A forgetting factor algorithm based on exponentially weighted moving average is used to adjust model weights. The formula is as follows: ,in For the updated weight values, The forgetting coefficient ranges from 0.9 to 0.95. The weight value, calculated based on the mapping relationship between the parameter set of the current successful sample and the input working condition vector, is triggered only when processing is successful. The system will... The parameter set and quality inspection data are written into the process database, and the model weights are updated. The above write-back and update operations are only performed if one of the following conditions is met: intervention occurred and the final quality is qualified, or no intervention occurred but the final quality is qualified. If the process fails, it is recorded as an abnormal sample for optimizing the range of process parameters, but the parameter mapping model is not directly updated.

[0029] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

[0030] Although this document uses a considerable amount of technical terms, the possibility of using other terms is not excluded. These terms are used merely for the convenience of describing and explaining the essence of the invention; interpreting them as any additional limitation would be contrary to the spirit of the invention.

Claims

1. An intelligent control system for the grinding process of a surface grinder, characterized in that, It includes a surface grinder and an intelligent grinding process module. The surface grinder has a linkage mechanism for executing X, Y, and Z linear axes, and also includes a grinding spindle and a grinding wheel. The intelligent grinding process module adopts a dual-loop processing architecture, including a slow-loop decision-making and self-learning layer and a fast-loop real-time monitoring and intervention layer. The slow-loop decision-making and self-learning layer, with a processing cycle on the order of a single workpiece machining cycle, is used to perform offline derivation of process parameters and iterative model updates, and includes: The process input and workpiece modeling unit is used to receive workpiece material and processing condition information and generate corresponding workpiece processing condition vectors. The intelligent process decision unit is communicatively connected to the process input and workpiece modeling unit. Based on the process database, it uses a parameter mapping model to derive and generate a set of grinding parameters and dynamic thresholds corresponding to each stage of rough grinding, semi-fine grinding, fine grinding, and sparkless grinding. An automatic cyclic grinding and wheel dressing unit is used to process the grinding parameter set to generate a machining program and dressing subroutine that can be executed by the machine tool, and then download it to the CNC system. The fast-loop real-time monitoring and intervention layer, with a processing cycle of 10ms to 100ms, is used to suppress sudden load increases during processing to prevent burns. It includes: The real-time signal acquisition and feature extraction unit is used to acquire the machine tool spindle power signal and extract real-time load features during the machining process; The process identification and dynamic threshold matching unit is used to identify the current processing stage and call the corresponding dynamic threshold. The dynamic burn prevention adaptive control unit is used to output graded intervention commands to the feed rate adjustment interface, cooling control interface and program jump interface of the CNC system based on the comparison results of real-time load characteristics and dynamic thresholds. The intelligent grinding process module also includes a cross-ring data interaction mechanism: The quality closed-loop and process database reverse writing unit is set at the end of the slow loop decision and self-learning layer. It is used to perform quality inspection after processing and to reverse write successful samples to the process database to update the parameter mapping model when the preset update conditions are met. The human-computer interaction and traceability unit is used to provide data visualization and product traceability.

2. The intelligent control system for surface grinding process according to claim 1, characterized in that, The dynamic threshold includes the maximum allowable load value for the current processing stage. and the upper limit of load rise rate , where k represents the rough grinding, semi-fine grinding, fine grinding or sparkless grinding stage; The threshold for the sparkless stage is set to 1.05-1.15 times the baseline of the machine tool's no-load power.

3. The intelligent control system for surface grinding process according to claim 1, characterized in that, The real-time load features extracted by the real-time signal acquisition and feature extraction unit include: Normalized Load Index This is used to characterize the relative level of the current grinding load; Load rise rate This is used to reflect the changing trend of grinding load; The real-time signal includes at least one of the following: a high-frequency acoustic emission signal, a spindle vibration signal, and a servo current signal, which are used to reflect the transient changes in abrasive cutting, and is used to assist in verifying the effectiveness of the load characteristics.

4. The intelligent control system for surface grinding process according to claim 1, characterized in that, The process identification and dynamic threshold matching unit uses a joint discrimination mechanism to identify the current processing stage, specifically: Read the program segment number currently being executed by the CNC system and determine whether it falls within the preset range of rough grinding, semi-fine grinding, fine grinding, or sparkless grinding program segments; Monitor the real-time position coordinates and rate of change of the grinding feed axis. When the grinding feed axis coordinates have not reached the end position and the feed speed is not zero, the feed grinding stage is determined by combining the stable fluctuation range of the load characteristics. The feed grinding stage includes rough grinding, semi-finish grinding or finish grinding stages. When the grinding feed axis reaches the end position and the load characteristics show a decreasing trend to approach the no-load value, it is determined to be the sparkless grinding stage. The dynamic threshold corresponding to this stage is automatically matched based on the judgment result. and .

5. The intelligent control system for surface grinding process according to claim 1, characterized in that, The dynamic burn prevention adaptive control unit executes a graded intervention strategy: Level 1 intervention, when Slightly exceeding the limit or When the trend is abnormal, the feed rate is automatically reduced to decrease the material removal rate and heat accumulation per unit time; Level 2 intervention: If the load continues to exceed the limit after Level 1 intervention is executed, output a pause feed command and extend the sparkless grinding time; Level 3 intervention: When the preset dressing trigger conditions are met, an automatic grinding wheel dressing subroutine instruction is output, and the cooling interface is activated for forced cooling. The adjustment triggering condition adopts a judgment logic that combines the energy integration method and the trend method. Adjustment is triggered when any of the following conditions are met: Continuous overtime ; Cumulative excess energy ; Load increase rate within the preset time window .

6. The intelligent control system for surface grinding process according to claim 1, characterized in that, The preset update conditions in the quality closed-loop and process database reverse writing unit are logical combination conditions: Condition A: The graded intervention command of the dynamic anti-burn adaptive control unit was triggered during the processing, and the final quality inspection results were all qualified. The quality inspection included dimensional accuracy, flatness, surface roughness and burn detection. Condition B: No graded intervention instructions were triggered during the processing, and all final quality inspection results were qualified; When either condition A or condition B is met, the system determines it as a successful sample and writes the current operating condition vector, the final set of execution parameters, and the quality inspection data back to the process database. If neither condition A nor condition B is met, it is determined as a failed sample, and abnormal data is recorded for optimizing the process parameter boundaries, but the parameter mapping model is not updated.

7. The intelligent control system for surface grinding process according to claim 1, characterized in that, The parameter mapping model is updated using an exponentially weighted moving average algorithm with a forgetting factor. The updated formula is ,in, Updated weight values, The model weights before the update. The gradient weights are calculated based on the mapping relationship of the current successful samples. This is the forgetting coefficient, with a value ranging from 0.9 to 0.

95.

8. The intelligent control system for surface grinding process according to claim 1, characterized in that, The human-computer interaction and traceability unit supports the traceability of the workpiece processing process and provides functions such as load monitoring, threshold monitoring, intervention log recording, and process version management.

9. A method for intelligent control of grinding process based on the system according to any one of claims 1 to 8, characterized in that, The method is executed collaboratively by the slow-loop decision-making and self-learning layer and the fast-loop real-time monitoring and intervention layer, and includes the following steps: Slow loop processing phase: S1: Input the workpiece material, geometry, grinding allowance, target surface roughness, and grinding wheel material to establish the initial working condition vector. ; S2: Based on the process database, a set of staged grinding and dressing parameters is derived through a parameter mapping model. The system calculates dynamic thresholds for burn risk at each stage, including a maximum load limit. and the upper limit of load rise rate ; S3: The parameter set is post-processed to generate an automatic cyclic grinding program and a grinding wheel dressing program, and then downloaded to the CNC system for execution; The fast-loop processing phase executes cyclically with a period of 10ms to 100ms. S4: During the grinding cycle, the spindle power signal and related auxiliary signals are acquired in real time, and the normalized load index is extracted. and load increase rate ; S5: Based on the current program segment number, grinding feed axis position coordinates, and feed status, identify the current process stage and match the corresponding dynamic threshold. and ; S6: If Slightly exceeding the limit or If the trend is abnormal, a level one intervention is triggered, and an instruction to automatically reduce the feed rate is output. If the load continues to exceed the limit after the first-level intervention, the second-level intervention is triggered, outputting a pause feed command and extending the sparkless grinding time; if the dressing trigger condition is met, the third-level intervention is triggered, outputting an automatic insertion of the grinding wheel dressing subroutine command and linking to open the cooling interface for forced cooling. After intervention, return to step S4 to continue monitoring; if no limits are exceeded, proceed normally until processing is complete; Slow-loop feedback phase: S7: After processing, perform quality inspections on dimensions, flatness, surface roughness, and burn condition; S8: Determine whether the preset update conditions are met, wherein the preset update conditions are "intervention has occurred and the quality is qualified" or "no intervention has occurred and the quality is qualified"; If satisfied, then , Quality data is written back to the process database, and the parameter mapping model is updated using an exponentially weighted moving average algorithm with a forgetting factor. If the conditions are not met, record the abnormal sample and terminate the current learning process.

Images