Thermal-mechanical-vibration synergistically inhibited centerless grinding machine intelligent monitoring and control system

Abstract

The application discloses a heat-force-vibration synergistic inhibition centerless grinder intelligent monitoring and regulation system, relates to the technical field of monitoring and regulation, and realizes real-time collection of heat, force and vibration signals of the centerless grinder through a plurality of source sensors, synchronously divides the signals by using a sliding time window, and extracts multimodal features; based on the features, complementary enhancement coefficients and time coupling coefficients are calculated, current fusion features are compared with historical training data distribution, and a drift measure is calculated; the fusion effect is judged comprehensively according to the above indexes, if qualified, real-time intelligent monitoring and regulation of the centerless grinder is realized according to the heat-force-vibration synergistic model prediction result, and stable and efficient operation of the system is ensured; in this way, an evaluation mechanism for directly judging the fusion quality according to the final prediction effect of the heat-force-vibration synergistic model is realized, fusion failure risk is accurately identified, it is guided whether re-fusion is needed, and the self-adaptive ability and stability are improved.

Description

Technical Field

[0001] This invention relates to the field of monitoring and control technology, specifically to an intelligent monitoring and control system for centerless grinding machines with synergistic suppression of thermal-mechanical-vibration forces. Background Technology

[0002] In the machining of high-precision parts, centerless grinders are widely used for grinding the outer diameter of shafts and sleeves due to their ability to operate without clamping and their suitability for large-volume continuous machining. However, the grinding process is essentially a complex multi-physics coupling process, involving thermal effects (grinding heat accumulation between the workpiece and the grinding wheel), mechanical loads (dynamic fluctuations in grinding force), and structural vibrations (micro-displacements and resonances between the grinding wheel and the workpiece). These factors interact and reinforce each other, easily leading to quality problems such as machining burns, chatter marks, and dimensional drift.

[0003] To address this, a collaborative thermo-mechanical-vibration suppression intelligent monitoring and control technology for centerless grinders has been proposed in recent years. Its core objective is to simultaneously acquire thermal, mechanical, and vibrational signals from multiple sensors, construct a thermo-mechanical-vibration coupling relationship using a multi-modal data fusion model, and based on this, predict and judge critical states (such as burn precursors, excessive vibration, and thermal deformation trends) in advance. Based on the prediction results, real-time control can be achieved by linking servo drives, hydraulic systems, and cooling units to prevent quality problems and ensure machining stability and accuracy. Currently, most practical deployments use deep learning or data-driven thermo-mechanical-vibration collaborative models for feature extraction and state prediction. These models fuse information from the three modes of thermo-mechanical-vibration to jointly predict and judge the state of the centerless grinder and adjust it in advance.

[0004] Therefore, in the thermo-mechanical-vibration synergistic model, the fusion effect of the three modal information types (thermal, mechanical, and vibration) directly determines the subsequent state prediction and control of the centerless grinder. If the fusion effect of these three modal information types is flawed, it may lead to problems in the subsequent state prediction and control of the centerless grinder.

[0005] Current methods typically rely solely on model prediction performance (such as accuracy and false alarm rate) to indirectly assess the quality of thermal-mechanical-oscillatory data fusion. This approach has two drawbacks: first, a decline in model prediction performance does not necessarily indicate fusion failure; deterioration can be caused by various factors, such as changes in operating conditions, sensor drift, data quality fluctuations, and sample distribution shifts, and does not necessarily mean the fusion quality itself is flawed; second, model retraining is costly. Once prediction performance is deemed "unsatisfactory," current methods often directly retrain the fusion model, which usually requires a large amount of new data, manual annotation, and redeployment, resulting in high costs, long cycles, and low efficiency. Therefore, there is an urgent need to construct an evaluation mechanism that does not rely on the final prediction result but can directly assess the quality of fusion to accurately identify the risk of fusion failure, guide whether re-fusion is necessary, and improve adaptability and stability. Summary of the Invention

[0006] The purpose of this invention is to solve the problems mentioned above and provide an intelligent monitoring and control system for centerless grinding machines with coordinated suppression of thermal, mechanical, and vibration forces.

[0007] In a first aspect of this invention, a smart monitoring and control system for centerless grinding machines with synergistic suppression of thermal, mechanical, and vibrational forces is first proposed, the system comprising:

[0008] Data acquisition module: Real-time acquisition of thermal, force, and vibration signals during the centerless grinding process using multi-source sensors;

[0009] The segmentation module: the three types of signals are synchronously segmented using a sliding time window, and the corresponding thermo-mechanical-vibrational multimodal features are extracted within each window;

[0010] Complementary enhancement module: Based on the extracted thermo-mechanical-vibrational multimodal features, the complementary enhancement coefficient of the multimodal features is calculated to evaluate the mutual information strength and redundancy between different modes;

[0011] Temporal Coupling Module: Performs correlation analysis on time-aligned multimodal features and calculates the temporal coupling coefficient of the multimodal features to measure the synchronicity and coordinated change trend between different modes;

[0012] Feature distribution drift module: compares the multimodal fusion features within the current time window with the fusion feature distribution of historical training data, and calculates the feature distribution drift metric;

[0013] Monitoring and control module: Based on the complementary enhancement coefficient, time coupling coefficient and characteristic distribution drift metric, determine whether the data fusion effect of the current thermo-mechanical-vibration synergistic model is qualified. If the data fusion effect is qualified, monitor and control the centerless grinder according to the results predicted by the thermo-mechanical-vibration synergistic model.

[0014] Optionally, the calculation steps for the complementary enhancement coefficients of multimodal features are as follows:

[0015] For a sliding window t, construct three types of feature vectors: thermal, mechanical, and vibrational. Thermal feature vector... Force eigenvectors Vibration characteristic vector ;

[0016] The mutual information between pairwise modes is calculated using nonparametric estimation methods, as follows: ; ; ; , and These represent the mutual information between thermal and mechanical characteristics, between thermal and vibrational characteristics, and between mechanical and vibrational characteristics, respectively.

[0017] The three modality feature vectors are concatenated into a fused feature. ;Calculation based on kernel density estimation entropy The calculation formula is: In the formula, Here, N is the kernel function (such as a Gaussian kernel), and N is the total number of time windows.

[0018] Calculate the complementary enhancement coefficient of multimodal features The calculation formula is: .

[0019] Optionally, the calculation steps for the temporal coupling coefficient of multimodal features are as follows:

[0020] The real-time collected heat, force, and vibration signals are respectively divided into sliding time windows. Perform synchronous segmentation; set the length of each time window to... Thus, in the t-th window, the modal characteristic sequences of the three types of signals can be obtained; thermal signal characteristic sequence: Force signal characteristic sequence: Vibration signal characteristic sequence: ,

[0021] N is the total number of time windows;

[0022] Calculate the first-order difference sequence of each modal signal over the time series: Thermal modal difference: Force mode difference: Vibration mode difference: ;

[0023] The difference sequences of the three modes were compared pairwise, and the temporal coupling degree was measured using the dynamic time warping distance (DTW). The three DTW distances are as follows: ; ; ;

[0024] Calculate the time coupling coefficient of multimodal features The calculation formula is: .

[0025] Optionally, the calculation steps for the characteristic distribution drift metric are as follows:

[0026] Within each sliding time window, the fused multimodal features are extracted from the thermal, force, and vibration signals and used as the current real-time feature set.

[0027] Extract fusion features from historical data saved during the model training phase and use them as a reference feature set;

[0028] We calculate the mean, covariance, and other basic statistics of the current feature set and the reference feature set respectively to characterize their probability distribution.

[0029] The KL divergence is used to calculate the relative entropy difference between the current feature distribution and the reference feature distribution, which serves as a measure of feature distribution drift.

[0030] Optionally, the data fusion effect of the current thermo-mechanical-vibrational synergy model is judged based on the complementary enhancement coefficient, time coupling coefficient, and characteristic distribution drift metric. If the data fusion effect is satisfactory, the steps for monitoring and controlling the centerless grinding machine based on the results predicted by the thermo-mechanical-vibrational synergy model are as follows:

[0031] The fusion pass coefficient is calculated based on the complementary enhancement coefficient, the temporal coupling coefficient, and the feature distribution drift metric. The fusion pass coefficient is compared with the preset fusion pass coefficient threshold, and the data fusion effect of the current thermo-mechanical-vibration synergistic model is judged based on the comparison results.

[0032] Optionally, the steps for calculating the fusion pass coefficient based on the complementary enhancement coefficient, the temporal coupling coefficient, and the feature distribution drift metric are as follows:

[0033] The complementary enhancement coefficient, temporal coupling coefficient, and feature distribution drift metric are dimensionlessly normalized to be within the range [0,1]. The weighted sum of these normalized coefficients yields the fusion pass coefficient, calculated using the following formula: In the formula, To achieve the qualification coefficient, These are the normalized complementary enhancement coefficient, temporal coupling coefficient, and characteristic distribution drift metric, respectively.

[0034] Optionally, the steps for comparing the fusion pass coefficient with a preset fusion pass coefficient threshold and determining whether the data fusion effect of the current thermal-mechanical-vibration synergistic model is qualified based on the comparison result are as follows:

[0035] The fusion pass coefficient is compared with the preset fusion pass coefficient threshold. If the fusion pass coefficient is not less than the preset fusion pass coefficient threshold, it means that the data fusion effect of the current thermo-mechanical-vibration collaborative model is qualified. Then, the centerless grinder is monitored and controlled according to the results predicted by the thermo-mechanical-vibration collaborative model.

[0036] If the fusion pass coefficient is less than the preset fusion pass coefficient threshold, it means that the data fusion effect of the current thermal-mechanical-vibration collaborative model is not qualified. The monitoring and control operations will no longer rely on the prediction results of the current thermal-mechanical-vibration collaborative model. At this time, the thermal-mechanical-vibration collaborative model can be automatically updated until the fusion pass coefficient is not less than the preset fusion pass coefficient threshold.

[0037] The beneficial effects of this invention are:

[0038] This invention proposes an intelligent monitoring and control system for centerless grinding machines with thermo-mechanical-vibration synergistic suppression. It collects thermal, mechanical, and vibration signals from the centerless grinding machine in real time using multi-source sensors, synchronously segments the signals using a sliding time window, and extracts multimodal features. Based on these features, a complementary enhancement coefficient is calculated to assess the complementarity of information between modes. A correlation analysis is used to calculate the temporal coupling coefficient to measure modal synchronization. A feature distribution drift module compares the current fused features with the distribution of historical training data and calculates the drift metric. The monitoring and control module comprehensively judges the fusion effect based on the above indicators. If it is satisfactory, the system performs real-time intelligent monitoring and control of the centerless grinding machine based on the prediction results of the thermo-mechanical-vibration synergistic model, ensuring stable and efficient system operation. This provides an evaluation mechanism that can directly judge the fusion quality without relying on the final prediction effect of the thermo-mechanical-vibration synergistic model, accurately identifying the risk of fusion failure, guiding whether re-fusion is needed, and improving adaptability and stability. Attached Figure Description

[0039] The invention will now be further described with reference to the accompanying drawings.

[0040] Figure 1 A framework diagram of an intelligent monitoring and control system for centerless grinding machines with synergistic suppression of thermal, mechanical, and vibrational forces. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] This invention provides an intelligent monitoring and control system for centerless grinding machines with coordinated suppression of thermal, mechanical, and vibration forces. See also... Figure 1 , Figure 1 This is a framework diagram of a centerless grinding machine intelligent monitoring and control system for coordinated suppression of thermal, mechanical, and vibration phenomena, provided in an embodiment of the present invention. The system includes the following steps:

[0043] Data acquisition module: Real-time acquisition of thermal, force, and vibration signals during the centerless grinding process using multi-source sensors;

[0044] The segmentation module: the three types of signals are synchronously segmented using a sliding time window, and the corresponding thermo-mechanical-vibrational multimodal features are extracted within each window;

[0045] Complementary enhancement module: Based on the extracted thermo-mechanical-vibrational multimodal features, the complementary enhancement coefficient of the multimodal features is calculated to evaluate the mutual information strength and redundancy between different modes;

[0046] Temporal Coupling Module: Performs correlation analysis on time-aligned multimodal features and calculates the temporal coupling coefficient of the multimodal features to measure the synchronicity and coordinated change trend between different modes;

[0047] Feature distribution drift module: compares the multimodal fusion features within the current time window with the fusion feature distribution of historical training data, and calculates the feature distribution drift metric;

[0048] Monitoring and control module: Based on the complementary enhancement coefficient, time coupling coefficient and characteristic distribution drift metric, determine whether the data fusion effect of the current thermo-mechanical-vibration synergistic model is qualified. If the data fusion effect is qualified, monitor and control the centerless grinder according to the results predicted by the thermo-mechanical-vibration synergistic model.

[0049] Based on the intelligent monitoring and control system for centerless grinding machines with thermo-mechanical-vibration synergistic suppression provided in this embodiment of the invention, an evaluation mechanism that can directly judge the quality of fusion without relying on the final prediction effect of the thermo-mechanical-vibration synergistic model can be established to accurately identify the risk of fusion failure, guide whether re-fusion is needed, and improve the adaptive capability and stability.

[0050] In one embodiment, thermal, force, and vibration signals during the centerless grinding process are acquired in real time using multi-source sensors; the three types of signals are synchronously segmented using a sliding time window, and the corresponding thermal-force-vibration multimodal features are extracted within each window;

[0051] In the centerless grinding process, to achieve comprehensive perception and intelligent analysis of the three types of physical quantities—thermal, mechanical, and vibration—multiple high-sensitivity sensors are first deployed:

[0052] Thermal signal acquisition: Infrared temperature sensors or thermocouples are installed near the contact area of ​​the grinding wheel to acquire the temperature changes of the workpiece surface and the grinding wheel area in real time during the grinding process. The sampling frequency is generally set at 50–100 Hz to balance response speed and signal stability.

[0053] Force signal acquisition: Strain gauges or multi-dimensional force sensors are installed on the support mechanism or loading area of ​​the grinding wheel spindle of the centerless grinder to monitor the cutting force, normal force, etc. generated during the grinding process and capture force fluctuations caused by changes in machining conditions. The force signal sampling rate is usually not less than 1 kHz to ensure responsiveness to short-term fluctuations.

[0054] Vibration signal acquisition: High-frequency accelerometers are installed on key structural components such as the grinding machine spindle box, bracket, and support wheel seat to acquire micro-vibration data during the grinding process. To ensure the capture of high-frequency characteristics such as chatter and impact, the vibration signal sampling rate is generally 5–20 kHz, and a high-pass filter is used to remove low-frequency structural noise.

[0055] After completing the acquisition of the above multimodal data, a sliding time window mechanism is introduced for synchronization processing to ensure the alignment and linkage analysis of the three types of signals in the time domain:

[0056] Sliding time window design: Align all modal signals according to a unified time reference to construct a fixed-length sliding time window (e.g., 1 second, 50% overlap). The window setting needs to comprehensively consider the sampling rate and physical response delay of different signals to ensure that the features of different modes within the same window are comparable and interconnected.

[0057] Synchronous segmentation and feature extraction: Within each sliding window, features are extracted after preprocessing (e.g., denoising, normalization) the three modal signals.

[0058] Extract temperature change trend indicators such as mean, slope, maximum temperature rise rate, and instantaneous fluctuation from the thermal signal;

[0059] The mean, root mean square, frequency energy distribution, and force change slope of the force signal are calculated as indicators reflecting the grinding state and material removal behavior.

[0060] Extract time-domain features (such as RMS, kurtosis, and kurtosis), frequency-domain features (main frequency band energy and spectral centroid), and time-frequency-domain features (such as wavelet packet energy ratio) from vibration signals to characterize processing stability.

[0061] Multimodal feature fusion construction: Finally, the features from the three types of signals—thermal, force, and vibration—within each time window are spliced ​​together in chronological order to form a set of multimodal collaborative feature vectors, which are used for subsequent fusion quality assessment, anomaly prediction, or model update mechanisms.

[0062] It should be noted that this step ensures the alignment of multiple modal information in the temporal dimension and their physical correlation, providing a solid data foundation for subsequent complementary enhancement analysis, coupling analysis, and fusion degradation discrimination. The sliding window mechanism not only improves the temporal representation capability of features but also enhances the model's response to instantaneous anomalies and trend changes, making it the core data preprocessing step in the entire intelligent monitoring process.

[0063] In one embodiment, based on the extracted thermo-mechanical-vibrational multimodal features, the complementary enhancement coefficient of the multimodal features is calculated to evaluate the mutual information strength and redundancy between different modes;

[0064] Specifically, the calculation steps for the complementary enhancement coefficient of multimodal features are as follows:

[0065] For a sliding window t, construct three types of feature vectors: thermal, mechanical, and vibrational. Thermal feature vector... Force eigenvectors Vibration characteristic vector ;

[0066] The mutual information between pairwise modes is calculated using nonparametric estimation methods (such as k-nearest neighbor estimation or kernel density estimation), as follows: ; ; ; , and These represent the mutual information between thermal and mechanical characteristics, between thermal and vibrational characteristics, and between mechanical and vibrational characteristics, respectively.

[0067] The three modality feature vectors are concatenated into a fused feature. ;Calculation based on kernel density estimation (such as Renyi entropy approximation) entropy The calculation formula is: In the formula, Here, N is the kernel function (such as a Gaussian kernel), and N is the total number of time windows.

[0068] Calculate the complementary enhancement coefficient of multimodal features The calculation formula is: .

[0069] It is important to note that the complementarity enhancement coefficient (CEC) of modal features reflects the degree of effective fusion of complementary information between different modal features. Essentially, it measures whether the fusion of multimodal features produces more information gain than when each mode acts alone. In the calculation, the CEC uses the pairwise mutual information between the three modes (thermal, mechanical, and vibration) as the denominator to represent the original redundancy and shared information between the modes; and the overall information entropy after the three-modal fusion is used as the numerator to measure the overall information content after fusion. A larger CEC indicates a significant increase in information entropy in the fused feature space, meaning that the fusion process effectively integrates unique and complementary features from multiple modes, rather than simply repeating or redundantly stacking signals. Conversely, a smaller CEC indicates strong redundancy among the three types of signals, resulting in no significant information gain after fusion, and may even indicate "modal conflict" or fusion failure. Especially in complex nonlinear dynamics like centerless grinding, where the thermal, mechanical, and vibration signals have different source mechanisms, the model only possesses high discriminative ability when they have sufficient complementarity and can collaboratively describe the grinding state. Therefore, a higher complementary enhancement coefficient indicates that the fusion process has effectively captured key differences in different modes, enhancing information richness and discrimination ability, indicating that the current fusion effect is good and the model has strong monitoring and control capabilities within this window. Conversely, if the complementary enhancement coefficient remains low, it may indicate that there is redundancy in the thermal-mechanical-vibration fusion features or that the fusion has failed, requiring a re-examination of the fusion strategy or an update of the model structure.

[0070] It should be noted that analyzing the complementarity enhancement coefficient of modal features has the following advantages in judging the fusion effect of the three modal information of thermo-mechanical-vibration in the thermo-mechanical-vibration synergistic model: First, it directly quantifies the degree of improvement in information entropy after fusion. Compared with relying solely on model prediction results, it can better reveal the true complementarity and information gain between multimodal features, avoiding attributing prediction errors to fusion failure. Second, it provides an immediate feedback mechanism for fusion quality, enabling the early detection of potential risks before the fusion effect deteriorates, which helps to dynamically optimize the fusion strategy or trigger an adaptive refusion process, improving robustness. Third, this coefficient has a certain degree of model independence, enabling cross-model and cross-task evaluation of the advantages and disadvantages of different fusion schemes, providing a stable and interpretable fusion quality reference indicator for subsequent decision-making modules. In summary, the complementarity enhancement coefficient, as a fusion quality assessment method, helps to determine the reliability of the fusion mechanism from the underlying feature level, supporting the evolution of centerless grinding machine monitoring towards intelligence, self-diagnosis, and self-recovery.

[0071] In one embodiment, correlation analysis is performed on time-aligned multimodal features to calculate the temporal coupling coefficient of the multimodal features, which is used to measure the synchronicity and cooperative change trend between different modes.

[0072] The calculation steps for the temporal coupling coefficient of multimodal features are as follows:

[0073] The real-time collected heat, force, and vibration signals are respectively divided into sliding time windows. Perform synchronous segmentation; set the length of each time window to... Thus, in the t-th window, the modal characteristic sequences of the three types of signals can be obtained; thermal signal characteristic sequence: Force signal characteristic sequence: Vibration signal characteristic sequence: ,

[0074] N is the total number of time windows;

[0075] Calculate the first-order difference sequence of each modal signal over the time series: Thermal modal difference: Force mode difference: Vibration mode difference: ;

[0076] The difference sequences of the three modes were compared pairwise, and the temporal coupling degree was measured using the dynamic time warping distance (DTW). The three DTW distances are as follows: ; ; ;

[0077] Calculate the time coupling coefficient of multimodal features The calculation formula is: .

[0078] It's important to note that the temporal coupling coefficient of multimodal features is an indicator used to measure the synchronicity and cooperative trend of changes in thermal, mechanical, and vibrational modal features over time. Essentially, it reflects the temporal coupling strength of different modes during dynamic processes. A larger coefficient indicates more consistent change patterns among the three modes, meaning they exhibit high consistency and linkage within the same time window. This suggests strong coupling between the physical processes of each mode, and the fused features better represent the actual grinding conditions. Conversely, a smaller temporal coupling coefficient indicates divergent change trends and inconsistent paces between modes, potentially indicating interference in some modes, sensor distortion, or coupling mechanism failure, leading to the fused features losing global representativeness. Therefore, TCC can directly reflect the quality of fusion from the underlying temporal structure without relying on model output. It is particularly suitable for monitoring the fusion stability and adaptability of the thermal-mechanical-vibration model during long-term operation, helping to determine whether the current fusion mechanism is still effective and whether model reconstruction or adjustment is needed to maintain high accuracy and robustness. This indicator has advantages such as strong physical interpretability, real-time assessment capability, and high model independence, making it a key structural reference parameter in fusion monitoring.

[0079] It is worth noting that analyzing the temporal coupling coefficient of multimodal features is beneficial for judging the fusion effect of the three modal information (thermal, mechanical, and vibration) in the thermo-mechanical-vibrational synergistic model. It reveals the synergistic response relationship of different modal signals during the grinding process from a temporal dynamic perspective, thus enabling structural diagnosis of fusion quality. Traditional fusion effects often rely on the final predicted performance for evaluation, making it difficult to trace the root cause of the problem. The temporal coupling coefficient, however, does not rely on model output; it only quantifies the trend of temporal consistency changes in the signals themselves, possessing greater independence and foresight. When thermal, mechanical, and vibrational signals exhibit a high degree of coupling within the same window, it indicates that the underlying physical mechanisms are interconnected, and the fused features have a consistent dynamic structural basis, which helps improve the model's ability to perceive and interpret changes in operating conditions. Therefore, the temporal coupling coefficient can not only monitor early signs of fusion failure in real time but also provide a basis for subsequent reconstruction of the fusion path or adjustment of weights, significantly improving the adaptability and robustness of the thermo-mechanical-vibrational synergy.

[0080] In one embodiment, the multimodal fusion features within the current time window are compared with the fusion feature distribution of historical training data to calculate the feature distribution drift metric.

[0081] The calculation steps for the characteristic distribution drift metric are as follows:

[0082] Within each sliding time window, the fused multimodal features are extracted from the thermal, force, and vibration signals and used as the current real-time feature set.

[0083] The fusion features are extracted from the historical data saved during the model training phase and used as a reference feature set (i.e., the feature distribution under the original working conditions).

[0084] We calculate the mean, covariance, and other basic statistics of the current feature set and the reference feature set respectively to characterize their probability distribution.

[0085] The KL divergence is used to calculate the relative entropy difference between the current feature distribution and the reference feature distribution, which serves as a measure of feature distribution drift.

[0086] It's important to note that feature distribution drift metric refers to the degree of difference in probability distribution between the currently collected fused features and the reference features from the model training phase. Essentially, it's an indicator of whether the statistical regularities in the feature space have changed significantly. Methods such as Kullback-Leibler divergence (KL divergence) are commonly used to measure the relative entropy difference between the two distributions. This metric is used to assess whether the multimodal fused features under the current operating conditions remain consistent with the original training conditions, reflecting the applicability and reliability of the model's fusion channels in new operating conditions. If the feature distribution drift metric is small, it indicates that the current fused features are highly consistent with the historical reference distribution, suggesting that the fusion relationship and expression patterns of the thermal, mechanical, and vibration modes have not undergone substantial shifts, the fusion mapping remains effective, and the information expression fully preserves the original semantics. The model can stably predict and regulate based on these fused features. Conversely, if the drift metric is large, it indicates that the fused features have become statistically disconnected from the original data, the fusion effect is weakened, and problems such as redundant enhancement or information loss may occur, making it difficult for the collaborative model to accurately perceive the actual operating conditions. Therefore, a smaller feature distribution drift metric indicates good fusion quality and is an important reference for judging the stability and generalization ability of the thermo-mechanical-vibrational synergistic model.

[0087] In one implementation,

[0088] It is important to note that feature distribution drift metric is crucial for assessing the fusion effectiveness of multimodal information in a thermo-mechanical-vibrational (TMV) collaborative model. Its advantage lies in providing an objective quantitative method independent of model prediction results, directly reflecting the degree of matching between the current actual operating conditions and the training conditions in the fused feature space. Compared to indirect judgments relying on output errors, drift metric can detect potential mismatches caused by changes in operating conditions, sensor errors, or signal anomalies during modal fusion earlier and more sensitively. This allows for proactive intervention before model prediction accuracy declines, enabling model updates or fusion reconstruction and improving the sensitivity and response speed to changes in the grinding machine's operating state. Furthermore, drift detection can be used to construct a model self-calibration mechanism, enabling online adaptive adjustment, reducing the risks of false alarms and false negatives, and improving robustness and stability. Therefore, introducing feature distribution drift metric provides a fundamental guarantee for fusion quality assessment and is of significant value in improving the generalization ability and long-term effectiveness of the TMV collaborative model.

[0089] In one embodiment, the data fusion effect of the current thermo-mechanical-vibration synergistic model is judged to be qualified based on the complementary enhancement coefficient, the time coupling coefficient, and the characteristic distribution drift metric. If the data fusion effect is qualified, the step of monitoring and controlling the centerless grinding machine according to the results predicted by the thermo-mechanical-vibration synergistic model is as follows:

[0090] The fusion pass coefficient is calculated based on the complementary enhancement coefficient, the temporal coupling coefficient, and the feature distribution drift metric. The fusion pass coefficient is compared with the preset fusion pass coefficient threshold, and the data fusion effect of the current thermo-mechanical-vibration synergistic model is judged based on the comparison results.

[0091] In one embodiment, the step of calculating the fusion pass coefficient based on the complementary enhancement coefficient, the temporal coupling coefficient, and the feature distribution drift metric is as follows:

[0092] The complementary enhancement coefficient, temporal coupling coefficient, and feature distribution drift metric are dimensionlessly normalized to be within the range [0,1]. The weighted sum of these normalized coefficients yields the fusion pass coefficient, calculated using the following formula: In the formula, To achieve the qualification coefficient, These are the normalized complementary enhancement coefficient, temporal coupling coefficient, and characteristic distribution drift metric, respectively.

[0093] It should be noted that, under normal circumstances, the complementary enhancement coefficient, temporal coupling coefficient, and feature distribution drift metric have equal weights when weighted and summed, and the sum of the weights is 1.

[0094] In one embodiment, the steps of comparing the fusion pass coefficient with a preset fusion pass coefficient threshold and determining whether the data fusion effect of the current thermal-mechanical-vibration collaborative model is qualified based on the comparison result are as follows:

[0095] The fusion pass coefficient is compared with the preset fusion pass coefficient threshold. If the fusion pass coefficient is not less than the preset fusion pass coefficient threshold, it means that the data fusion effect of the current thermo-mechanical-vibration collaborative model is qualified. Then, the centerless grinder is monitored and controlled according to the results predicted by the thermo-mechanical-vibration collaborative model.

[0096] If the fusion pass coefficient is less than the preset fusion pass coefficient threshold, it means that the data fusion effect of the current thermal-mechanical-vibration collaborative model is not qualified. The monitoring and control operations will no longer rely on the prediction results of the current thermal-mechanical-vibration collaborative model. At this time, the thermal-mechanical-vibration collaborative model can be automatically updated until the fusion pass coefficient is not less than the preset fusion pass coefficient threshold.

[0097] It should be noted that comparing the fusion pass coefficient with the preset fusion pass coefficient threshold is a crucial step in judging the fusion quality of the current thermo-mechanical-vibrational synergistic model. This method achieves quantitative judgment of the model fusion status by setting a reasonable fusion effect benchmark value (i.e., the fusion pass coefficient threshold). If the currently calculated fusion pass coefficient (derived from a comprehensive evaluation of the complementary enhancement coefficient, temporal coupling coefficient, and feature distribution drift metric) is not less than the threshold, it indicates that the multimodal information has good complementarity and temporal synergy, and the current operating conditions are basically consistent with the distribution of the training data. The data fusion effect is stable and reliable, and the prediction results of the thermo-mechanical-vibrational synergistic model can continue to be trusted, and the centerless grinding machine can be effectively monitored and controlled accordingly. Conversely, if the fusion pass coefficient is lower than the set threshold, it indicates that there may be problems such as information loss, asynchrony, mismatch, or distribution drift in the modal fusion, and the fusion quality is no longer reliable. At this point, the system will no longer rely on the prediction results of the current model to prevent serious consequences such as misjudgment, false alarms, and incorrect control. Simultaneously, it will automatically trigger the model fusion update process, performing incremental corrections, modal reconstruction, or feature filtering on the fusion structure, and continuously monitor the new fusion effect until the fusion pass coefficient rises above the set threshold. This mechanism effectively avoids the risks caused by blindly relying on prediction performance, while avoiding the cost burden of frequent, full-scale retraining. It achieves intelligent control with fusion quality awareness, adaptive update rhythm, and controllable risks, effectively enhancing the robustness, economy, and self-healing capability of the thermal-mechanical-vibration monitoring system for centerless grinding machines during long-term operation.

[0098] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A heat-force-vibration synergistically inhibited centerless grinding machine intelligent monitoring and control system, characterized in that, The system includes: Data acquisition module: Real-time acquisition of thermal, force, and vibration signals during the centerless grinding process using multi-source sensors; The segmentation module: the three types of signals are synchronously segmented using a sliding time window, and the corresponding thermo-mechanical-vibrational multimodal features are extracted within each window; Complementary enhancement module: Based on the extracted thermo-mechanical-vibrational multimodal features, the complementary enhancement coefficient of the multimodal features is calculated to evaluate the mutual information strength and redundancy between different modes; Temporal Coupling Module: Performs correlation analysis on time-aligned multimodal features and calculates the temporal coupling coefficient of the multimodal features to measure the synchronicity and coordinated change trend between different modes; Feature distribution drift module: compares the multimodal fusion features within the current time window with the fusion feature distribution of historical training data, and calculates the feature distribution drift metric; Monitoring and control module: Based on the complementary enhancement coefficient, time coupling coefficient and characteristic distribution drift metric, it determines whether the data fusion effect of the current thermo-mechanical-vibration synergistic model is qualified. If the data fusion effect is qualified, the centerless grinder is monitored and controlled according to the results predicted by the thermo-mechanical-vibration synergistic model. The data fusion effect of the current thermo-mechanical-vibration synergistic model is judged based on the complementary enhancement coefficient, time coupling coefficient, and characteristic distribution drift metric. If the data fusion effect is satisfactory, the steps for monitoring and controlling the centerless grinder based on the results predicted by the thermo-mechanical-vibration synergistic model are as follows: The fusion pass coefficient is calculated based on the complementary enhancement coefficient, the temporal coupling coefficient, and the feature distribution drift metric. The fusion pass coefficient is compared with the preset fusion pass coefficient threshold, and the data fusion effect of the current thermal-mechanical-vibration synergistic model is judged based on the comparison results. The steps for calculating the fusion pass rate based on the complementary enhancement coefficient, temporal coupling coefficient, and feature distribution drift metric are as follows: The complementary enhancement coefficient, temporal coupling coefficient, and feature distribution drift metric are dimensionlessly normalized to be within the range [0,1]. The weighted sum of these normalized coefficients yields the fusion pass coefficient, calculated using the following formula: In the formula, To achieve the qualification coefficient, These are the normalized complementary enhancement coefficient, temporal coupling coefficient, and characteristic distribution drift metric, respectively.

2. The heat-force-vibration synergistically inhibited centerless grinding machine intelligent monitoring and regulation system according to claim 1, characterized in that, The steps for calculating the complementary enhancement coefficient of multimodal features are as follows: For the sliding window t, three types of feature vectors are constructed: thermal, force, and vibration feature vectors ; force eigenvector ; vibration feature vector ; The mutual information between pairwise modes is calculated using nonparametric estimation methods, as follows: ; ; ; , and These represent the mutual information between thermal and mechanical characteristics, between thermal and vibrational characteristics, and between mechanical and vibrational characteristics, respectively. The three modality feature vectors are concatenated into a fused feature. ;Calculation based on kernel density estimation entropy The calculation formula is: In the formula, Here, N is the kernel function, and N is the total number of time windows. Computing a complementary enhancement coefficient for multi-modal features The formula for the computation is: 。 3. The heat-force-vibration synergistically suppressed centerless grinding machine intelligent monitoring and regulation system according to claim 1, characterized in that, The steps for calculating the temporal coupling coefficient of multimodal features are as follows: The real-time collected heat, force, and vibration signals are respectively divided into sliding time windows. Perform synchronous segmentation; set the length of each time window to... Thus, in the t-th window, the modal characteristic sequences of the three types of signals can be obtained; thermal signal characteristic sequence: Force signal characteristic sequence: Vibration signal characteristic sequence: , N is the total number of time windows; Compute first difference sequence of each modality signal over time series: thermal modality difference: , force modality difference: ; vibration modality difference: ; The three types of modal difference sequences are compared with each other, and the dynamic time warping distance is used to measure the time coupling degree. The three groups of DTW distances are: ; ; ; Computing time coupling coefficients for multi-modal features The formula for the computation is: 。 4. The heat-force-vibration synergistically suppressed centerless grinding machine intelligent monitoring and regulation system according to claim 1, characterized in that, The calculation steps for the characteristic distribution drift metric are as follows: Within each sliding time window, the fused multimodal features are extracted from the thermal, force, and vibration signals and used as the current real-time feature set. Extract fusion features from historical data saved during the model training phase and use them as a reference feature set; The mean and covariance of the current feature set and the reference feature set are calculated separately to characterize their probability distribution. The KL divergence is used to calculate the relative entropy difference between the current feature distribution and the reference feature distribution, which serves as a measure of feature distribution drift.

5. The heat-force-vibration synergistically suppressed centerless grinding machine intelligent monitoring and regulation system according to claim 1, characterized in that, The steps for comparing the fusion pass coefficient with the preset fusion pass coefficient threshold and determining whether the data fusion effect of the current thermal-mechanical-vibration co-model is qualified based on the comparison result are as follows: The fusion pass coefficient is compared with the preset fusion pass coefficient threshold. If the fusion pass coefficient is not less than the preset fusion pass coefficient threshold, it means that the data fusion effect of the current thermo-mechanical-vibration collaborative model is qualified. Then, the centerless grinder is monitored and controlled according to the results predicted by the thermo-mechanical-vibration collaborative model. If the fusion pass coefficient is less than the preset fusion pass coefficient threshold, it means that the data fusion effect of the current thermal-mechanical-vibration collaborative model is not qualified. The monitoring and control operations will no longer rely on the prediction results of the current thermal-mechanical-vibration collaborative model. At this time, the thermal-mechanical-vibration collaborative model can be automatically updated until the fusion pass coefficient is not less than the preset fusion pass coefficient threshold.

Images