A multi-source data-based multi-level maintenance strategy optimization system for numerical control machine tools

Abstract

The application discloses a multi-source data-based multi-level maintenance strategy optimization system for numerical control machine tools, belongs to the technical field of numerical control machine tool maintenance, and comprises a data acquisition and preprocessing unit, a health state quantization unit, a deterioration gradient evaluation unit, a potential risk evaluation unit and a maintenance strategy mapping unit.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of numerical control machine tool maintenance, in particular to a multi-level maintenance strategy optimization system for numerical control machine tools based on multi-source data. BACKGROUND

[0002] As the core equipment of high-end manufacturing industry, the stability and reliability of numerical control machine tools directly affect product quality and production efficiency. The traditional maintenance mode is mainly divided into post-failure maintenance and preventive maintenance based on fixed time period. The former has a lagging response and huge economic losses. The latter does not consider the actual working conditions and individual health status of the machine tool, often leading to insufficient or excessive maintenance. Existing predictive maintenance technologies focus on improving the prediction accuracy of single fault points, but lack quantitative description of the deterioration process of system health status, especially the quantification of the key dynamic indicator of deterioration rate. Maintenance decisions are usually based on rigid thresholds, which cannot match the intensity of maintenance intervention with the gradient of system deviation from the normal state. Moreover, there is a lack of comprehensive optimization of maintenance cost and potential risk. Therefore, how to go beyond single-point fault prediction and build a system that can quantify the deterioration gradient of machine tool health status and continuously optimize multi-level maintenance strategies is a technical problem that needs to be solved in the current field.

[0003] The above information disclosed in the above background section is only used to strengthen the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0004] The purpose of the present application is to provide a multi-level maintenance strategy optimization system for numerical control machine tools based on multi-source data to solve the problems raised in the above background.

[0005] The technical solution of the present application is as follows: a data acquisition and preprocessing unit is used to acquire the running data of the numerical control machine tool in real time and preprocess the running data to extract a quantitative feature set;

[0006] A health status quantification unit is used to perform operation processing based on the quantitative feature set to obtain a system unhealthiness. The health stability index is generated by subtracting the system unhealthiness from the preset ideal health status value;

[0007] A deterioration gradient evaluation unit is used to perform operation processing based on the health stability index to obtain a health status difference value. The health deterioration gradient is generated by dividing the health status difference value by the preset calculation time window.

[0008] The potential risk assessment unit is configured to calculate the system deviation, the deterioration rate value and the bottleneck risk value by combining the health stability index and the health deterioration gradient, and to generate a potential risk level by performing a weighted summation operation on the system deviation, the deterioration rate value and the bottleneck risk value.

[0009] The maintenance strategy mapping unit is configured to compare and analyze the potential risk level with a preset risk threshold sequence to generate a multilevel maintenance strategy instruction.

[0010] The feedback correction unit is configured to record the change of the health stability index caused by the maintenance activity after the execution of the multilevel maintenance strategy instruction, and to iteratively optimize the internal parameters of the system according to the change.

[0011] Preferably, the process of generating the health stability index by the health state quantification unit includes:

[0012] normalizing the feature data in the set of quantified features extracted by the data acquisition and preprocessing unit to obtain normalized feature values, calling preset feature weight coefficients corresponding to the normalized feature values, multiplying each normalized feature value with the corresponding feature weight coefficient to obtain a single item unhealthiness, summing the single item unhealthinesses to obtain a system unhealthiness, and calculating the health stability index by subtracting the system unhealthiness from a preset ideal health state value.

[0013] Preferably, the process of generating the health deterioration gradient by the deterioration gradient assessment unit includes:

[0014] obtaining the health stability index at the current time and the historical health stability index at the previous calculation period, subtracting the historical health stability index from the health stability index at the current time to obtain a health state difference, and dividing the health state difference by a preset calculation time window to obtain the health deterioration gradient.

[0015] Preferably, the system deviation is determined by subtracting the health stability index from a preset ideal health state value.

[0016] Preferably, the process of determining the deterioration rate value includes:

[0017] judging the health deterioration gradient;

[0018] when the health deterioration gradient is a negative value, extracting the negative value as the deterioration rate value;

[0019] when the health deterioration gradient is a non-negative value, setting the deterioration rate value to zero.

[0020] Preferably, the process of determining the bottleneck risk value includes:

[0021] The normalized characteristic values generated by the health state quantification unit and preset characteristic weight coefficients are called, each normalized characteristic value is multiplied by the corresponding characteristic weight coefficient to obtain a single risk contribution value, and the maximum value of the single risk contribution values is extracted as the bottleneck risk value.

[0022] Preferably, the process of generating the multi-level maintenance strategy instruction by the maintenance strategy mapping unit comprises:

[0023] The preset risk threshold sequence and the corresponding multi-level maintenance strategy set are called, the potential risk level is compared with the risk threshold sequence to determine the risk interval, the corresponding maintenance strategy is selected from the multi-level maintenance strategy set according to the determined risk interval, and the multi-level maintenance strategy instruction is generated.

[0024] Preferably, the process of iterative optimization of the feedback correction unit comprises:

[0025] After the maintenance activity is executed, the health stability index after maintenance is collected and compared with the health stability index before maintenance to quantify the maintenance effect, the data record containing the state before maintenance, the maintenance activity and the quantified maintenance effect is taken as a knowledge sample, and the feature weight coefficient in the health state quantification unit or the risk threshold sequence in the maintenance strategy mapping unit is updated by feedback.

[0026] The present application provides a multi-level maintenance strategy optimization system for numerical control machine tools based on multi-source data, which has the following improvements and advantages compared with the prior art:

[0027] 1. The present application quantifies the health level of the machine tool at a certain time by constructing a health state quantification unit and a deterioration gradient evaluation unit, calculates the change rate of the index, and expands the focus of maintenance from isolated and static health point information to continuous and dynamic health evolution process information; The system can distinguish between two completely different risk scenarios: one is that the health index is low but stable, and the other is that the health index is acceptable but is rapidly declining; The prior art can only identify the former, but ignores the sudden failure risk indicated by the latter;

[0028] 2. The application establishes a comprehensive potential risk level evaluation model; the model integrates three key risk dimensions: system deviation degree determined by health stability index, reflecting the overall stock risk of the system; deterioration rate value determined by health deterioration gradient, reflecting the dynamic incremental risk of the system health; and bottleneck risk value determined by the maximum value of each unhealthy degree, aiming to capture the extreme risk of local components; the progress of this design is that it effectively avoids the average effect in traditional risk assessment, ensuring that the sharp degradation of a single component or the rapid decline of overall health trend can be timely and accurately reflected in the final risk level, greatly improving the sensitivity and accuracy of risk identification;

[0029] 3. The accurate matching of maintenance intervention and risk level is achieved: the maintenance strategy mapping unit matches the interval in the preset risk threshold sequence according to the calculated potential risk level to generate multi-level maintenance strategy instructions; the advancement of this mechanism lies in that it abandons the one-size-fits-all maintenance decision based on rigid threshold in the prior art; instead, it adopts a fine-grained hierarchical intervention mode that dynamically matches the risk level, from enhanced monitoring and online compensation to planned maintenance and component replacement, ensuring that the maintenance resources invested are just right for the actual risk, achieving the optimal balance between maintenance cost and equipment reliability;

[0030] 4. The system is endowed with self-learning and adaptive evolution capability: the feedback correction unit of the application constitutes a closed-loop learning circuit of the system; after the execution of multi-level maintenance strategy instructions, the unit forms knowledge samples containing state, activity and effect by quantifying the maintenance effect; these samples are used to continuously iterate and optimize the internal parameters of the system, such as feature weight coefficients or risk threshold sequences; this function enables the system to learn from its own maintenance history, and its model will become more and more consistent with the running characteristics and failure modes of the specific machine tool over time, which is a significant adaptive evolution advantage that the existing static maintenance model does not have. BRIEF DESCRIPTION OF DRAWINGS

[0031] The application will be further explained in conjunction with the drawings and examples:

[0032] Figure 1 is the flow chart of the system of the application. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical scheme and advantages of the application more clear and explicit, the application will be further explained in detail in conjunction with specific examples.

[0034] Example 1:

[0035] Please refer to Figure 1The application provides a multi-source data-based multi-level maintenance strategy optimization system for a numerical control machine tool, comprising: a data acquisition and preprocessing unit for acquiring running data of the numerical control machine tool in real time and preprocessing the running data to extract a quantitative feature set;

[0036] A health state quantization unit is configured to perform operation processing based on the quantitative feature set to obtain a system unhealthiness degree; and subtract the system unhealthiness degree from a preset ideal health state value to generate a health stability index;

[0037] A deterioration gradient evaluation unit is configured to perform operation processing based on the health stability index to obtain a health state difference value; and divide the health state difference value by a preset calculation time window to generate a health deterioration gradient;

[0038] A potential risk evaluation unit is configured to perform operation processing in combination with the health stability index and the health deterioration gradient to determine a system deviation degree, a deterioration rate value and a bottleneck risk value, and perform weighted summation operation in combination with the system deviation degree, the deterioration rate value and the bottleneck risk value to generate a potential risk level;

[0039] A maintenance strategy mapping unit is configured to compare and analyze the potential risk level with a preset risk threshold sequence to generate a multi-level maintenance strategy instruction;

[0040] A feedback correction unit is configured to record changes of the health stability index caused by maintenance activities after execution of the multi-level maintenance strategy instruction, and iteratively optimize system internal parameters according to the changes.

[0041] In the embodiment scenario, a five-axis linkage machining center is taken as a monitoring object, and the system realizes dynamic optimization of the maintenance strategy through cooperative work of internal units; the data acquisition and preprocessing unit of the system continuously acquires running information from key parts of the machine tool, such as spindle vibration spectrum, driving motor current fluctuation and tool soundprint characteristics, and processes them into a quantitative feature set;

[0042] The processing process comprises: for the spindle vibration spectrum, the vibration spectrum can be converted to a frequency domain through fast Fourier transform, and preset fault feature frequencies such as fault frequencies of bearing inner and outer rings, rolling elements and retainer cages, energy integral values in corresponding frequency bands, and time domain root mean square values, kurtosis and margin of the vibration signal are extracted as quantitative features; for the driving motor current, the mean value and standard deviation of the fluctuation signal of the driving motor current can be extracted; for the tool soundprint characteristics, the sound pressure level energy of the tool soundprint characteristics in a specific high frequency band can be extracted; these extracted numerical values with clear physical meanings jointly constitute the quantitative feature set;

[0043] Based on this set, the health state quantification unit converts multi-dimensional data into a single health stability index, and then the deterioration gradient evaluation unit calculates the rate of change of this index to provide dynamic input; the potential risk assessment unit integrates the static health level and the dynamic deterioration trend to generate a comprehensive potential risk level; the repair strategy mapping unit matches the optimal repair intervention measure according to the risk level; the feedback correction unit uses repair effect data to optimize system model parameters in reverse after the repair activity is executed, thereby forming a complete, adaptive closed-loop management and control system; The internal advantage of this architecture is that it upgrades the maintenance decision from isolated fault prediction to continuous management and control of the overall health evolution process of the equipment, achieving global optimization of machine tool operation and maintenance.

[0044] Embodiment 2:

[0045] The process of generating the health stability index by the health state quantification unit includes:

[0046] The feature data in the quantitative feature set extracted by the data acquisition and preprocessing unit is normalized to obtain normalized feature values; the preset feature weight coefficients corresponding to each normalized feature value are called; each normalized feature value is multiplied by the corresponding feature weight coefficient to obtain each single unhealthy degree; the sum of each single unhealthy degree is obtained to obtain the system unhealthy degree; the health stability index is calculated by subtracting the system unhealthy degree from the preset ideal health state value;

[0047] The process of generating the health deterioration gradient by the deterioration gradient evaluation unit includes:

[0048] The health stability index at the current time is obtained, and the historical health stability index of the previous calculation period is obtained; the health stability index at the current time is subtracted from the historical health stability index to obtain a health state difference value; the health state difference value is divided by the preset calculation time window to obtain the health deterioration gradient.

[0049] In order to convert the discrete multi-dimensional performance characteristics of the numerical control machine tool into a health condition that can be uniformly evaluated, the health state quantification unit introduces a health stability index model, and the deterioration gradient evaluation unit introduces a health deterioration gradient model, which together constitute the basis for a complete description of the machine tool state;

[0050] The health stability index The technical motivation of the health stability index

[0051] ;

[0052] Among them, represents the time The health stability index is , 1 represents an ideal health state, and 0 represents complete failure; It is the first Original features Normalized eigenvalues ​​after min-max normalization; is the preset feature weight coefficient corresponding to each normalized eigenvalue. The determination method can use the hierarchical analysis method combined with expert experience, or objectively assign values ​​based on data-driven methods such as principal component analysis, and meet the requirements. ; : refers to the moment; : represents the total number of quantitative features; : Index used for summation, representing the Features

[0053] The ideal health status value is usually preset to 1, indicating no unhealthy state; the HSI value range is [0,1], and the closer the value is to 1, the healthier the state;

[0054] When using the hierarchical analysis method, a three-level hierarchical structure can be constructed, with the machine tool health status as the target layer, the vibration status, electrical status, etc. as the criterion layer, and each specific quantitative feature as the solution layer; multiple field experts are invited to compare the importance of each feature under the same criterion, build a judgment matrix, and calculate the normalized eigenvector corresponding to the maximum eigenvalue of the matrix as the weight of each feature under the criterion. Finally, a consistency test is performed, such as the consistency ratio , to ensure the validity of expert judgment. When principal component analysis is used, the normalized characteristic matrix composed of a large amount of historical normal operation data can be analyzed. Usually, the first principal component explains the largest part of the data variability, and its load vector It reflects the contribution of each original feature to PC1; therefore, the absolute value of each load coefficient of PC1 can be normalized to calculate the weight of each feature, where : represents the transpose of a matrix or vector, and the calculation formula is: ;

[0055] in, : No. The loading coefficient of the original features on the first principal component; : Index used for summation, representing the Features : No. The loading coefficient of the original features on the first principal component; The total number of features; the application effect of this formula is that it integrates complex multi-source data streams into a curve of health state evolution over time, providing a quantitative foundation for subsequent decision-making;

[0056] Health deterioration gradient The model is logically an extension of the model, if is regarded as the position function of the system health state, then is the velocity function of the state change, the technical motivation is to accurately quantify the change rate of the health state and capture its dynamic evolution trend; in discrete-time systems, the form is:

[0057] ;

[0058] where, is the health deterioration gradient at time ; and are the current and previous calculation period health stability indexes respectively; is the preset calculation time window, this parameter is set according to the failure mode of the monitored object, a smaller value is taken for fast failure mode, and a larger value is taken for slow degradation process; the application of this model is a dynamic supplement to static indicators, the combination of the two enables the system to grasp the current health level and deterioration trend of the machine tool, realizing the expansion from static indicators to dynamic evaluation.

[0059] The potential risk assessment unit is used to combine the health stability index and the health deterioration gradient for operation and processing to determine the system deviation degree, deterioration rate value and bottleneck risk value, and to combine the system deviation degree, deterioration rate value and bottleneck risk value for weighted summation operation to generate a potential risk level;

[0060] The system deviation degree is determined by subtracting the health stability index from the preset ideal health state value;

[0061] The determination process of the deterioration rate value includes:

[0062] judging the health deterioration gradient;

[0063] When the health deterioration gradient is negative, the negative value is extracted as the deterioration rate value;

[0064] When the health deterioration gradient is non-negative, set the deterioration rate value to zero;

[0065] The determination process of the bottleneck risk value includes:

[0066] Retrieve the normalized eigenvalues ​​and preset eigenweight coefficients generated by the health status quantification unit; multiply each normalized eigenvalue by the corresponding eigenweight coefficient to obtain each individual risk contribution value; extract the maximum value from each individual risk contribution value as the bottleneck risk value.

[0067] The core function of the potential risk assessment unit is to generate potential risk levels The technical motivation behind the adopted model is to overcome the flaw in traditional risk assessment that the averaging of indicators masks local extreme risks. It deconstructs total risk into a weighted sum of three dimensions: stock risk represented by system deviation, incremental risk represented by the deterioration rate value, and sudden risk dominated by a single weakest link represented by the bottleneck risk value.

[0068] The potential risk level The evaluation model is composed as follows:

[0069] ;

[0070] The structure and parameters of this formula are as follows: The first item is the overall state risk item, the core of which is the system deviation , measuring the gap between the system and the ideal health state; the second item is the overall gradient risk item, the core of which is the deterioration rate value , ensuring that only The risk is only included when it is a negative value; the third item is the single point bottleneck risk item, the core of which is the bottleneck risk value ,pass The operator directly identifies and quantifies the single risk source that currently contributes most to the system's unhealthiness. The parameters in the formula all have a clear source: , , , All inherited from the above modules; is the weight coefficient of the three risks, satisfying , configured according to the focus of the maintenance strategy; is the risk magnification factor, which is greater than 1 and is set according to the risk aversion of the application scenario; The reference deterioration gradient is a significant deterioration rate benchmark value based on historical data statistics or expert definition, which is used to perform dimensionless processing on the deterioration rate value. Its physical dimension is the same as Stay consistent; :time potential risk level; :time Health stability index; : Reference deterioration gradient, a benchmark value used to perform dimensionless processing on the deterioration rate value; :time No. a normalized feature value;

[0071] a risk weight coefficient The weight can be determined by inviting experts in the field to compare the relative importance of the system deviation risk, the deterioration rate risk and the bottleneck risk using the analytic hierarchy process; the value of the risk amplification coefficient can be set according to the machining precision requirements, safety level requirements of different machine tools or workpieces, for example, for aerospace part machining scenes requiring high reliability and low failure risk, a higher risk amplification coefficient value such as 1.5 or 2 can be set;

[0072] An objective determination method based on historical data statistics is: collect all health deterioration gradients of the machine tool in multiple complete life cycles or long-term running processes , filter out all negative gradient values representing health decline, and calculate the absolute values of these negative gradients. Statistically, take the 95th percentile of these absolute values as the reference deterioration gradient , that is ;

[0073] This method ensures that the most severe 5% of the deterioration rate has occurred in history, thereby providing a data-supported and reproducible benchmark for significant deterioration; : the 95th percentile in statistics; : health deterioration gradient;

[0074] In an application example, it is assumed that at a certain time: and ; set the model parameters , and the reference deterioration gradient ; the potential risk assessment unit determines that the system deviation degree is , and the deterioration rate value is ; then perform weighted summation operation:

[0075] ;

[0076] The calculation result is the potential risk level at the current time; this model produces a highly sensitive and informative risk indicator, not only reflecting the existing health status of the machine tool, but also incorporating the consideration of its deterioration speed, thereby enabling earlier and more accurate early warning of potential problems.

[0077] Embodiment 3:

[0078] The process of the maintenance strategy mapping unit generating multi-level maintenance strategy instructions includes:

[0079] Retrieve a preset risk threshold sequence and its corresponding multi-level maintenance strategy set; compare the potential risk level with the risk threshold sequence to determine the risk interval to which it belongs; and select the corresponding maintenance strategy from the multi-level maintenance strategy set based on the determined risk interval to generate a multi-level maintenance strategy instruction.

[0080] The maintenance strategy mapping unit is designed to convert the quantified risk into specific operational instructions; the core operating logic is to convert the calculated potential risk level Value, and the preset risk threshold sequence The method for determining the risk threshold sequence does not rely on subjective experience, but is based on statistical modeling of historical data and finding the optimal split point through cost-benefit analysis;

[0081] This cost-benefit analysis can be performed by constructing the total expected cost function This function takes into account both the preventive maintenance cost and the failure loss cost. For example, for a two-level strategy L0-no maintenance, L1-maintenance, the total expected cost can be defined as:

[0082] ;

[0083] in, is the risk threshold to be optimized; is the average cost of performing L1 repairs; is the average loss caused by an unexpected failure; It is the probability density function of the potential risk level obtained through historical data statistics; is the conditional probability of a machine tool failure in the next cycle under a specific PRL value. This function can be fitted by historical failure data and the corresponding PRL value through methods such as logistic regression. Get the minimum value The optimal split point can be determined by ;For multi-level thresholds, this optimization can be performed level by level; :by is the total expected cost function of the threshold; : potential risk level; : Average loss caused by unexpected failure; A differential or infinitesimal increment representing the potential level of risk;

[0084] The analysis process comprehensively considers various levels of maintenance strategies, such as - Enhanced data monitoring frequency, -Online parameter compensation, - Cost of executing planned downtime inspections and the probability of major failure and the corresponding loss cost under different risk levels; taking the above example, assume that the maintenance strategy threshold is set as ; since the calculated potential risk level satisfies the condition of , the system determines that the risk belongs to the first interval and selects and generates the level maintenance strategy instruction from the multi-level maintenance strategy set; the technical effect of the mapping mechanism lies in the on-demand allocation and intensity matching of maintenance intervention, ensuring that the maintenance resources are dynamically and accurately adapted to the actual risk level, thereby achieving lean management of maintenance activities.

[0085] Embodiment 4:

[0086] The process of iterative optimization of the feedback correction unit includes:

[0087] After the execution of the maintenance activity, the health stability index after maintenance is collected and compared with the health stability index before maintenance to quantify the maintenance effect; the data record containing the state before maintenance, the maintenance activity and the quantified maintenance effect is taken as a knowledge sample; the feature weight coefficient in the health state quantification unit or the risk threshold sequence in the maintenance strategy mapping unit is updated by feedback using the knowledge sample.

[0088] For the update of the feature weight coefficient , a gradient-based optimization method can be used. When a maintenance activity, such as replacement of component i, is confirmed to be effective, i.e., the health stability index after maintenance is significantly higher than that before maintenance, the system can regard this as a positive feedback. The feature weight related to this component can be incrementally adjusted, for example, using the following update rule:

[0089] ;

[0090] where is a small learning rate, and other weights are proportionally reduced to ensure that the sum of all weights is still 1; for the update of the risk threshold sequence , after a certain number of new knowledge samples, such as 100, are accumulated, these samples are supplemented to the historical database, and the cost-benefit analysis model in Embodiment 4 is re-executed to calculate the optimal threshold that better fits the current working condition; : the new feature weight after update; : the old feature weight before adjustment; : the health stability index after execution of the maintenance activity; : the health stability index before execution of the maintenance activity;

[0091] The feedback correction unit is the core component of the adaptive optimization system. When a multi-level maintenance strategy instruction is executed, for example, the main shaft bearing is replaced during a planned downtime, the function of the unit is activated. The health stability index of the machine tool after the maintenance activity is collected and recorded, and compared with the index before the maintenance, so as to quantify the actual effect of the maintenance activity, for example, it is recorded that before the maintenance For , the maintenance after the recovery to ; A complete data record containing the state before maintenance, maintenance activity content and quantitative maintenance effect will be structured and stored as a new knowledge sample; The system uses the accumulated knowledge samples to update and optimize the key internal parameters in the upstream model, including updating the weight coefficients of each feature in the health state quantification unit in reverse , or correcting the risk threshold sequence in the maintenance strategy mapping unit ; The technical effect of the closed-loop feedback correction mechanism is to give the entire optimization system the ability of self-learning and self-adaptation, so that it can learn from each maintenance practice, and the model parameters will become more suitable for the specific machine tool and the specific working condition over time, and finally realize the continuous and automatic improvement of the prediction accuracy and decision level.

[0092] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A multi-level maintenance strategy optimization system for CNC machine tools based on multi-source data, characterized in that: include: A data acquisition and preprocessing unit is used to obtain the operating data of the CNC machine tool in real time and preprocess the operating data to extract a quantitative feature set; A health status quantification unit is used to perform calculations based on the quantitative feature set to obtain a system unhealthiness; the health stability index is generated by subtracting the system unhealthiness from a preset ideal health status value; A deterioration gradient evaluation unit is used to perform calculations based on the health stability index to obtain a health status difference; Dividing the health state difference by a preset calculation time window to generate a health deterioration gradient; The potential risk assessment unit is used to combine the health stability index and the health deterioration gradient to perform calculations to determine the system deviation, deterioration rate value, and bottleneck risk value, and to perform a weighted summation operation on the system deviation, deterioration rate value, and bottleneck risk value to generate a potential risk level; A maintenance strategy mapping unit is used to compare and analyze the potential risk level with a preset risk threshold sequence to generate a multi-level maintenance strategy instruction; A feedback correction unit is used to record the changes in the health stability index caused by maintenance activities after the multi-level maintenance strategy instructions are executed, and iteratively optimize the internal parameters of the system based on the changes; The system deviation is determined by subtracting the health stability index from the preset ideal health state value; The process of determining the deterioration rate value includes: Determine the gradient of health deterioration; When the health deterioration gradient is negative, the negative value is extracted as the deterioration rate value; When the health deterioration gradient is non-negative, the deterioration rate is set to zero; The process of determining the bottleneck risk value includes: Retrieve the normalized eigenvalues ​​and preset eigenweight coefficients generated by the health status quantification unit; multiply each normalized eigenvalue by the corresponding eigenweight coefficient to obtain each individual risk contribution value; extract the maximum value from each individual risk contribution value as the bottleneck risk value.

2. The multi-level maintenance strategy optimization system for CNC machine tools based on multi-source data according to claim 1 is characterized in that: The process of generating the health stability index by the health status quantification unit includes: Normalizing the feature data in the quantitative feature set extracted by the data acquisition and preprocessing unit to obtain normalized feature values; retrieving the preset feature weight coefficients corresponding to each normalized feature value; multiplying each normalized feature value by the corresponding feature weight coefficient to obtain each individual unhealthy degree; summing each individual unhealthy degree to obtain the system unhealthy degree; and calculating the health stability index by subtracting the system unhealthy degree from the preset ideal health state value.

3. The multi-level maintenance strategy optimization system for CNC machine tools based on multi-source data according to claim 1 is characterized in that: The process of generating the health deterioration gradient by the deterioration gradient evaluation unit includes: Obtain the health stability index at the current moment and the historical health stability index of the previous calculation cycle; subtract the historical health stability index from the health stability index at the current moment to obtain the health status difference; divide the health status difference by the preset calculation time window to obtain the health deterioration gradient.

4. The multi-level maintenance strategy optimization system for CNC machine tools based on multi-source data according to claim 1 is characterized in that: The process of the maintenance strategy mapping unit generating a multi-level maintenance strategy instruction includes: Retrieve a preset risk threshold sequence and its corresponding multi-level maintenance strategy set; compare the potential risk level with the risk threshold sequence to determine the risk interval to which it belongs; and select the corresponding maintenance strategy from the multi-level maintenance strategy set based on the determined risk interval to generate a multi-level maintenance strategy instruction.

5. The multi-level maintenance strategy optimization system for CNC machine tools based on multi-source data according to claim 1 is characterized in that: The process of iterative optimization performed by the feedback correction unit includes: After the maintenance activity is performed, the health and stability index after maintenance is collected and compared with the health and stability index before maintenance to quantify the maintenance effect; the data records containing the pre-maintenance status, maintenance activities and quantified maintenance effects are used as knowledge samples; and the knowledge samples are used to feedback and update the feature weight coefficients in the health status quantification unit or the risk threshold sequence in the maintenance strategy mapping unit.

Images