System for predicting running state of automatic metal processing machine tool

Abstract

A metal automatic processing machine tool operation state prediction system of the present invention belongs to the technical field of automatic processing, and comprises a data acquisition module used for acquiring multi-scale physical parameters representing the machine tool operation state in real time, the multi-scale physical parameters comprise a structure thermal gradient tensor, a kurtosis change rate of a tool interface signal, a relative fluctuation value of a cooling liquid dielectric constant and an environment low-frequency seismic noise spectrum; the dynamic instability risk coupling module is used for determining a harmonic locking risk index based on the multi-scale physical parameters acquired by the data acquisition module; and the risk level evaluation module is used for determining a current system operation risk level based on the harmonic locking risk index determined by the dynamic instability risk coupling module and generating a normalized risk evaluation index.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automated processing, in particular to a metal automated processing machine tool running state prediction system. BACKGROUND

[0002] With the continuous development of automated processing technology, the stability control and safety protection of metal machine tool processing process are facing increasing challenges. This complexity brings many problems, especially in the management and monitoring of machine tool running state.

[0003] Currently, the running state of the machine tool is generally checked by technical personnel regularly, and detailed information of each inspection and maintenance is recorded. However, the traditional passive monitoring and maintenance method relies on manual judgment and physical measurement, and requires technical personnel to check and record in person. Although these methods can provide information about the running state of the machine tool, they are usually inefficient and may not be able to find problems in real time.

[0004] Therefore, how to timely discover and predict potential problems in machine tool operation has become a problem that needs to be solved in the field. SUMMARY

[0005] The purpose of the present application is to provide a metal automated processing machine tool running state prediction system to solve the problems raised in the background.

[0006] The technical solution of the present application is as follows: The data acquisition module is used to acquire multi-scale physical parameters representing the running state of the machine tool in real time, including structural thermal gradient tensor, kurtosis change rate of tool interface signal, relative fluctuation value of cooling liquid dielectric constant and environmental low-frequency seismic noise spectrum. The dynamic instability risk coupling module is used to determine the harmonic locking risk index based on the multi-scale physical parameters acquired by the data acquisition module. The risk level assessment module is used to determine the current system running risk level based on the harmonic locking risk index determined by the dynamic instability risk coupling module, and generate a normalized risk assessment index. The closed-loop active control module is used to predict the risk evolution gradient based on the historical multi-scale physical parameters and the risk assessment index generated by the risk level assessment module, and perform hierarchical intervention to adjust the machine tool running parameters according to the current system running risk level and the risk evolution gradient.

[0007] Preferably, the data acquisition module is specifically used for: Acquiring the structural thermal gradient tensor through the thermocouple array arranged on the key structure of the machine tool; Collecting the tool interface signal through the acoustic emission sensor installed on the spindle box and calculating the kurtosis change rate; The dielectric constant of the coolant is monitored by a dielectric sensor arranged in the coolant circulation pipeline, and a relative fluctuation value is determined; The low-frequency seismic noise spectrum is acquired by an environmental vibration sensor arranged on the machine tool base.

[0008] Preferably, the dynamic instability risk coupling module determines a harmonic lock risk index, comprising: Based on the structural thermal gradient tensor and the preset initial modal frequency, the system modal frequency after drift is calculated; Based on the kurtosis change rate and the relative fluctuation value, the cutting process excitation intensity factor is calculated; The harmonic lock risk index is determined by combining the system modal frequency after drift, the cutting process excitation intensity factor, and the low-frequency seismic noise spectrum.

[0009] Preferably, the system modal frequency after drift is calculated, comprising: The dynamic stiffness weakening effect caused by the structural thermal gradient tensor is quantified by using a preset thermal softening coefficient, and the system modal frequency after drift is determined as a representation.

[0010] Preferably, the cutting process excitation intensity factor is calculated, comprising: The preset kurtosis change rate critical value and the dielectric constant relative fluctuation critical threshold value are used to normalize the real-time acquired kurtosis change rate and relative fluctuation value, respectively; According to the preset weight coefficient, the two normalized values are weighted and summed to generate the cutting process excitation intensity factor.

[0011] Preferably, the risk level evaluation module determines the current system operation risk level, comprising: The harmonic lock risk index is compared with the preset risk index critical threshold value to generate a normalized risk assessment index; When the risk assessment index is lower than the first preset threshold value, the system operation risk level is determined as the safety level; When the risk assessment index is between the first preset threshold value and the second preset threshold value, the system operation risk level is determined as the first-level warning level; When the risk assessment index is higher than the second preset threshold value, the system operation risk level is determined as the second-level warning level.

[0012] Preferably, the closed-loop active control module predicts the risk evolution gradient, comprising: A pre-trained recurrent neural network model is used to input a multi-dimensional time series feature data set containing historical multi-scale physical parameters and risk assessment indexes; The recurrent neural network model outputs a risk index prediction value after a future preset time step; Based on the risk index prediction value and the current risk assessment index, the risk evolution gradient is calculated.

[0013] Preferably, the closed-loop active control module performs hierarchical intervention, including: When the system operation risk level is a first warning level and the risk evolution gradient exceeds a preset gradient threshold, a first intervention is triggered. The first intervention includes: generating a modulation signal based on the risk assessment index and the risk evolution gradient; and superimposing the modulation signal on the spindle command speed to start spindle speed fine tuning.

[0014] Preferably, the closed-loop active control module performs hierarchical intervention, further including: When the system operation risk level is a second warning level, a second intervention is triggered. The second intervention includes: forcibly reducing the spindle speed and the feed rate to a preset safe idle speed state; and automatically executing an emergency stop when the risk assessment index fails to fall within a preset time.

[0015] The present application provides a metal automatic machining tool running state prediction system, which has the following improvements and advantages compared with the prior art: 1. The machine tool running state is comprehensively and deeply perceived, which greatly improves the early instability precursor capture ability; the prior art usually relies on a single macro-vibration signal for state monitoring, and the perception dimension is single, and it is not sensitive to the microscopic deterioration process; the present scheme acquires a multi-scale physical parameter acquisition system executed by a data acquisition module, which cooperatively monitors from four physical dimensions: the structural thermal gradient tensor quantifies the influence of thermal deformation on the stiffness of the whole machine from the macro-structure layer; the kurtosis change rate of the tool interface signal captures the transient characteristics of tool damage from the micro-interface layer; the relative fluctuation value of the cooling liquid dielectric constant indirectly reflects the deterioration of the lubrication and chip removal state in the cutting area from the chemical auxiliary layer; and the environmental low-frequency seismic noise spectrum evaluates the potential threat of random excitation from the external environment layer; this multi-physical field fusion perception constructs a complete portrait of the machine tool state, which can identify the instability precursor in the embryonic stage that cannot be found by traditional single signal method, significantly enhancing the sensitivity and reliability of the system perception. 2. A set of dynamic instability risk coupling evaluation model with clear physical meaning and high precision is established, which overcomes the limitations of static model or simple data weighting in the prior art; the dynamic instability risk coupling module of the present scheme is not a simple linear superposition of collected data, but a deep coupling: firstly, it calculates the system modal frequency after drift based on the structural thermal gradient tensor, solves the model misalignment problem caused by regarding the machine tool modal frequency as a static constant in the prior art, and provides a dynamic and accurate resonance benchmark for risk assessment; secondly, it fuses the kurtosis change rate and the relative fluctuation value of the coolant dielectric constant into a cutting process excitation strength factor, which quantifies the strength of internal excitation source uniformly; by combining the dynamic system modal frequency, the internal excitation strength factor and the external environmental noise spectrum, the harmonic locking risk index is determined; the index deeply reveals the physical nature of the mutual coupling between thermal-induced stiffness weakening, internal excitation enhancement and external excitation triggering, which leads to brittle instability, and the evaluation result is more accurate and has more predictability than the traditional threshold alarm method; 3. The introduction of the forward-looking prediction of the risk evolution trend changes the control system from passive response to predictive intervention; the closed-loop active control module of the present scheme not only evaluates the current risk, but more importantly, it can predict the risk evolution gradient; through a recurrent neural network model pre-trained based on a historical multi-dimensional time series feature data set, the system can predict the future risk index and calculate the change rate of the current risk accordingly; this risk evolution gradient provides a forward-looking in the time dimension for the control decision, enabling the system to identify those dangerous working conditions with low current risk but extremely fast deterioration rate, so that intervention can be carried out before the risk reaches the peak, realizing the change from passive response to active prevention in the technical concept; 4. A set of hierarchical intervention strategies dynamically matched with the risk level and evolution gradient is designed, realizing the best balance between processing efficiency and operation safety; according to the evaluation and prediction results, the closed-loop active control module of the present scheme performs fine hierarchical intervention: when the system is in the first warning level and the risk is rapidly increasing, the first-level intervention is started, and the spindle command speed is adjusted to actively break the harmonic locking condition in a way that has minimal impact on processing; when the system enters the second warning level, the second-level intervention is triggered, and protective measures such as forced reduction of spindle speed and feed rate or even automatic emergency shutdown are taken; this rigid-flexible control logic avoids the drawbacks of traditional methods, such as insufficient intervention leading to accidents or excessive intervention such as shutting down the machine, which affects production efficiency, and realizes intelligent closed-loop control that maximizes the maintenance of processing continuity and efficiency while ensuring absolute safety. BRIEF DESCRIPTION OF DRAWINGS

[0016] The present application will be further explained in conjunction with the accompanying drawings and examples: Figure 1 is the flow chart of the system of the present application. DETAILED DESCRIPTION

[0017] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to specific embodiments.

[0018] Embodiment 1 Please refer to Figure 1 , the present application provides a metal automatic machining tool running state prediction system, comprising: A data acquisition module is configured to acquire multi-scale physical parameters representing the running state of the tool in real time, wherein the multi-scale physical parameters include a structural thermal gradient tensor, a kurtosis change rate of a tool interface signal, a relative fluctuation value of a coolant dielectric constant, and an environmental low-frequency seismic noise spectrum. A dynamic instability risk coupling module is configured to determine a harmonic locking risk index based on the multi-scale physical parameters acquired by the data acquisition module. A risk level assessment module is configured to determine a current system running risk level based on the harmonic locking risk index determined by the dynamic instability risk coupling module, and generate a normalized risk assessment index. A closed-loop active control module is configured to predict a risk evolution gradient based on historical multi-scale physical parameters and the risk assessment index generated by the risk level assessment module, and perform a hierarchical intervention to adjust the tool running parameters according to the current system running risk level and the risk evolution gradient. A metal automatic machining tool running state prediction system, the technical purpose is to establish an intelligent protection system that can change from passive response to predictive intervention, through deep coupling analysis and evolution trend prediction of multi-scale physical parameters, actively adjusting the tool running parameters before catastrophic chatter occurs, thereby greatly improving the stability, precision and safety of the machining process; in this embodiment, the system is composed of four core modules, forming a complete technical link of information perception, risk assessment, trend prediction and closed-loop control; The data acquisition module is configured to capture underlying physical information representing the running state of the tool in real time and comprehensively; the module acquires multi-scale physical parameters in four dimensions through a heterogeneous sensor network deployed at different key positions of the tool; these parameters include a structural thermal gradient tensor representing the thermal deformation state of the whole machine , a kurtosis change rate of a tool interface signal representing the micro-damage state of the tool interface , a relative fluctuation value of a coolant dielectric constant representing the performance change of the cooling and lubricating system , and an environmental low-frequency seismic noise spectrum representing external environmental excitation ; these four parameters jointly construct a complete portrait of the tool running state from four scales of macro-structure, micro-interface, chemical assistance and external environment. The dynamic instability risk coupling module has the core function of fusing the above-mentioned multi-scale and heterogeneous physical parameters into a single index capable of accurately quantifying the dynamic instability risk through a mathematical model with clear physical meaning; the module receives real-time multi-scale physical parameters provided by the data acquisition module, and calculates and finally determines the core risk index, the harmonic locking risk index, based on these parameters The innovation of the index lies in that it is not simply data weighting, but deeply couples the thermal stiffness softening effect, the micro failure excitation enhancement effect and the external noise triggering effect, so as to accurately capture the risk of brittle mutation caused by the positive feedback cycle of multiple factors; The risk level assessment module has the function of converting the continuously changing risk index into a discrete, clear and directly decision-making risk level for the control system; the module receives the harmonic locking risk index determined by the dynamic instability risk coupling module , compares it with the preset critical threshold, and generates a normalized risk assessment index ; according to different numerical intervals of the normalized index, the current system operation risk level is determined as a safety level, a first warning level or a second warning level; this process provides clear and objective decision-making basis for subsequent intervention measures of what strength; The closed-loop active control module has the function of executing the final preventive intervention action and is the execution terminal for realizing active safety of the system; the work of the module is not a simple passive response, but a dual consideration of the current state and future trend; the risk evolution gradient is predicted by using a pre-trained prediction model based on historical multi-scale physical parameters and risk assessment indexes ; then, the module executes a carefully designed hierarchical intervention strategy according to the current system operation risk level and the risk evolution gradient, and actively breaks the instability condition to be formed by fine-tuning or forcibly adjusting the key machine tool operation parameters such as spindle speed and feed rate, so as to suppress the chatter at the budding stage; The above-mentioned four modules work together to build a complete closed loop from multi-dimensional perception, coupling analysis, level assessment to predictive control; overcome the limitations of the prior art which only relies on a single signal or passive response, realize accurate and forward-looking management of the machine tool dynamic instability risk by deeply coupling multi-scale physical mechanisms and introducing the prediction of the risk evolution gradient; the technical effects achieved are: without affecting the normal processing efficiency, significantly reducing the probability of sudden chatter, protecting the surface quality and dimensional accuracy of the processed parts, and effectively avoiding damage to the tool and key components of the machine tool caused by severe vibration, realizing intelligent active safety protection of the metal automatic processing process.

[0019] The data acquisition module is specifically used for: The thermal gradient tensor of the structure is acquired through a thermocouple array arranged on a key structure of the machine tool; The tool interface signal is collected through an acoustic emission sensor mounted on the spindle box, and the kurtosis change rate is calculated; The dielectric constant of the cooling liquid is monitored through a dielectric sensor deployed on the cooling liquid circulating pipeline, and the relative fluctuation value is determined; The low-frequency seismic noise spectrum is collected through an environmental vibration sensor arranged on the machine tool base; The embodiment is a specific implementation of the above-mentioned data acquisition module; in order to ensure the accuracy and representativeness of the collected data, a specific sensor deployment scheme is adopted; The data acquisition module is specifically used for: The thermal gradient tensor of the structure is acquired through a thermocouple array arranged on a key structure of the machine tool; the thermal gradient tensor of the structure is used to reflect the unevenness of the overall temperature distribution of the machine tool; in this embodiment, the thermocouple array is arranged on the spindle box, the column, the bed body and other key paths of machining heat generation and transmission; by collecting the temperature values of these discrete points in real time, a continuous temperature field is reconstructed by using an interpolation algorithm, and then a temperature gradient tensor is calculated ; the purpose of this is to accurately capture the dynamic stiffness change of the whole machine caused by thermal deformation, which is the basis for subsequent frequency drift calculation; Preferably, the interpolation algorithm is radial basis function interpolation or Kriging interpolation; RBF interpolation is suitable for temperature field reconstruction of complex structure surface due to its high efficiency on non-uniformly distributed data points; Kriging interpolation can provide the best linear unbiased estimation based on spatial autocorrelation and can give the error variance of the interpolation result, thereby quantifying the uncertainty of reconstruction; The tool interface signal is collected through an acoustic emission sensor mounted on the spindle box, and the kurtosis change rate is calculated; the kurtosis change rate is used to represent the index of the impact or sharpness of the signal changing with time; the acoustic emission sensor is installed near the spindle bearing due to its high sensitivity to high-frequency transient events such as micro-cracks in the material, so as to minimize signal attenuation; the signal processing unit in the module calculates the kurtosis value of the collected acoustic emission signal in real time, and performs time differentiation on the kurtosis value, to obtain the kurtosis change rate ; the purpose of this is to capture the deterioration precursors such as tool wear and edge collapse at the micro level in advance when macro vibration is not obvious; The calculation of the kurtosis change rate uses central difference method: ; wherein, is the kurtosis value at the current time, is the kurtosis value at the last time step, is the sampling time interval; in order to smooth the data and reduce the influence of noise, the kurtosis value sequence can be subjected to a moving average filter before derivation; The dielectric constant of the cooling liquid is monitored by a dielectric sensor deployed in the cooling liquid circulation pipeline, and a relative fluctuation value is determined; the relative fluctuation value of the dielectric constant of the cooling liquid is used to represent the degree of deviation of the electrical characteristics of the cooling liquid from the normal state; the dielectric sensor is installed in the cooling liquid return pipeline to monitor the dielectric constant in real time; when the tool or workpiece wears out, the concentration of metal chips suspended in the cooling liquid increases, causing the dielectric constant to change; the module calculates the relative fluctuation value by comparing the real-time value with the preset reference value of the clean cooling liquid ; the purpose of this is to indirectly monitor the deterioration of the processing state from the perspective of electrochemistry; Low-frequency seismic noise spectrum is collected by an environmental vibration sensor arranged on the machine tool base; the low-frequency seismic noise spectrum is used to represent the energy distribution of the vibration excitation from the external environment such as the workshop floor and the operation of adjacent equipment in the frequency domain; a high-sensitivity accelerometer or geophone is directly installed on the machine tool base which is isolated from the ground, focusing on collecting environmental vibration signals in the low-frequency band, usually 1-100Hz; the module converts the time-domain signal into a low-frequency seismic noise spectrum by fast Fourier transform ; the purpose of this is to quantify the external energy input that may resonate with the machine tool modal frequency, which is an important trigger factor for harmonic locking; The system aims to establish an efficient and practical predictive protection system through in-depth coupling analysis of four key multi-scale physical parameters. Although the complexity of the machining process involves many factors, the system mainly focuses on the few core parameters that have the most significant impact on chatter risk to balance between model accuracy and real-time computing capability; Through the above specific and targeted sensor deployment and signal processing method, the data acquisition module of the embodiment can ensure the real-time, accuracy and explicitness of the physical meaning of the four core physical parameters obtained; compared with traditional methods that rely on single or macro vibration signals, this multi-scale, multi-physical field sensing scheme provides high-quality, high-dimensional input information for subsequent precise risk coupling modeling, greatly improving the perceptual sensitivity and reliability of the system to early unstable states.

[0020] Embodiment 2 The dynamic instability risk coupling module determines the harmonic locking risk index, including: Based on the structural thermal gradient tensor and the preset initial modal frequency, the system modal frequency after drift is calculated; Based on the kurtosis change rate and the relative fluctuation value, the cutting process excitation intensity factor is calculated; The harmonic locking risk index is determined in combination with the system modal frequency after drift, the cutting process excitation intensity factor and the low-frequency seismic noise spectrum; The system modal frequency after drift is calculated, including: The dynamic stiffness weakening effect caused by the structural thermal gradient tensor is quantified by using a preset thermal softening coefficient, and the system modal frequency after drift is determined as a representation; The cutting process excitation intensity factor is calculated, including: The kurtosis change rate and the relative fluctuation value are normalized by using preset kurtosis change rate critical value and dielectric constant relative fluctuation critical threshold; The two normalized values are weighted and summed according to the preset weight coefficient to generate the cutting process excitation intensity factor; The embodiment is a specific implementation of the dynamic instability risk coupling module of embodiment 1 for determining the harmonic locking risk index , which integrates the above technical features and exhibits a logical progression and clear physical meaning modeling process; the purpose of the process is to unify multiple physical quantities that seem unrelated into a single index that can accurately predict the risk of chatter; The process of the module for determining the harmonic locking risk index includes the following core steps: The system modal frequency after drift is calculated based on the structural thermal gradient tensor and the preset initial modal frequency The purpose of this step is to quantify the influence of thermal stiffness softening effect on the system resonance benchmark; the inherent property of the machine tool, the modal frequency, is not constant, but changes dynamically with the heat accumulation during the machining process; in this embodiment, the system modal frequency after drift is calculated as follows: To quantify this physical process, a core thermal modal frequency drift formula is introduced: ; Wherein: : the system modal frequency after drift, the physical dimension is hertz, the source is calculated by the formula, which is dynamically changed and represents the frequency of the system resonance point under the current thermal state; : the initial modal frequency, the physical dimension is hertz, the source is pre-calibrated by experiment, which refers to the benchmark modal frequency of the machine tool in cold state or thermal equilibrium state, and can be obtained by modal testing means such as hammering method or sweep analysis, which is the inherent property of the machine tool; : the Frobenius norm of the thermal gradient tensor, the physical dimension is Kelvin / m, the source is calculated in real time by the data acquisition module according to the sensor array data, which represents the overall tensor The quantification of the temperature inhomogeneity degree contained is a scalar value; : thermal softening coefficient, physical dimension of meter / Kelvin, source is pre-calibrated by experiment; the thermal softening coefficient refers to a preset parameter representing the sensitivity of the dynamic stiffness of the machine tool to the change of thermal gradient; the calibration method is as follows: in the calibration experiment, a data point set composed of thermal gradient norm measurement values under different working conditions and corresponding modal frequency measurement values is collected, and then the coefficient is fitted by a regression analysis method such as least square method to best approximate the function relationship described in the above formula; The formula is an engineering approximation used to capture the main influence trend of the thermal gradient on the modal frequency; The data point set should at least contain three different working conditions: cold state, pre-starting, stable thermal equilibrium state, long-time empty running and typical machining working conditions such as rough machining and finish machining; under each working condition, the Frobenius norm of the structure thermal gradient tensor and the system modal frequency measured by the sweep frequency method or the hammering method should be measured at the same time; the size of the data point set should be not less than 20 groups to ensure the statistical significance of the regression result; Technical motivation and fusion logic: this step quantifies the weakening effect of the dynamic stiffness of the whole machine caused by the structure thermal gradient tensor by using the preset thermal softening coefficient , and determines the system modal frequency after the drift is represented; this solves the model misalignment problem in the prior art caused by regarding the modal frequency as a static value, provides a dynamic and accurate resonance benchmark for risk assessment, and is a physical prerequisite for accurate prediction; Based on the kurtosis change rate and the relative fluctuation value, the cutting process excitation intensity factor is calculated The purpose of this step is to comprehensively evaluate the internal excitation intensity caused by the state deterioration on the micro level in the cutting process; system instability not only needs frequency proximity, but also needs sufficient energy injection; in this embodiment, the cutting process excitation intensity factor is calculated as follows: Therefore, the empirical model definition formula of the cutting process excitation intensity factor is introduced: ; Wherein: : cutting process excitation intensity factor, dimensionless, source is calculated by the formula, which comprehensively reflects the intensity of internal excitation; : kurtosis change rate, physical dimension of -1 , source is calculated in real time by the data acquisition module; : relative fluctuation of dielectric constant, dimensionless, derived from real-time monitoring and calculation by data acquisition module; : critical value of kurtosis change rate, physical dimension is -1 , derived from presetting through historical data statistics; the critical value of kurtosis change rate refers to the statistical threshold observed in a large number of historical experimental data, which indicates that the cutting state is about to change from stable to critical instability; : critical threshold of relative fluctuation of dielectric constant, dimensionless, derived from presetting through historical data statistics; similar to , it is a critical change threshold indicating state deterioration; : weight coefficient, dimensionless, derived from presetting through machine learning model training; the weight coefficient refers to the coefficient for balancing the importance of two different source precursor signals; the determination method is: using a historical data set containing chatter and stable cutting states, training through classification algorithms such as logistic regression to find the combination of and that can achieve the best classification effect; This formula is an empirical model that aims to approximately integrate two different signals through weighted summation; The historical data set used for training should contain at least 50 groups of typical chatter precursor data before chatter occurs and 100 groups of stable normal cutting data. Each group of data should contain time series of kurtosis change rate and relative fluctuation value of dielectric constant ; the model training goal is to minimize the cross-entropy loss function, and the F1 score is used as the main evaluation indicator to ensure that the model can effectively identify chatter precursors when dealing with unbalanced data sets, with fewer chatter samples than normal samples; Technical motivation and fusion logic: this step uses the preset critical value of kurtosis change rate and the relative fluctuation critical threshold of dielectric constant to normalize the real-time acquired kurtosis change rate and relative fluctuation value ; according to the preset weight coefficient , the two normalized values are weighted and summed to generate the cutting process excitation intensity factor ; the innovation of this step is to integrate two different physical sources of microscopic precursor signals indicating instability into a unified, dimensionless excitation intensity indicator, solving the problem of difficult comprehensive evaluation of internal excitation sources; Combined with the system modal frequency after drift, the cutting process excitation intensity factor and the low-frequency seismic noise spectrum, the harmonic locking risk index is determined​​ The small quantity introduced ensures that the denominator is not zero even when the main shaft rotational speed harmonic frequency is exactly equal to the system modal frequency, thus maintaining numerical stability of the calculation; it ensures that in the extreme case where the frequency is extremely close to the resonance point, the risk index can rise sharply but smoothly, in line with physical intuition; This step is the final risk coupling step, which aims to combine the likelihood of system resonance, the frequency proximity, the strength of internal excitation and the triggering effect of external excitation to form the final risk index; Harmonic lock risk index is defined as: ; Wherein: : mth harmonic lock risk index, dimensionless, derived from the calculation by this formula; : cutting process excitation strength factor, dimensionless, derived from the calculation by the previous step 2; : normalized ambient noise energy spectrum, dimensionless, derived from the data acquisition module and processed, specifically the noise energy value at the current shifted system modal frequency , reflecting the coupling degree of external excitation and system resonance point; by dividing the ambient low-frequency seismic noise spectrum by its maximum peak or total energy, the normalized ambient noise energy spectrum is obtained; : harmonic order, an integer, representing an integer multiple of the main shaft rotational speed, derived from a preset range, for example , to focus on the low-order main shaft rotational speed harmonics that have the greatest impact on chatter; : main shaft rotational speed, physical dimension of hertz, derived from the real-time acquisition by the machine tool numerical control system; : shifted system modal frequency, physical dimension of hertz, derived from the calculation by the previous step 1; : regularization small quantity, dimensionless, derived from a preset small positive number, such as ; the role is to avoid the denominator being zero when is exactly equal to , ensuring numerical stability of the calculation; The value of the regularization small quantity should be much smaller than the frequency mistuning degree , which is the minimum value in the normal working range, and the recommended value range is to ; a too large The denominator will weaken the sensitivity of the denominator to the frequency close to the resonance, resulting in inaccurate calculation of the risk index; too small Then there may be a floating-point precision problem; Fusion logic: the physical meaning of the formula is the ratio of the effective excitation energy to the frequency mismatch; the numerator is the product of the internal excitation And external excitation , representing the total excitation strength, here, the normalized environmental noise energy spectrum Quantifies the excitation effect of the external environment on the system resonance point; the denominator Represents the normalized difference between the m-th harmonic frequency of the main shaft speed and the real modal frequency of the system, and the smaller the value, the closer the system is to the resonance condition; the total harmonic lock risk index Take the maximum value of the risk in all concerned orders, that is: ; Through the close coupling of the above three steps, the dynamic instability risk coupling module of the embodiment successfully integrates multiple cross-scale physical quantities such as thermal gradient, micro-damage, coolant state, environmental noise and main shaft speed into a single index that can accurately and sensitively represent the catastrophic chatter risk ; The model not only has clear physical meaning, but also can capture the complex nonlinear positive feedback relationship between various factors, and its prediction accuracy and predictability have been improved compared with the existing technology based on single threshold or linear model.

[0021] Embodiment 3 The risk level assessment module determines the current system operation risk level, including: Compare the harmonic lock risk index with the preset risk index critical threshold to generate a normalized risk assessment index; When the risk assessment index is lower than the first preset threshold, the system operation risk level is determined to be the safety level; When the risk assessment index is between the first preset threshold and the second preset threshold, the system operation risk level is determined to be the first warning level; When the risk assessment index is higher than the second preset threshold, the system operation risk level is determined to be the second warning level; This embodiment is a specific implementation of the risk level assessment module of embodiment 1 to determine the current system operation risk level; the purpose is to convert the continuous harmonic lock risk index output by the dynamic instability risk coupling module Into discrete risk levels that are easy to decide and control; The process of determining the current system operation risk level by the risk level assessment module is as follows: Normalization of risk index: This step aims to eliminate the influence of specific working conditions and individual differences of machine tools, and obtain a universal risk assessment standard; the module will calculate the harmonic locking risk index in real time and compare it with the preset risk index critical threshold value to generate a normalized risk assessment index ; Risk index critical threshold value refers to the critical value determined by a large amount of historical data analysis or simulation experiment, at which the system transitions from a critical unstable state to a sudden chatter state ; represents the upper limit of risk that the system can withstand The determination method is: in a controlled experimental environment, systematically adjust the spindle speed and feed rate until critical chatter occurs; in this process, record and save the harmonic locking risk index value in real time; statistically analyze all the values of the harmonic locking risk index at the moment of critical chatter, and take the 95% quantile as the critical threshold value ; this statistical-based method can effectively eliminate data noise and randomness, making the determined threshold more robust The calculation formula is: ; is a dimensionless value, which theoretically indicates that chatter is about to occur when it equals 1 : real-time calculated harmonic locking risk index : risk index critical threshold value, which is the critical value determined by historical data analysis or simulation experiment, at which the system transitions from a critical unstable state to a sudden chatter state ; Risk level determination: This step divides the system operation risk into three clear levels according to the value of the normalized risk assessment index ; in this embodiment, two key preset thresholds are defined: the first preset threshold is preferably set to 0.7 and the second preset threshold is preferably set to 1.0 When the risk assessment index is lower than the first preset threshold value , it is determined that the system operation risk level is safe; at this level, the system is considered to be in a stable cutting state and does not require any intervention When the risk assessment index is between the first preset threshold value and the second preset threshold value , the system operation risk level is determined as a first warning level; this level indicates that the system has entered a critical instability region, although no severe flutter has occurred, the risk is significantly increased, and preventive intervention measures need to be started; When the risk assessment index is higher than a second preset threshold, , the system operation risk level is determined as a second warning level; this level indicates that the system has or will have a sudden flutter, the risk is extremely high, and strong intervention or emergency shutdown procedures must be immediately executed; This embodiment establishes a clear and quantitative risk assessment standard through normalization of the risk index and setting of the grading threshold; this approach converts the complex and continuous risk dynamics into three levels of safety, first warning, and second warning that can be clearly understood by both the machine and the operator, providing direct and reliable decision input for the subsequent closed-loop active control module to execute accurate and appropriate intervention strategies, avoiding the problems of insufficient or excessive intervention.

[0022] The closed-loop active control module predicts the risk evolution gradient, including: A pre-trained recurrent neural network model is used to input a multi-dimensional time series feature data set containing historical multi-scale physical parameters and risk assessment indexes; The recurrent neural network model outputs a risk index prediction value after a preset time step in the future; Based on the risk index prediction value and the current risk assessment index, the risk evolution gradient is calculated; This embodiment is a specific implementation of the risk evolution gradient prediction function in the closed-loop active control module of embodiment 1; the core purpose is to give the system the ability to look ahead, that is, not only to know how high the current risk is, but also to predict the change speed and trend of the risk, so as to realize the change from repairing the lost sheep to preparing for the rainy day; The process of the closed-loop active control module predicting the risk evolution gradient is as follows: Model construction and training: This embodiment uses a pre-trained recurrent neural network model; recurrent neural network refers to a class of deep learning models that are particularly suitable for processing time series data, because the internal recurrent structure can effectively capture the dependence of data in the time dimension; Preferably, the model is composed of an LSTM layer containing 64 neurons and a fully connected layer, and is trained using the Adam optimizer; The training of the model is completed offline; the multi-dimensional time series feature data set used for training contains complete data related to system state evolution collected from a large number of historical processing processes; each sample of the data set is a time series segment, and the feature dimension contains historical multi-scale physical parameters, i.e. Kurtosis change rate Relative dielectric constant fluctuation value Risk assessment index The label is the actual risk assessment index at a future time point after the end point of the time series; Online prediction and gradient calculation: In the real-time running phase of the system, the module performs the following operations: Input multi-dimensional time series feature data set containing historical multi-scale physical parameters and risk assessment indexes: the module continuously caches a small time window in the past, for example, all related data in the past 5 seconds, to form an input sequence consistent with the training format; Output risk index prediction value after a preset time step in the future by the recurrent neural network model: provide the above input sequence to the pre-trained RNN model, and the model performs a forward propagation calculation, and the output is the risk index prediction value after a preset time step in the future After that The preset time step is a parameter that can be optimized and adjusted according to the response speed of the control system and the specific process characteristics, for example, it can be set to 0.5 seconds; Based on the risk index prediction value and the current risk assessment index, the risk evolution gradient is calculated: the module obtains the current risk assessment index provided by the risk level assessment module, and calculates the risk evolution gradient using the following formula : ; Wherein: The risk evolution gradient, the physical dimension is -1 , the positive and negative signs represent the growth or decline trend of the risk, and the absolute value size represents the degree of trend intensity; The risk index prediction value is dimensionless, and is obtained from the RNN model output; The current risk assessment index is dimensionless, and is obtained from the risk level assessment module; The prediction time step is in seconds, and is obtained from the preset parameter; By introducing the prediction model based on RNN, the embodiment successfully adds the time dimension of foresight to the system; the calculated risk evolution gradient becomes a key decision variable independent of the current risk level; a larger positive Even if the current risk level is still in the early stage of the first warning, it can predict that the system state is deteriorating rapidly, thereby triggering earlier and more decisive intervention; this prediction capability enables the control strategy to intervene before the risk reaches its peak, greatly improving the effectiveness and proactivity of the intervention measures, and is the core technical innovation to achieve predictive maintenance and control.

[0023] Embodiment 4 The closed-loop active control module performs hierarchical intervention, including: When the system operation risk level is a first warning level and the risk evolution gradient exceeds the preset gradient threshold, a first intervention is triggered; The first intervention includes: generating a modulation signal based on the risk assessment index and the risk evolution gradient; and superimposing the modulation signal on the spindle command speed to start spindle speed fine tuning; The closed-loop active control module performs hierarchical intervention, further including: When the system operation risk level is a second warning level, a second intervention is triggered; The second intervention includes: forcibly reducing the spindle speed and the feed rate to a preset safe idle speed state; and automatically executing an emergency stop when the risk assessment index fails to fall within a preset time; This embodiment is a specific implementation of the hierarchical intervention function in the closed-loop active control module of Embodiment 1, integrating the above technical features to form a complete closed-loop control strategy dynamically matched with the risk level and evolution gradient; the purpose is to achieve the most effective risk suppression with the least impact on processing based on the evaluation and prediction results; The logic of the closed-loop active control module performing hierarchical intervention is as follows: First intervention, active suppression Triggering condition: when the system operation risk level is a first warning level, i.e. And the risk evolution gradient exceeds the preset gradient threshold That is, , trigger the first intervention; the preset gradient threshold is a threshold for determining whether the risk growth rate is worth attention, and the setting basis is to balance the sensitivity of intervention and avoid excessive intervention affecting processing efficiency, for example, it can be set to 0.1⁻¹ ; This double condition ensures that intervention is started only when risk exists and is rapidly increasing, avoiding unnecessary adjustments; Execution action: the core of the first intervention is to start spindle speed fine tuning, actively changing the excitation frequency to break the harmonic locking condition; Based on the risk assessment index And the risk evolution gradient , a modulation signal is generated; The modulation signal is superimposed on the spindle command speed ; real-time rotating speed The adjustment logic is as follows: ; Wherein: : spindle command speed, physical dimension of Hz, from the basic rotating speed set by numerical control program; : modulation amplitude, dimensionless, is a function positively related to risk degree, which determines the range of speed fine-tuning; in the embodiment, it is designed as a piecewise linear function to realize accurate matching of intervention intensity: The relationship between modulation amplitude and frequency in the formula and risk index is an empirical function calibrated through experiment, aiming to effectively capture the dose-effect relationship between intervention intensity and risk degree in a simplified way; When , ; When , ; Initial amplitude and gain coefficient The calibration of the initial amplitude and the gain coefficient is completed through a closed-loop step response experiment. When the system is in the first early warning state, such as , the modulation amplitude is gradually increased from zero until the system risk index begins to fall significantly. Record this amplitude as , when approaching the second early warning threshold, such as , repeat the process to determine , by repeating the experiment at different values, the gain coefficient is fitted using the least squares method; this method ensures accurate matching of intervention intensity and risk degree; Wherein, is the initial amplitude of each interval, is the gain coefficient, which are parameters calibrated through experiment according to the specific machine tool dynamics; : modulation frequency, physical dimension of Hz, determines the speed of change; in a preferred embodiment, the frequency can be set to the optimal fixed frequency that can most effectively break the harmonic lock ; to further enhance adaptability, it can also be fine-tuned according to the risk evolution gradient , for example: ; Wherein is the sensitivity coefficient of frequency adjustment; Second intervention, forced protection Trigger condition: when the system runs at the secondary warning level of risk rating, i.e. , this is an emergency situation that the system has entered or is about to enter a dangerous state, triggering the secondary intervention immediately; this is the highest priority instruction, indicating that the system has entered or is about to enter a dangerous state of flutter; Execution action: the secondary intervention takes more decisive protective measures; First stage: forced speed reduction; the system immediately issues an instruction to forcibly reduce the spindle speed and feed rate to a preset safe idle speed state; the safe idle speed state refers to a low-speed running state that is much lower than the normal machining parameters and has been experimentally verified to be absolutely free of flutter; The parameters of the safe idle speed state should be determined by the vibration-speed-feed rate map during machine tool acceptance or system debugging; according to the map, the spindle speed and the feed rate are selected to be in the absolutely stable region, i.e. values that have no risk of flutter, such as 100 Hz for the spindle speed and 50 mm / min for the feed rate; these values should be referenced to the natural modal frequency of the machine tool to avoid all possible harmonic resonance points; Second stage: emergency stop; the system starts an internal timer; if the risk assessment index fails to fall back to the safe level within a preset time, e.g. 2 seconds, the system will automatically execute an emergency stop to cut off the power supply of the spindle and feed servo, maximizing the safety of personnel, workpieces and equipment; This embodiment perfectly balances machining efficiency and absolute safety through this set of hierarchical intervention logic; the primary intervention is a surgical precision adjustment that skillfully resolves potential risks without interrupting machining through minor parameter perturbations, reflecting the level of system intelligence and fine control; the secondary intervention is a solid safety bottom line that avoids catastrophic consequences through decisive forced measures in the extreme case of risk out of control; this control strategy of combining hardness and softness in a step-by-step progression makes the system have high efficiency, precision and robustness in actual application.

[0024] 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, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A metal automated machine tool operation state prediction system, characterized by, The method comprises the following steps: a data acquisition module for acquiring multi-scale physical parameters representing the operating state of the machine tool in real time, the multi-scale physical parameters comprising a structural thermal gradient tensor, a kurtosis change rate of a tool interface signal, a relative fluctuation value of a coolant dielectric constant, and a low-frequency seismic noise spectrum of an environment; a dynamic instability risk coupling module for determining a harmonic locking risk index based on the multi-scale physical parameters acquired by the data acquisition module; a risk level assessment module for determining a current system operating risk level based on the harmonic locking risk index determined by the dynamic instability risk coupling module, and generating a normalized risk assessment index; a closed-loop active control module for predicting a risk evolution gradient based on historical multi-scale physical parameters and the risk assessment index generated by the risk level assessment module, and performing hierarchical intervention to adjust the operating parameters of the machine tool according to the current system operating risk level and the risk evolution gradient.

2. The metal automatic processing machine tool operation state prediction system according to claim 1, characterized in that, The data acquisition module is specifically configured to: acquire the structural thermal gradient tensor through a thermocouple array arranged on a key structure of the machine tool; acquire the tool interface signal through an acoustic emission sensor installed on a spindle box, and calculate the kurtosis change rate; monitor the coolant dielectric constant through a dielectric sensor deployed on a coolant circulation pipeline, and determine the relative fluctuation value; acquire the low-frequency seismic noise spectrum through an environmental vibration sensor arranged on a machine tool base.

3. The metal automatic processing machine tool operation state prediction system according to claim 1, characterized in that, The dynamic instability risk coupling module determines the harmonic locking risk index, comprising: calculating the shifted system modal frequency based on the structural thermal gradient tensor and a preset initial modal frequency; calculating the cutting process excitation intensity factor based on the kurtosis change rate and the relative fluctuation value; determining the harmonic locking risk index in combination with the shifted system modal frequency, the cutting process excitation intensity factor, and the low-frequency seismic noise spectrum.

4. The metal automatic processing machine tool operation state prediction system according to claim 3, characterized by, The calculation of the shifted system modal frequency comprises: quantifying the overall machine dynamic stiffness weakening effect caused by the structural thermal gradient tensor using a preset thermal softening coefficient, and determining the shifted system modal frequency as a representation.

5. The metal automatic processing machine tool operation state prediction system according to claim 3, characterized in that, The calculation of the cutting process excitation intensity factor comprises: performing normalization processing on the real-time acquired kurtosis change rate and relative fluctuation value respectively using preset kurtosis change rate critical values and dielectric constant relative fluctuation critical thresholds; performing weighted summation on the two normalized values according to preset weight coefficients to generate the cutting process excitation intensity factor.

6. The metal automated machine tool operation state prediction system according to claim 1, wherein The risk level assessment module determines the current system operating risk level, comprising: comparing the harmonic locking risk index with a preset risk index critical threshold to generate a normalized risk assessment index; when the risk assessment index is lower than a first preset threshold, determining that the system operating risk level is a safe level; when the risk assessment index is between the first preset threshold and a second preset threshold, determining that the system operating risk level is a first-level early warning level; when the risk assessment index is higher than the second preset threshold, determining that the system operating risk level is a second-level early warning level.

7. The metal automated machine tool operation state prediction system according to claim 1, wherein The closed-loop active control module predicts the risk evolution gradient, comprising: inputting a multi-dimensional time sequence feature data set containing historical multi-scale physical parameters and risk assessment indexes into a pre-trained recurrent neural network model; output a risk index prediction value for a future preset time step by the recurrent neural network model; based on the risk index prediction value and the current risk assessment index, calculate a risk evolution gradient.

8. The metal automated machine tool operation state prediction system according to claim 1, wherein, The closed-loop active control module performs hierarchical intervention, including: When the system operation risk level is a first warning level and the risk evolution gradient exceeds a preset gradient threshold, a first intervention is triggered. The first intervention includes: generating a modulation signal based on the risk assessment index and the risk evolution gradient; and superimposing the modulation signal on the spindle command speed to start spindle speed fine tuning.

9. The metal automated machine tool operation state prediction system according to claim 8, wherein, The closed-loop active control module performs hierarchical intervention, including: When the system operation risk level is a second warning level, a second intervention is triggered. The second intervention includes: forcibly reducing the spindle speed and the feed rate to a preset safe idle speed state; and automatically executing an emergency stop when the risk assessment index fails to fall within a preset time.

Images