A system for predicting energy consumption of a turning process of a numerical control lathe

Abstract

The numerical control lathe turning process energy consumption prediction system belongs to the technical field of data processing and prediction, comprises a multi-source data acquisition unit, is used for synchronously collecting high-frequency mechanical vibration signals and low-frequency electrical power signals of a machine tool, and collecting processing process parameters, a steady-state power prediction unit is used for generating a steady-state power baseline based on the low-frequency electrical power signals and the processing process parameters collected by the multi-source data acquisition unit, and the prediction accuracy under abnormal conditions is improved, the transient vibration information of the physical world is fused, the limitation of the energy consumption slow change hypothesis is broken through, a new paradigm of mutation perception type energy consumption prediction is created, the transformation from passive response to active predictive maintenance is realized, and the workpiece rejection and unexpected downtime risk is greatly reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing and prediction, in particular to a numerical control lathe turning process energy consumption prediction system. BACKGROUND

[0002] The energy consumption prediction of numerical control lathes is a key link to realize lean production and green manufacturing. The existing prediction methods are mostly based on the assumption of steady-state operation, using regression or time series models to process the low-frequency electrical power data of the machine tool. However, such methods completely ignore the physical phenomena caused by tool state mutations such as wear, edge collapse, and fracture during the cutting process. These mutation events can cause a dramatic change in cutting force on a millisecond time scale, resulting in transient energy consumption peaks in the machine tool servo system. Since the existing model paradigm cannot capture such high-frequency transient events, its prediction accuracy is very low under abnormal working conditions, losing the ability to warn, and energy consumption management is left to the passive post-statistics stage. Therefore, there is an urgent need in the field for a new technical solution that can integrate transient physical characteristics and accurately predict energy consumption mutations.

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

[0004] The purpose of the present application is to provide a numerical control lathe turning process energy consumption prediction system to solve the problems raised in the above BACKGROUND.

[0005] The technical solution of the present application is as follows: a multi-source data acquisition unit is used to synchronously acquire high-frequency mechanical vibration signals and low-frequency electrical power signals of the machine tool, and to acquire processing parameters;

[0006] A steady-state power prediction unit is used to generate a steady-state power baseline based on the low-frequency electrical power signals and the processing parameters acquired by the multi-source data acquisition unit;

[0007] A transient impact quantification unit is used to generate a transient mutation impact factor representing tool state mutations based on the high-frequency mechanical vibration signals acquired by the multi-source data acquisition unit;

[0008] A fusion prediction generation unit is used to generate a final power prediction value by combining the steady-state power baseline generated by the steady-state power prediction unit, the transient mutation impact factor generated by the transient impact quantification unit, and the processing parameters acquired by the multi-source data acquisition unit.

[0009] Preferably, the steady-state power prediction unit is specifically used for:

[0010] An autoregressive model with external inputs is adopted; historical values of the low-frequency electrical power signal and current values of the machining process parameters are set as inputs of the model; and the steady-state power baseline is calculated based on the inputs of the model.

[0011] Preferably, the transient impact quantification unit is further configured to:

[0012] The high-frequency mechanical vibration signal is decomposed using a wavelet packet transform to generate sub-signals covering different frequency bands; and a feature frequency band set is determined based on the sub-signals and the feature frequency band set, and a feature frequency band related to a tool state mutation is determined.

[0013] Preferably, the transient impact quantification unit is specifically configured to:

[0014] The sum of short-time energies in the feature frequency band is calculated; a statistical mean of the sum of short-time energies in a healthy cutting state is determined; a statistical standard deviation of the sum of short-time energies in the healthy cutting state is determined; and the transient mutation impact factor is calculated using a Z-score normalization method in combination with the sum of short-time energies, the statistical mean, and the statistical standard deviation.

[0015] Preferably, the transient impact quantification unit further comprises a dynamic parameter updating module, which is configured to:

[0016] The transient mutation impact factor is compared with a preset stability threshold;

[0017] If the transient mutation impact factor is lower than the stability threshold, the statistical mean and the statistical standard deviation are continuously updated using a sliding time window;

[0018] If the transient mutation impact factor is not lower than the stability threshold, the statistical mean and the statistical standard deviation are kept unchanged.

[0019] Preferably, the fusion prediction generation unit is specifically configured to:

[0020] The transient mutation impact factor is compared with a preset impact factor activation threshold;

[0021] If the transient mutation impact factor is greater than the impact factor activation threshold, a transient compensation term is determined, and the final power prediction value is generated in combination with the steady-state power baseline and the transient compensation term; the transient compensation term is generated based on a difference between the transient mutation impact factor and the impact factor activation threshold, a preset energy impact reference conversion coefficient, and a working condition adaptive weight;

[0022] If the transient mutation impact factor is not greater than the impact factor activation threshold, the steady-state power baseline is directly determined as the final power prediction value.

[0023] Preferably, the working condition adaptive weight is a dimensionless dynamic weight function; the dynamic weight function is used to dynamically adjust the amplitude of the transient compensation term according to the deviation of the machining process parameter relative to the preset reference process parameter.

[0024] Preferably, the dynamic weight function includes a dimensionless sensitivity coefficient; the determination process of the dimensionless sensitivity coefficient includes:

[0025] Repeating the mutation experiment under different machining process parameters to obtain a data point set of energy impact reference conversion coefficients; linear regression fitting is performed on the data point set to determine the dimensionless sensitivity coefficient.

[0026] The present application improves the energy consumption prediction system for the turning process of a numerical control lathe, and has the following improvements and advantages compared with the prior art:

[0027] 1. The prediction accuracy under abnormal working conditions is improved, and a new paradigm of mutation perception type energy consumption prediction is established by fusing the transient vibration information of the physical world, breaking through the limitation of the energy consumption slow change hypothesis;

[0028] 2. The system can capture the precursor events such as micro-chipping of the tool before the tool is severely damaged by monitoring the transient compensation term in the predicted energy consumption , realizing the transition from passive response to active predictive maintenance, and greatly reducing the risk of workpiece rejection and unexpected downtime;

[0029] 3. Each energy consumption mutation event recorded by the present application, time, amplitude and machining parameter constitute a high-value data point, through correlation analysis with the workpiece quality detection data produced in the same time period, the quantitative relationship between the specific energy consumption mutation mode and the surface vibration lines, size deviation and other quality defects can be established, providing unprecedented data insight for the closed-loop optimization of process parameters. BRIEF DESCRIPTION OF DRAWINGS

[0030] The present application will be further explained in conjunction with the drawings and examples:

[0031] Figure 1 is a flowchart of the energy consumption prediction system for the turning process of a numerical control lathe of the present application;

[0032] Figure 2 is a step flowchart of the fusion prediction generation unit of the present application. DETAILED DESCRIPTION

[0033] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with specific examples.

[0034] Embodiment 1

[0035] Please refer to Figure 1 The present application provides a numerical control lathe turning process energy consumption prediction system, comprising: a multi-source data acquisition unit, used for synchronously acquiring high-frequency mechanical vibration signals and low-frequency electrical power signals of a machine tool, and acquiring machining process parameters;

[0036] A steady-state power prediction unit is used for generating a steady-state power baseline based on the low-frequency electrical power signals and the machining process parameters acquired by the multi-source data acquisition unit;

[0037] A transient impact quantification unit is used for generating a transient mutation impact factor representing tool state mutation based on the high-frequency mechanical vibration signals acquired by the multi-source data acquisition unit;

[0038] A fusion prediction generation unit is used for generating a final power prediction value by combining the steady-state power baseline generated by the steady-state power prediction unit, the transient mutation impact factor generated by the transient impact quantification unit, and the machining process parameters acquired by the multi-source data acquisition unit.

[0039] The present application realizes energy consumption monitoring and management of the turning process by predicting instantaneous power. In order to overcome the technical defects of the prior art that only rely on slowly varying electrical signals and cannot capture instantaneous power impact caused by tool state mutation, the numerical control lathe turning process energy consumption prediction system disclosed in the embodiment constructs an innovative dual-flow information fusion framework. The system realizes comprehensive and accurate prediction of energy consumption through the four core units, i.e. the multi-source data acquisition unit, the steady-state power prediction unit, the transient impact quantification unit and the fusion prediction generation unit, which are tightly coupled by internal logic. The fundamental advantage of this system architecture is that it not only continues the steady-state modeling of macro energy consumption trends, but also revolutionarily introduces quantitative representation of micro mutations of physical processes. By coupling low-frequency electrical power data representing macro trends with high-frequency mechanical vibration data revealing micro physical impacts across scales, the system can generate a final power prediction value containing a steady-state baseline and transient compensation. This design breaks through the steady-state assumption limitation of traditional energy consumption prediction models, so that the prediction result can accurately reflect the power peak caused by sudden events such as tool collapse and fracture, thereby improving the prediction accuracy under abnormal conditions from less than 50% of traditional methods to more than 90%, and creating a new paradigm of mutation-aware energy consumption prediction.

[0040] Embodiment 2

[0041] The steady-state power prediction unit is specifically used for:

[0042] An autoregressive model with external input is adopted; the historical values ​​of the low-frequency electrical power signal and the current values ​​of the processing parameters are set as the input of the model; based on the input of the model, the steady-state power baseline is calculated.

[0043] In this embodiment, the steady-state power prediction unit aims to construct a theoretical power benchmark that can accurately reflect the machine tool under specific process parameters when no abnormal events occur. To achieve this goal, the unit uses an autoregressive model with external input, which deeply reveals the historical dependence of power itself and the physical correlation with external process parameters. The unit receives the low-frequency electrical power signal provided by the multi-source data acquisition unit. The historical sequence values ​​and the processing parameters analyzed in real time from the machine tool CNC system, including the spindle speed , feed rate and cutting depth , and solve for the steady-state power baseline based on these inputs ;

[0044] The process is rigorously defined by the following formula:

[0045] ;

[0046] This method organically combines classic autoregressive time series analysis with multivariate linear regression. The technical motivation is to establish a prediction model that can distinguish between intrinsic inertia and external driving factors of power, thereby providing a stable and adaptive reference benchmark for subsequent stripping of abnormal shocks.

[0047] in, Indicates The predicted steady-state power baseline at each moment, with the physical dimension being watts (W); Representatives in The moment before The historical power value of the sampling point, the physical dimension is watt (W), is the index of the historical time step, and n is the model order; is the autoregressive coefficient, which quantifies the influence of historical power on current power and is a dimensionless parameter; , , They are Spindle speed (r / min), feed rate (mm / min) and cutting depth (mm) at the moment; are the physical influence coefficients of the corresponding process parameters, and their dimensions must be matched according to the physical quantities to ensure that the dimension of the corresponding terms on the right side of the equation is Watt; is the constant bias term of the model, representing the basic standby power, with the physical dimension of watt (W); all model coefficients With constant term are trained by using system identification method or multivariate linear regression algorithm on a large amount of historical normal machining data;

[0048] The application and technical effects are that, by performing the calculation, the steady-state power prediction unit can output a steady-state power baseline highly fitted to the current actual working condition ; the baseline is not a static value, but is dynamically adjusted with real-time changes of the process parameters , , The core technical effect is to provide a power reference in an ideal state for the entire prediction system, so that the subsequent fusion prediction unit can accurately judge whether the actual power deviates from the normal range and accurately calculate the degree of deviation, which is a key prerequisite for realizing high-precision abnormal impact quantification.

[0049] Embodiment 3:

[0050] Before generating the transient mutation impact factor, the transient impact quantification unit is further used for:

[0051] performing decomposition on the high-frequency mechanical vibration signal by using wavelet packet transform to generate sub-signals covering different frequency bands; determining a feature frequency band related to tool state mutation based on the preset feature frequency band set and the sub-signals;

[0052] The transient impact quantification unit is specifically used for:

[0053] calculating the sum of short-time energies in the feature frequency band; determining a statistical mean of the sum of short-time energies in the healthy cutting state; determining a statistical standard deviation of the sum of short-time energies in the healthy cutting state; and combining the sum of short-time energies, the statistical mean and the statistical standard deviation, and using a Z-score standardization method to solve the transient mutation impact factor.

[0054] In this embodiment, the transient impact quantification unit undertakes the key task of refining core mutation information from complex high-frequency mechanical vibration signals ; in order to accurately capture the instantaneous stress wave generated by tool micro-damage, the unit first performs multi-layer decomposition on the collected high-frequency vibration signals by using wavelet packet transform, and the purpose of this is to decompose the original signal of a wide frequency band into a series of sub-signals with concentrated energy and different frequency ranges, so as to realize fine time-frequency analysis of the signal; according to the previous offline calibration experiment, that is, by comparing the vibration energy spectrum of a new tool, a normally worn tool and a collapsed tool, the frequency band with the most significant energy change when the tool state deteriorates is selected to form a preset feature frequency band set ; this step ensures that the focus of subsequent analysis is concentrated on the signal component most related to tool state mutation;

[0055] After determining the characteristic frequency band, the transient impact quantification unit converts the vibration energy with physical dimensions into a dimensionless impact factor with clear statistical significance through a standardized statistical measurement method. ; The calculation process is defined by the following formula:

[0056] ;

[0057] The origin and technical motivation of this formula stem from the Z-score normalization method in statistics. The technical motivation lies in the fact that raw vibration energy fluctuates with cutting conditions, such as speed and feed rate, and directly using energy values ​​cannot provide fair comparisons between different conditions. Through Z-score normalization, the vibration energy under any condition can be converted into a standardized deviation index. The value directly reflects the degree to which the current state deviates from its normal baseline, thus achieving unified measurement and real-time alerting of abnormal events.

[0058] in, for The dimensionless transient mutation impact factor output at every moment; is Moment, located at Decomposition layer, The short-time signal energy within a characteristic frequency band is calculated by the square sum of the corresponding wavelet packet coefficients, and its direct source is the high-frequency vibration signal ; It is a preset set of feature band indexes, which comes from offline calibration experiments; is the total energy of all characteristic frequency bands in healthy cutting state The statistical mean of , as a dynamic benchmark parameter; The statistical standard deviation of the total energy under healthy cutting conditions is also a dynamic benchmark parameter;

[0059] The transient impact quantification unit of this formula successfully extracts a complex physical signal that changes with working conditions into a concise, standardized, dimensionless indicator through this calculation. ; The output value of this indicator directly and objectively reflects the health of the tool state; when When the value is very small, for example, close to 0, it indicates that the system is in a stable and healthy cutting state; when its value increases sharply, it clearly indicates the occurrence of sudden events such as tool chipping or severe wear; this result provides a core and quantitative input basis for the subsequent fusion prediction unit to perform accurate energy compensation.

[0060] Example 4:

[0061] The transient impact quantization unit further includes a dynamic parameter updating module, which is used to:

[0062] Compare the transient mutation impact factor with a preset stability threshold;

[0063] If the transient mutation impact factor is lower than the stability threshold, the sliding time window is used to continuously update the statistical mean and statistical standard deviation;

[0064] If the transient mutation impact factor is not lower than the stability threshold, the statistical mean and statistical standard deviation remain unchanged.

[0065] In order to further enhance the robustness and adaptability of the system, a dynamic parameter update module is integrated into the transient impact quantization unit in this embodiment; the core function of this module is to ensure that the parameters used to calculate the impact factor Statistical benchmark parameters and It can dynamically adapt to the normal drift of the working conditions while avoiding being contaminated by the abnormal events themselves; the operating logic of this module is: in the initial stable stage of the new tool starting processing, the system collects a period of data and calculates and The initial value of; in subsequent processing, the module will calculate the transient mutation impact factor in real time The comparison is performed continuously with a preset stability threshold, for example, 1.5 based on statistical confidence level; when When the value is lower than the stability threshold, the system determines that it is in a healthy cutting state. At this time, the module will enable a sliding time window to continuously and gradually recalculate and update and On the contrary, once If the value is not lower than or even exceeds the stability threshold, the system will determine that an abnormal impact may have occurred, and the module will freeze immediately. and This conditional update mechanism can effectively track the normal drift of the vibration baseline caused by slight changes in workpiece material or macro adjustments of processing parameters, ensuring At the same time, by pausing updates during anomalies, it avoids incorporating abnormal data into the normal model, ensuring the detection sensitivity of subsequent real abnormal events.

[0066] Example 5:

[0067] like Figure 2 As shown, the fusion prediction generation unit is specifically used for:

[0068] Compare and analyze the transient mutation impact factor with the preset impact factor activation threshold;

[0069] If the transient mutation impact factor is greater than the impact factor activation threshold, a transient compensation term is determined and combined with the steady-state power baseline to generate a final power prediction value; the transient compensation term is generated based on the difference between the transient mutation impact factor and the impact factor activation threshold, the preset energy impact benchmark conversion coefficient, and the operating condition adaptive weight;

[0070] If the transient mutation impact factor is not greater than the impact factor activation threshold, the steady-state power baseline is directly determined as the final power prediction value;

[0071] The working condition adaptive weight is a dimensionless dynamic weight function; the dynamic weight function is used to dynamically adjust the amplitude of the transient compensation term according to the deviation of the processing parameters relative to the preset reference process parameters.

[0072] As the core innovation hub of the present invention, the fusion prediction generation unit performs the final power prediction calculation; this unit adopts a baseline + compensation logic framework to convert the steady-state power baseline from the steady-state power prediction unit into a and the transient sudden impact factor from the transient impact quantization unit Perform nonlinear coupling to generate the final predicted energy consumption value In order to realize an intelligent correction mechanism that does not intervene unless necessary, the unit will obtain with a preset shock factor activation threshold This threshold is based on the comparison of normal and abnormal samples in a large amount of historical data. Distribution, determined by statistical analysis, such as setting it to 3 times the standard deviation, corresponding to a 99.7% confidence interval or receiver operating characteristic curve analysis, aims to optimally balance false positives and false negatives;

[0073] Only when The value exceeds The system will then recognize that a significant physical shock has occurred and activate the transient compensation item. If the threshold is not exceeded, the final predicted value is directly equal to the steady-state power baseline. This fusion process is driven by the following core formula:

[0074] ;

[0075] This formula is a gated weighted compensation model driven by physical meaning; its technical motivation is that the increase in instantaneous power should be proportional to the severity of the impact, but this relationship does not hold true under all small disturbances; the introduction of the gating function and activation threshold , the system can effectively avoid overreaction to normal signal noise, ensuring that the compensation term is only activated when a significant mutation actually occurs; at the same time, considering that the same physical impact under different processing loads will result in different power increments, a working condition adaptive weight is introduced to dynamically adjust the amplitude of the compensation .

[0076] wherein, is the total power prediction value of the system final output, with the physical dimension of watt (W); is the power baseline from the steady-state power prediction unit (W); is the dimensionless impact factor from the transient impact quantification unit; is the dimensionless impact factor activation threshold; is the core energy impact reference conversion coefficient, with the physical dimension of watt (W), which defines the instantaneous power increment corresponding to a unit impact factor (i.e. exceeding one unit) under the reference working condition; is a dimensionless working condition adaptive weight function, which dynamically adjusts the amplitude of the compensation term according to the deviation of the current process parameters from the preset reference process parameters, ensuring the consistency of the physical dimension;

[0077] Through this fusion mechanism, the system realizes intelligent and fine management of energy consumption prediction; during normal processing, the prediction value is smoothly equal to the steady-state baseline; when an abnormality such as tool collapse occurs, the system can instantly calculate a transient compensation term proportional to the impact severity and the current processing load, and superimpose it on the baseline; this enables the final prediction curve to accurately reproduce the real power spike, greatly improving the prediction accuracy, and more importantly, through monitoring the occurrence and size of the transient compensation term , active early warning of tool damage can be achieved, providing unprecedented decision support for preventive maintenance and ensuring processing quality.

[0078] Embodiment 6:

[0079] The dynamic weight function includes a dimensionless sensitivity coefficient; the determination process of the dimensionless sensitivity coefficient includes:

[0080] Repeating the mutation experiment under different processing parameters to obtain a data point set of the energy impact reference conversion coefficient; performing linear regression fitting on the data point set to determine the dimensionless sensitivity coefficient.

[0081] To ensure that the working condition adaptive weight The physical accuracy and engineering reliability, the embodiment further discloses a robust calibration method of its internal dimensionless sensitivity coefficient; to determine the sensitivity coefficient related to the spindle speed For example, its calibration process abandons the method that may be unstable due to single-point measurement error, and instead adopts a more statistically significant linear regression fitting strategy; The steps are as follows: the feed rate And the cutting depth Fixed at their reference values And ; Secondly, a series of, for example, at least 3 different spindle speed values Covering the common working range are selected ; Then, at each selected speed , the experiment for calibrating the energy impact reference conversion coefficient is repeated, for example, by artificially manufacturing a series of different degrees of mutation in the cutting process, so as to obtain the conversion coefficient value under the specific working condition ; After the experiment is completed, a set of key data points is obtained, in the form of , wherein the abscissa represents the dimensionless relative deviation degree of the spindle speed, and the ordinate is the energy conversion coefficient measured under the deviation degree; Finally, based on the relationship , the linear regression fitting is carried out on the set of data points; In this fitting, the intercept of the obtained straight line should theoretically approach the reference conversion coefficient , and the slope of the straight line is a robust estimate of the product ; Therefore, the final dimensionless sensitivity coefficient Can be calculated by the formula fitting slope And Determination also adopts the same process; This calibration method based on multiple-point regression greatly enhances the accuracy and anti-interference ability of the sensitivity coefficient, ensures that the working condition adaptive weight Can accurately and dynamically scale the transient energy consumption compensation under various machining conditions, thereby providing a solid guarantee for the entire prediction system to maintain high precision in complex and variable production environments;

[0082] To further clarify the composition of the working condition adaptive weight function, the mathematical expression can be linearly superimposed by the relative deviation degree of each process parameter and its corresponding sensitivity coefficient, in the form of:

[0083] ;

[0084] Wherein, Respectively, the preset reference spindle speed, the reference feed rate and the reference cutting depth, as the reference for normalization processing; : A dimensionless working condition adaptive weight function, which is used to dynamically adjust the amplitude of the transient compensation term according to the deviation degree of the current machining process parameters; : A preset reference feed rate, which is used as a reference value when calibrating the sensitivity coefficient; : A preset reference cutting depth, which is used as a reference value when calibrating the sensitivity coefficient; the explicit function expression ensures that the calibrated sensitivity coefficient can be applied to the final power prediction calculation without ambiguity, thereby fully implementing the present application.

[0085] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. 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 present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A CNC lathe turning process energy consumption prediction system, characterized in that: include: Multi-source data acquisition unit, used to synchronously collect high-frequency mechanical vibration signals and low-frequency electrical power signals of machine tools, and collect processing parameters; a steady-state power prediction unit, configured to generate a steady-state power baseline based on the low-frequency electrical power signal and the machining process parameters acquired by the multi-source data acquisition unit; a transient impact quantification unit, configured to generate a transient mutation impact factor representing a sudden change in a tool state based on the high-frequency mechanical vibration signal acquired by the multi-source data acquisition unit; a fusion prediction generation unit, configured to combine the steady-state power baseline generated by the steady-state power prediction unit, the transient sudden change impact factor generated by the transient impact quantification unit, and the processing parameters collected by the multi-source data collection unit to generate a final power prediction value; Before generating the transient sudden change impact factor, the transient impact quantification unit is further configured to: Decomposing the high-frequency mechanical vibration signal using wavelet packet transform to generate sub-signals covering different frequency bands; determining a characteristic frequency band associated with a sudden change in the tool state based on a preset characteristic frequency band set and the sub-signals; The transient impact quantization unit is specifically used for: Calculating the sum of short-time energies within the characteristic frequency band; determining a statistical mean of the sum of short-time energies in a healthy cutting state; determining a statistical standard deviation of the sum of short-time energies in a healthy cutting state; and calculating the transient mutation impact factor using a Z-score normalization method based on the sum of short-time energies, the statistical mean, and the statistical standard deviation; The fusion prediction generation unit is specifically used for: Comparing and analyzing the transient mutation impact factor with a preset impact factor activation threshold; If the transient mutation impact factor is greater than the impact factor activation threshold, a transient compensation term is determined, and the final power prediction value is generated by combining the steady-state power baseline and the transient compensation term; the transient compensation term is generated based on the difference between the transient mutation impact factor and the impact factor activation threshold, a preset energy impact benchmark conversion coefficient, and an operating condition adaptive weight; If the transient mutation impact factor is not greater than the impact factor activation threshold, directly determining the steady-state power baseline as the final power prediction value; This fusion process is driven by the following formula: ； in, is the predicted value of the total power output of the system, with the physical dimension being watt (W); is the power baseline (W) from the steady-state power prediction unit; is the dimensionless shock factor from the transient shock quantization unit; is the dimensionless shock factor activation threshold; It is the core energy impact benchmark conversion coefficient, and its physical dimension is Watt (W). It defines the instantaneous power increment corresponding to the unit impact factor under the reference working condition. The unit impact factor represents Beyond a unit; It is a dimensionless working condition adaptive weight function, which is based on the current process parameters The amplitude of the compensation term is dynamically adjusted relative to the deviation of the preset reference process parameters to ensure the consistency of the physical dimensions.

2. The energy consumption prediction system for a CNC lathe turning process according to claim 1 is characterized in that: The steady-state power prediction unit is specifically used for: An autoregressive model with external input is adopted; historical values ​​of the low-frequency electrical power signal and current values ​​of the processing parameters are set as inputs of the model; and the steady-state power baseline is calculated based on the inputs of the model.

3. The energy consumption prediction system for a CNC lathe turning process according to claim 1 is characterized in that: The transient impulse quantization unit further includes a dynamic parameter updating module, which is configured to: Comparing the transient mutation impact factor with a preset stability threshold; If the transient mutation impact factor is lower than the stability threshold, continuously updating the statistical mean and the statistical standard deviation using a sliding time window; If the transient mutation impact factor is not lower than the stability threshold, the statistical mean and the statistical standard deviation are kept unchanged.

4. The energy consumption prediction system for a CNC lathe turning process according to claim 3 is characterized in that: The working condition adaptive weight is a dimensionless dynamic weight function; the dynamic weight function is used to dynamically adjust the amplitude of the transient compensation term according to the deviation of the processing parameters relative to the preset reference process parameters.

5. The energy consumption prediction system for a CNC lathe turning process according to claim 4 is characterized in that: The dynamic weight function includes a dimensionless sensitivity coefficient; the process of determining the dimensionless sensitivity coefficient includes: The mutation experiment is repeatedly performed under different processing parameters to obtain a data point set of the energy impact benchmark conversion coefficient; and a linear regression fitting is performed on the data point set to determine the dimensionless sensitivity coefficient.

Images