Early warning method for tool wear status in machining centers based on multi-source data fusion

Abstract

This invention belongs to the field of tool wear measurement technology, specifically relating to a tool wear status early warning method for machining centers based on multi-source data fusion. The method includes: calculating cutting power consumption characteristics using a spindle load energy mapping model based on the original sampled current and real-time feed rate of the spindle; sampling the cutting power consumption characteristics using a sliding window and calculating the cutting state disorder; nonlinearly coupling the power consumption gradient term with the cutting state disorder to obtain a wear instability coefficient; establishing a dynamic statistical reference surface; and providing tool status early warning based on the deviation of the real-time monitored wear instability coefficient. This invention effectively identifies and eliminates instantaneous interference caused by uneven material hardness, while simultaneously capturing the signal structure disorder in the early stages of tool failure. It solves the technical problems of high false alarm rates and difficulty in capturing early failure signs in mold steel machining due to material microstructure segregation and varying working conditions, significantly improving the monitoring accuracy of precision machining processes.

Description

Technical Field

[0001] This invention relates to the field of tool wear measurement technology. More specifically, this invention relates to a method for early warning of tool wear status in machining centers based on multi-source data fusion. Background Technology

[0002] In the field of precision mold steel machining, the condition of machining center cutting tools directly determines the dimensional accuracy and surface roughness of the mold surface. Especially when precision machining high-hardness mold steel, the tools need to operate under high-intensity loads and complex cutting thermal environments for extended periods. To ensure machining quality and prevent sudden tool breakage that could lead to the scrapping of expensive mold blanks, real-time monitoring of tool wear has become a core requirement of intelligent machining.

[0003] Currently, the closest existing technology in the industry usually adopts a monitoring scheme based on the amplitude of spindle motor current. The monitoring logic is based on the physical principle that "cutting force and current are positively correlated". When the effective value of the current is detected to exceed a preset single hard threshold, the system determines that the tool has suffered severe wear or failure and immediately triggers a shutdown warning.

[0004] However, this technique has significant limitations in actual mold processing scenarios: mold steel materials often exhibit microstructure segregation or uneven carbide distribution during heat treatment, leading to obvious local hardness fluctuations within the material. When the cutting tool cuts into such high-hardness areas at high speed, even if the tool is still in good condition, the spindle current will generate instantaneous spikes. Existing technologies, focusing only on the macroscopic characteristic of current amplitude, struggle to physically distinguish between random load pulses caused by material inhomogeneity and substantial load increases caused by tool geometric wear, resulting in frequent false alarms. Furthermore, when machining the convex surfaces of precision molds, the system often needs to dynamically adjust the feed parameters based on the radius of curvature to maintain a constant cutting speed. This causes the machine tool's operating power consumption to fluctuate linearly with the feed speed. Fixed-threshold monitoring schemes cannot adapt to this variable load condition, often triggering false alarms during normal speed adjustment phases, significantly disrupting production continuity. Summary of the Invention

[0005] To address the technical problem that existing technologies lack in-depth analysis of signal structure characteristics and dynamic decoupling capability for machine tool operating losses, thus failing to effectively shield against spurious load interference caused by material hardness segregation in complex mold steel machining environments, resulting in insufficient accuracy in tool failure monitoring and inability to meet the requirements of high-reliability precision machining, this invention provides a tool wear state early warning method for machining centers based on multi-source data fusion. The method includes: synchronously acquiring the original sampling current and real-time feed rate of the spindle from the machining center controller bus in real time; calculating the cutting power consumption characteristics using a spindle load energy mapping model; sampling the cutting power consumption characteristics using a sliding window; calculating the cutting state disorder based on the uniformity of signal distribution in the frequency domain within the sliding window; obtaining the rate of change of the cutting power consumption characteristics over time within the current sliding window as a power consumption gradient term; nonlinearly coupling the power consumption gradient term with the cutting state disorder to calculate the wear instability coefficient; collecting the wear instability coefficient sequence at the start of machining to establish a dynamic statistical reference surface; and providing tool state early warning based on the deviation between the real-time monitored wear instability coefficient and the dynamic statistical reference surface.

[0006] This invention eliminates machine tool idle loss by using a spindle load energy mapping model, thus removing interference from variable feed conditions on the signal. It also captures qualitative changes in the signal's frequency domain structure using cutting state disorder, overcoming the insensitivity of traditional amplitude monitoring to early failures. Furthermore, it effectively identifies and eliminates random interference caused by uneven material hardness by nonlinearly coupling the power consumption gradient and disorder through a wear loss instability coefficient. This invention achieves precise quantification and early warning of tool condition during mold steel machining, significantly reducing the risk of precision mold scrapping due to sudden tool breakage while ensuring mold surface accuracy.

[0007] Preferably, the step of calculating the cutting power consumption characteristics through the spindle load energy mapping model includes: In the formula, This refers to the cutting power consumption characteristics; Main spindle input voltage; This is the original sampling current; The motor transmission efficiency coefficient is obtained from the motor nameplate. The real-time power factor of the main spindle is obtained by reading specific system variables or PLC registers built into the system. For real-time feed rate; For machine tools at real-time feed rate The no-load power loss.

[0008] This invention introduces voltage, current, and power factor, and deducts the no-load power loss corresponding to the real-time feed rate, so that the calculated cutting power consumption characteristics can purely reflect the physical interaction between the tool and the workpiece contact interface, thus solving the problem of load reference drift caused by motion power consumption fluctuations during variable parameter machining.

[0009] Preferably, before calculating the cutting power consumption characteristics through the spindle load energy mapping model, the process includes: measuring the no-load power data of the machine tool under different speed gradients; and fitting the no-load power consumption function using the least squares method.

[0010] Preferably, the formula for calculating the degree of disorder in the cutting state is: In the formula, The degree of disorder in the cutting state; The first one in the current sliding window Power spectral density of each spectral component; This is the sum of the power spectral densities of all spectral components within the current sliding window; This represents the total number of spectrum statistical intervals. It is the natural logarithm function.

[0011] This invention introduces energy distribution entropy to calculate the dispersion of spectral components, which can keenly capture the signal structure disorder caused by high-frequency random vibration in the early stage of blade breakage. This makes up for the shortcomings of traditional methods that only focus on macroscopic energy amplitude and ignore microscopic signal evolution, thus improving the sensitivity of failure early warning.

[0012] Preferably, the formula for calculating the wear instability coefficient is: In the formula, The coefficient for stable wear loss; This is the system sensitivity adjustment factor; This refers to the cutting power consumption characteristics; This represents the rate of change of cutting power consumption characteristics over time within the current sliding window; The degree of disorder in the cutting state is calculated in real time; The average cutting state disorder during the initial reference period of machining; These are the nonlinear gated compensation coefficients; It is the hyperbolic tangent function; It is a function for maximizing the value.

[0013] This invention constructs nonlinear gated logic using a hyperbolic tangent function, which enables the system to activate the wear instability coefficient only when energy growth and signal disturbance occur simultaneously. This mathematically forces the removal of spurious increments caused by material hardness deviations, greatly improving the signal-to-noise ratio of the monitoring system.

[0014] Preferably, establishing a dynamic statistical benchmark surface includes: obtaining the mean and standard deviation of the wear loss instability coefficient during the benchmark period; and constructing a dynamic discrimination boundary based on the mean and standard deviation of the wear loss instability coefficient during the benchmark period.

[0015] Preferably, the expression for the dynamic discrimination boundary is: In the formula, The wear loss stability coefficient is for real-time monitoring; This represents the average wear loss stability coefficient during the baseline period. The standard deviation of the wear loss stability coefficient during the baseline period; This is the confidence factor.

[0016] This invention utilizes a statistical discriminant based on mean, standard deviation, and confidence factor to transform the warning threshold from a rigid threshold into a dynamic threshold band with probabilistic statistical significance, effectively avoiding the impact of statistical noise in the production environment on the judgment result.

[0017] Preferably, the sliding window sampling has a length of 1024 sampling points and a step size of 256 sampling points.

[0018] Preferably, the calibration process of the confidence factor includes: continuous cutting under standard process conditions, monitoring the wear amount on the tool flank, recording the dispersion multiple of the wear instability coefficient and the mean value of the wear instability coefficient in the reference period when the wear amount reaches a critical value, and introducing a safety factor to cover the influence of cooling efficiency.

[0019] Preferably, the tool status early warning includes: when the wear instability coefficient monitored in real time meets the discrimination criteria, the machining center controller issues a stop command and simultaneously triggers the audible and visual alarm terminal.

[0020] The beneficial effects of this invention are as follows:

[0021] (1) This invention obtains the original sampling current and real-time feed speed of the spindle in real time from the machining center controller bus, and processes them using the spindle load energy mapping model, thereby separating the no-load power loss of the machine tool itself from the total power, solving the problem of load reference offset under variable parameter machining conditions, so that the calculated cutting power consumption characteristics can eliminate the interference caused by feed speed changes, and purely reflect the energy consumption of the tool and workpiece contact interface, thus laying the physical basis for subsequent accurate monitoring.

[0022] (2) After obtaining a pure cutting signal, the present invention further performs sliding window sampling on the cutting power consumption characteristics and calculates the cutting state disorder, and uses the uniformity of the frequency domain distribution to characterize the microscopic stability of the cutting process. Compared with the existing technology that only relies on amplitude monitoring, the calculation of disorder can capture the qualitative change of the signal structure. Even when the energy increment is not obvious in the early stage of wear, the early damage precursor of the tool can be identified through the dispersion phenomenon of frequency domain energy, which effectively solves the technical defect of the existing technology that is not sensitive to small chipping.

[0023] (3) This invention nonlinearly couples the power consumption gradient term with the cutting state disorder to calculate the wear instability coefficient. The coupling mechanism acts as an intelligent filter, requiring that abnormal energy growth must be accompanied by signal structure disorder before it is judged as instability. This corresponds to the interference problem caused by material hardness segregation in the background technology. Although uneven hardness can cause instantaneous current spikes, the cutting state still maintains relatively stable spectral characteristics and will not trigger large fluctuations in disorder. Through this logical constraint, the system can automatically eliminate false load pulses, greatly reducing the false alarm rate.

[0024] (4) By establishing a dynamic statistical reference surface and combining it with the discrimination criteria to perform automatic early warning and shutdown linkage, the present invention realizes closed-loop control from data calculation to hardware execution, ensuring that the feed is cut off within milliseconds when the risk of tool failure is confirmed, solving the safety hazards caused by the lag of manual inspection, and achieving effective protection of the core components of precision machine tools while ensuring the accuracy of the mold surface. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the tool wear status early warning method for machining centers based on multi-source data fusion in this invention;

[0026] Figure 2 This is a schematic diagram illustrating the comparison between the original sampled current signal and the threshold determination of existing technology;

[0027] Figure 3 It is a schematic diagram showing the distribution of cutting state disorder as the tool wear state evolves;

[0028] Figure 4 This is a schematic diagram illustrating the early warning effect of the wear loss instability coefficient based on the nonlinear entropy control suppression model. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0031] This invention discloses a method for early warning of tool wear status in machining centers based on multi-source data fusion, referring to... Figure 1 This includes steps S1 to S4:

[0032] S1. The original sampling current and real-time feed speed of the spindle are obtained synchronously from the machining center controller bus in real time, and the cutting power consumption characteristics are calculated through the spindle load energy mapping model.

[0033] It should be noted that, since the spindle current signal of a machining center during operation is affected by the superposition of the idle power of various transmission components of the machine tool and different machining parameters, it is impossible to effectively extract the load energy component contributed by tool wear by directly using the original current amplitude. This makes it difficult for the monitoring system to accurately obtain the real machining load fluctuation. Therefore, this invention establishes a spindle load energy mapping model and uses a preset idle power consumption function to deduct the inherent losses of the machine tool, thereby extracting dynamic features that can characterize pure cutting power consumption, providing a high signal-to-noise ratio data basis for subsequent accurate assessment of tool wear status.

[0034] Specifically, the raw sampling current of the spindle is obtained in real time synchronously from the machining center controller bus. With real-time feed rate By conducting preliminary experiments on the machine tool, the no-load power data of the machine tool was measured under different speed gradients, and the no-load power consumption function was obtained by fitting using the least squares method. .

[0035] Furthermore, a calculation model is constructed based on the power balance relationship to obtain the cutting power consumption characteristics. The formula for its calculation is:

[0036]

[0037] In the formula, This refers to the cutting power consumption characteristics; It is a constant coefficient for calculating three-phase power; The spindle input voltage is preferably a dynamic input voltage that is synchronously obtained in real time through the machining center controller bus to eliminate calculation errors caused by power grid fluctuations; under specific voltage stabilization conditions, the rated voltage of the machining center can also be used as the spindle input voltage for calculation. This is the original sampling current; The motor transmission efficiency coefficient is obtained from the motor nameplate. As the real-time power factor of the spindle, the power module of the CNC system has the function of real-time monitoring of voltage and current phase angle. By reading specific system variables or PLC registers built into the system, the real-time power factor data of the current spindle motor can be directly and synchronously obtained. For real-time feed rate; For machine tools at real-time feed rate The no-load power loss.

[0038] This calculation formula achieves precise decoupling of the actual energy consumed at the cutting interface by subtracting the no-load power loss under the corresponding motion state from the total electrical power of the spindle; cutting power consumption characteristics. The larger the value, the stronger the interaction force between the tool and the workpiece. This characteristic eliminates the influence of real-time feed rate. The inherent operating losses of the changing machine tool ensure that the energy increment comes entirely from the load changes during the machining process.

[0039] For example, Figure 2 This is a schematic diagram comparing the original sampled current signal with the threshold judgment of existing technology. The diagram shows the waveform of the original sampled current signal obtained by the current transformer over time during a complete processing cycle. The curve at the top of the diagram that fluctuates violently is the original sampled current signal. The amplitude jump that occurs in the interval between 100s and 115s represents the load pulse caused by the fluctuation of material hardness. The area in the diagram where the current signal exceeds the threshold line and is covered by shaded areas represents the false alarm judgment range caused by the inability of existing technology to distinguish the source of the load.

[0040] S2. Perform sliding window sampling on the cutting power consumption characteristics, and calculate the cutting state disorder based on the uniformity of the signal distribution in the frequency domain within the sliding window.

[0041] It should be noted that when a cutting tool experiences wear or minor chipping, the stability of the cutting process undergoes a qualitative change. This physical failure is not only manifested as an increase in energy amplitude, but also as an increase in the complexity of the signal structure in the frequency domain. If only the energy value is monitored while ignoring the evolution of the signal structure, the subtle disorder characteristics in the early stages of tool failure cannot be captured. Therefore, this invention introduces the energy distribution entropy from informatics to quantify the dispersion of cutting power consumption characteristics in the frequency domain, thereby enabling sensitive detection of signal anomalies caused by changes in the tool's geometric state.

[0042] Specifically, regarding the characteristics of cutting power consumption Continuous sliding window sampling is performed, with the length of the sliding window set to 1024 sampling points and the step size to 256 sampling points. A fast Fourier transform is performed on the signal within each sliding window to convert the power signal in the time domain to the frequency domain.

[0043] Furthermore, the degree of disorder in the cutting state is calculated. The formula for its calculation is:

[0044]

[0045] In the formula, The degree of disorder in the cutting state; The first one in the current sliding window Power spectral density of each spectral component; This is the sum of the power spectral densities of all spectral components within the current sliding window; This represents the total number of spectrum statistical intervals. It is the natural logarithm function.

[0046] The calculation formula utilizes the entropy weight method to measure the uniformity of cutting energy distribution in the frequency domain. When the tool is in a healthy state, the cutting vibration exhibits significant periodicity, with energy concentrated near a specific fundamental frequency and its harmonics. The distribution is extremely uneven, leading to a disordered cutting state. At a relatively low level; when tool wear or chipping leads to instability in the cutting process, the enhanced high-frequency random vibration causes energy to diffuse across the entire frequency band, with the energy proportion of each frequency band tending to be equal, thereby driving the degree of turbulence in the cutting state. The value of increases significantly, which indicates that the disorder of the cutting signal structure is intensified.

[0047] S3. Nonlinearly couple the time-domain gradient of the cutting power consumption characteristics with the cutting state disorder to calculate the wear instability coefficient.

[0048] It should be noted that, due to the uneven internal hardness during the machining of mold steel, instantaneous cutting load spikes will be generated. In order to avoid misjudging them as wear, this invention constructs a coupled model containing nonlinear gating compensation coefficients. It utilizes the physical difference between the irreversible frequency domain disorder caused by tool wear and the quasi-static energy change caused by material hardness fluctuations to correct the monitoring indicators, thereby achieving accurate locking of the actual wear state.

[0049] Specifically, obtain the cutting power consumption characteristics within the current sliding window. The rate of change over time is defined as the power consumption gradient term; simultaneously, the cutting state disorder calculated at the current moment is obtained. .

[0050] Furthermore, the wear loss stability coefficient was calculated. The formula for its calculation is:

[0051]

[0052] In the formula, The coefficient for stable wear loss; This is the system sensitivity adjustment factor, measured in seconds per watt (s / W), used to adjust the sensitivity of the system. Perform dimensional cancellation; This represents the rate of change of cutting power consumption characteristics over time within the current sliding window; The degree of disorder in the cutting state is calculated in real time; The average cutting state disorder during the initial reference period of machining; These are the nonlinear gated compensation coefficients; It is the hyperbolic tangent function; It is a function for maximizing the value.

[0053] Wherein, the system sensitivity adjustment factor It is a property parameter used to balance the dimensions of calculations and adjust the system's response speed to failure characteristics; since the first derivative of cutting power consumption is much larger in magnitude than the degree of disorder in the cutting state, this factor is set to normalize the wear loss instability coefficient. The value is kept within a preset statistical monitoring range. If this value is set too small, the wear instability coefficient will be insensitive to energy fluctuations caused by early minor chipping, resulting in missed detections. If it is set too large, it will amplify the random interference of background Gaussian noise, causing the monitoring indicators to frequently touch the statistical boundary during the healthy processing stage, triggering false alarms. Therefore, its value range is set to 0.0001 to 0.0010. In this embodiment, the average background energy gradient under healthy conditions is collected at the beginning of the processing stage to keep the wear instability coefficient benchmark value within a certain range. Based on the order of magnitude, the factor is set to 0.0003 to ensure both the robustness of the judgment result and the sensitivity of the response. In other embodiments, the implementer can make flexible fine adjustments according to the sampling frequency of the data acquisition system and the damping characteristics of the machine tool transmission chain: when the sampling frequency is increased, the value of the factor should be appropriately reduced to filter high-frequency glitch interference.

[0054] Wherein, the nonlinear gated compensation coefficient This is a key parameter used to balance the contribution weight of power growth rate and signal disorder to wear determination. If this parameter is set too small, the system's ability to capture signal frequency dispersion will be insufficient, leading to a lag in the early warning response to severe wear stages. If this parameter is set too large, the system will be overly sensitive to vibrations during normal cutting processes, resulting in a wear instability coefficient. Significant fluctuations occur during normal processing; therefore, the nonlinear gating compensation coefficient... The value range is [2.0, 5.0], and in this embodiment, it is preferably 3.5; in other embodiments, the implementer can adjust it within this range according to the structural rigidity and damping characteristics of the machining center.

[0055] The calculation formula utilizes the nonlinear mapping properties of the hyperbolic tangent function to construct differential gated logic; when the tool cuts into a high-hardness region, a power consumption gradient is generated. When the instantaneous increase occurs, if the cutting state disorder... No deviation from the benchmark level ,but The output of the term is extremely small, which forcibly suppresses the wear loss instability coefficient. The amplitude is thus immune to hardness interference; the wear instability coefficient only decreases when the power consumption gradient and signal structure disorder occur synchronously. This will lead to exponential growth, and the larger the value, the higher the probability that the tool will enter a failure state.

[0056] For example, Figure 3 The figure shows the distribution of cutting state disorder as the tool wear state evolves. The figure illustrates the evolution trajectory of the frequency domain energy distribution entropy value calculated in step S2. The step-like curve in the figure represents the cutting state disorder: in the healthy stage of early machining and the hardness fluctuation range, the curve remains in a low and stable state, proving that the change in pure energy amplitude does not lead to a qualitative change in the signal spectrum structure. However, in the wear evolution stage after 180s, due to the injection of high-frequency chatter signal, the curve produces a significant nonlinear rise, characterizing the physical process of the signal changing from order to disorder.

[0057] S4. Establish a dynamic statistical reference surface and provide tool status warning based on the degree of deviation between the wear instability coefficient and the reference surface.

[0058] It should be noted that, due to the inherent physical differences between different models of machine tools and different batches of mold steel workpieces, a single fixed threshold cannot take into account the detection sensitivity under all working conditions, which can easily lead to missed detections or false alarms. Therefore, this invention establishes a dynamic statistical reference surface in the early stage of processing, obtains the statistical feature envelope of the health status under the current working conditions, and combines it with a confidence factor including a safety factor to determine the final warning triggering time, thereby ensuring the robustness of the monitoring system under complex working conditions.

[0059] Specifically, the system collects a data sequence of 50 sets of wear loss instability coefficients during the initial processing stage, and calculates the mean value of the wear loss instability coefficients during the baseline period based on this data sequence. and the standard deviation of the wear loss stability coefficient during the baseline period. .

[0060] Furthermore, an early warning discrimination criterion is set, and the discriminant for the dynamic discrimination boundary is: In the formula, The wear loss stability coefficient is for real-time monitoring; This represents the average wear loss stability coefficient during the baseline period. The standard deviation of the wear loss stability coefficient during the baseline period; This is the confidence factor.

[0061] Wherein, the confidence factor The tolerance boundary of the early warning system is determined; its calibration logic is as follows: Continuous cutting experiments are conducted under standard process conditions, and the wear on the tool flank is monitored using a metallographic microscope. When the wear reaches the failure critical value of 0.3 mm, the wear instability coefficient and mean value at this point are recorded. The discrete multiples are determined, and a safety factor of 1.2 is introduced to cover the effects of different cutting fluid cooling efficiencies.

[0062] It should be noted that by establishing a dynamic statistical reference surface in the early stage of processing and constructing a dynamic discrimination boundary using confidence factors, the present invention enables the system to automatically adjust the discrimination threshold according to the actual cutting characteristics of each group of workpieces. This ensures accurate capture of the physical event of the tool entering the failure stage while eliminating interference from individual differences, thus achieving reliable monitoring of the precision machining process.

[0063] It should be further noted that during the precision machining of mold steel, once the wear instability coefficient exceeds the safety threshold, it means that the tool has entered a state of severe wear or chipping. If the machining task is not interrupted immediately, it will lead to the workpiece surface accuracy exceeding the tolerance or even cause mechanical damage to the spindle. Therefore, this invention establishes a real-time response mechanism from the digital signal calculation layer to the hardware execution control layer. Through instruction linkage, it ensures that the system can cut off the feed path within milliseconds and simultaneously mobilize the sensing terminal to notify the operator.

[0064] Specifically, the system compares the currently calculated wear instability coefficient with the dynamic discrimination boundary in real time; when it detects that the real-time monitored wear instability coefficient is greater than the dynamic discrimination boundary, the system determines that the tool failure risk is established.

[0065] Furthermore, the system sends a high-priority external interrupt stop command to the machining center controller via the industrial Ethernet communication protocol. Upon receiving the stop command, the machining center controller immediately executes the feed hold operation of the current CNC program, causing the feed axis motor to decelerate rapidly to a stop and locking the spindle speed to prevent secondary collisions between the tool and the workpiece. Simultaneously, the system drives the audible and visual alarm terminal through the input / output interface, triggering a high-frequency buzzer to sound and a red warning light to flash, prompting the operator to perform a tool change.

[0066] It should be further added that, in order to ensure the traceability of the early warning process, the system automatically captures and stores the original sampled current, cutting power consumption characteristics and corresponding cutting state disorder data in the current window when triggering the shutdown command. The above information is then written into the system fault log after being associated with the timestamp, providing data support for the optimization of subsequent processing technology.

[0067] It should be noted that this invention directly maps the logical judgment result of the wear loss instability coefficient to the hardware action of the machining center controller, and supplements it with real-time prompts from the audible and visual alarm terminal, thus realizing closed-loop control from implicit data monitoring to explicit safety protection. This not only avoids the risk of precision mold scrapping caused by the lag of manual inspection, but also ensures the operational safety of the core precision components of the machining center through an automatic shutdown mechanism.

[0068] For example, Figure 4 This is a diagram showing the early warning effect of the wear instability coefficient based on the nonlinear entropy control suppression model. The diagram illustrates the evolution characteristics of the final judgment index after reconstruction in step S3. The solid line at the bottom of the diagram represents the wear instability coefficient, and the horizontally arranged chain-like dashed lines represent the dynamic statistical threshold set by this invention. During the hardness fluctuation period at 100s, due to the truncation effect of the nonlinear entropy control gating operator, the wear instability coefficient curve remains below the threshold line, successfully eliminating false interference. In the actual wear range after 180s, the curve undergoes an exponential transition under entropy increase triggering. The position marked by the dots in the diagram is the precise early warning trigger point locked by the system through statistical discrimination criteria.

Claims

1. A method for early warning of tool wear status in machining centers based on multi-source data fusion, characterized in that, include: The original sampled current and real-time feed rate of the spindle are synchronously acquired from the machining center controller bus in real time. The cutting power consumption characteristics are calculated through the spindle load energy mapping model, satisfying the following: ； This refers to the power consumption characteristics of cutting. Main spindle input voltage; This is the original sampling current; The motor transmission efficiency coefficient is obtained from the motor nameplate. The real-time power factor of the main spindle is obtained by reading specific system variables or PLC registers built into the system. For real-time feed rate; For machine tools at real-time feed rate The no-load power loss; The cutting power consumption characteristics are sampled using a sliding window. The cutting state disorder is calculated based on the uniformity of the signal distribution in the frequency domain within the sliding window, satisfying the following conditions: ； The degree of disorder in the cutting state; The first one in the current sliding window Power spectral density of each spectral component; This is the sum of the power spectral densities of all spectral components within the current sliding window; This represents the total number of spectrum statistical intervals. It is the natural logarithm function; The rate of change of the cutting power consumption characteristic over time within the current sliding window is obtained as the power consumption gradient term. The power consumption gradient term is nonlinearly coupled with the cutting state disorder to calculate the wear instability coefficient, satisfying the following: ； The coefficient for wear loss stability; This is the system sensitivity adjustment factor; This represents the rate of change of cutting power consumption characteristics over time within the current sliding window; The average cutting state disorder during the initial reference period of machining; These are nonlinear gated compensation coefficients; It is the hyperbolic tangent function; It is a function for maximizing the value; The wear instability coefficient sequence at the beginning of the machining process is collected to establish a dynamic statistical reference surface. Tool status warnings are given based on the deviation of the real-time monitored wear instability coefficient from the dynamic statistical reference surface.

2. The method for early warning of tool wear status in machining centers based on multi-source data fusion according to claim 1, characterized in that, Before calculating the cutting power consumption characteristics through the spindle load energy mapping model, the process includes: measuring the no-load power data of the machine tool under different speed gradients; and fitting the no-load power consumption function using the least squares method.

3. The method for early warning of tool wear status in machining centers based on multi-source data fusion according to claim 1, characterized in that, The establishment of the dynamic statistical reference surface includes: Obtain the mean and standard deviation of the wear loss stability coefficient during the baseline period; A dynamic discrimination boundary is constructed based on the mean and standard deviation of the wear loss instability coefficient during the baseline period.

4. The method for early warning of tool wear status in machining centers based on multi-source data fusion according to claim 3, characterized in that, The expression for the dynamic discrimination boundary is: ； In the formula, The wear loss stability coefficient is for real-time monitoring; This represents the average wear loss stability coefficient during the baseline period. The standard deviation of the wear loss stability coefficient during the baseline period; This is the confidence factor.

5. The method for early warning of tool wear status in machining centers based on multi-source data fusion according to claim 1, characterized in that, The sliding window sampling has a length of 1024 sampling points and a step size of 256 sampling points.

6. The method for early warning of tool wear status in machining centers based on multi-source data fusion according to claim 4, characterized in that, The calibration process of the confidence factor includes: continuous cutting under standard process conditions, monitoring the wear amount on the tool flank, recording the dispersion multiple of the wear instability coefficient and the mean value of the wear instability coefficient in the reference period when the wear amount reaches the critical value, and introducing a safety factor to cover the influence of cooling efficiency.

7. The method for early warning of tool wear status in machining centers based on multi-source data fusion according to claim 1, characterized in that, The tool status early warning includes: when the real-time monitored wear instability coefficient meets the discrimination criteria, the machining center controller issues a stop command and simultaneously triggers the audible and visual alarm terminal.

Images