A method for automatically switching control of machining process parameters

Abstract

The application discloses a kind of machining process parameter automatic switching control methods, belong to machining parameter control technical field.The method is first defined and initialized and includes the state vector of cutting force, flutter probability and tool wear degree, then obtains multichannel sensing data and decouples into independent physical source component.Signal quality index is calculated and physical source component is adaptively denoised, and effective cutting force component is obtained.Effective cutting force component is decomposed into instantaneous component and trend component, and state variable is updated using probability iterative correction model, and state vector containing state confidence is obtained.Finally, state deviation is calculated and dynamic deviation threshold is adjusted according to state confidence, and if state deviation exceeds corresponding dynamic deviation threshold, process parameter is switched.The application realizes the adaptive switching of process parameter by matching dynamic threshold with working condition, significantly improves the stability and intelligent level of machining process under variable working condition.

Description

Technical Field

[0001] This invention relates to the field of machining parameter control technology, and in particular to an automatic switching control method for machining process parameters. Background Technology

[0002] In CNC milling, turning, and gear hobbing processes, process parameters such as spindle speed, feed rate, and feed speed directly affect machining efficiency, quality, and tool life. Currently, the determination of process parameters mainly relies on two approaches: one is to set fixed parameters for production based on process manuals or operational experience; the other is to establish a static mapping relationship between process parameters and machining targets through surrogate model methods such as orthogonal experiments, response surface methodology, or neural networks, and to obtain a set of fixed parameters for actual machining. However, in actual machining, tool wear accumulates continuously, and cutting forces and chatter fluctuate in real time. The above-mentioned static optimization strategies are difficult to adapt to the dynamic time-varying factors in the machining process, causing the preset parameters to gradually deviate from the optimal range, resulting in decreased machining stability and quality deterioration. Existing control strategies lack the closed-loop control capability to sense the machining status online and actively switch process parameters.

[0003] For example, Chinese patent application CN109407614A discloses a method for optimizing the process parameters of CNC gear hobbing machines. This method selects hob rotation speed, feed rate, and feed speed as research objects, uses spindle motor current to characterize the motor's machining capability, and establishes a multi-dimensional grey model for parameter optimization. However, the current signal it relies on is an indirect physical quantity, making it difficult to accurately perceive core state information such as instantaneous changes in cutting force and chatter probability. Furthermore, the obtained optimal parameters are fixed values, unable to automatically switch to a better parameter combination online as tool wear intensifies or cutting conditions change abruptly. Moreover, it lacks a comprehensive state assessment and dynamic threshold determination system that includes multiple variables such as cutting force, chatter probability, and tool wear degree, making it difficult to achieve early warning of machining state degradation and proactive parameter control. Therefore, this method needs further improvement. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides an automatic switching control method for machining process parameters. This invention improves the accuracy of cutting force acquisition by decoupling multi-channel signals and adaptive noise reduction, and combines probabilistic iterative estimation and dynamic deviation threshold to achieve working condition discrimination and automatic switching of process parameters, forming a closed-loop control, which effectively improves the stability and intelligence level of the machining process under varying working conditions.

[0005] The objective of this invention can be achieved through the following technical means: An automatic switching control method for machining process parameters includes the following steps: Step 1: Define and initialize the state vector, where the state variables are cutting force, chatter probability, and tool wear degree; Step 2: Acquire multi-channel sensor data in real time during the machining process, and use a signal decoupling mechanism to separate the multi-channel sensor data into multiple independent physical source components. The physical source components include at least cutting force source components and noise source components. Step 3: Calculate the signal quality index based on the physical source components, adjust the noise suppression intensity according to the signal quality index, and denoise the cutting force source components based on the noise suppression intensity to obtain the effective cutting force components; Step 4: Decompose the effective cutting force component into instantaneous component and trend component; Step 5: Use the probabilistic iterative correction model to update the state variables by combining the effective cutting force component, instantaneous component and trend component to obtain the state vector containing the state confidence at the current moment; Step 6: Calculate the state deviation between each state variable and the target state variable at the current time, and adjust the dynamic deviation threshold of each state variable according to the state confidence. Step 7: If any state deviation exceeds the corresponding dynamic deviation threshold, match the corresponding operating condition level according to the distribution characteristics of the current state deviation, switch the process parameters to the process parameter combination corresponding to the operating condition level, and return to Step 2; otherwise, keep the current process parameters unchanged and return to Step 2.

[0006] In this invention, in step 2, the multi-channel sensing data includes at least two of the following: current signal, vibration signal, and acoustic emission signal.

[0007] In this invention, in step 2, the signal decoupling mechanism adopts a blind source separation algorithm, specifically including: establishing a mixed signal model X=AS, where X is multi-channel sensing data, S is physical source components, A is a mixing matrix, estimating the inverse matrix W of the mixing matrix A by maximizing the statistical independence criterion, and reconstructing mutually independent physical source components S=WX. The cutting force source component in the physical source components corresponds to an independent signal source related to the spindle load, and the noise source component corresponds to an independent signal source caused by environmental vibration or sensor thermal noise.

[0008] In this invention, in step 3, the signal-to-noise ratio and spectral entropy value of the cutting force source component are calculated based on the noise source component. The signal-to-noise ratio and spectral entropy value are normalized and weighted and summed to obtain the signal quality index. When the signal quality index is lower than the preset strength threshold, the noise suppression strength is increased. When the signal quality index is higher than the preset strength threshold, the noise suppression strength is decreased.

[0009] In this invention, in step 4, wavelet packet transform or multi-scale decomposition algorithm is used to decompose the effective cutting force component into sub-band signals of different frequency bands, reconstruct the high-frequency sub-band signals into instantaneous components, and use the low-frequency sub-band signals or sub-band signals after moving average filtering as trend components.

[0010] In this invention, in step 5, the probabilistic iterative correction model includes a state transition equation and an observation equation. The state transition equation describes the evolution of cutting force, chatter probability, and tool wear degree, and the observation equation describes the mapping relationship between the effective cutting force component, instantaneous component, and trend component and each state variable.

[0011] In this invention, in step 5, the state vector of the previous moment is used as the initial basis. The prior estimates of each state variable and the corresponding uncertainty range are predicted by the state transition equation. The effective cutting force component, instantaneous component and trend component are used as the input of the observation equation. The correction weight of each state variable is calculated. The prior estimates are weighted and fused based on the correction weight to obtain the optimal estimates of each state variable at the current moment. The statistic of the uncertainty range is used as the corresponding state confidence.

[0012] In this invention, in step 6, a baseline deviation threshold Δ0 is set. For each state variable, its corresponding state confidence C is extracted, and the dynamic deviation threshold Δ1 = Δ0 + K(1-C) is calculated, where K is the adjustment coefficient and C ∈ [0,1].

[0013] In this invention, in step 7, the working condition level includes at least a stable machining condition, a slight vibration condition, a chatter critical condition, and a severe tool wear condition. Each working condition level corresponds to a set of preset process parameter combinations. Each set of process parameter combinations includes the set values ​​of spindle speed, feed rate, depth of cut, and tool geometric compensation.

[0014] In this invention, in step 7, the distribution characteristics of each state deviation include the magnitude and relative proportion of the cutting force deviation, chatter probability deviation, and tool wear deviation. A weighted comprehensive evaluation value of the cutting force deviation, chatter probability deviation, and tool wear deviation is calculated, wherein the weight coefficient of each state deviation is adjusted according to the relative proportion. The weighted comprehensive evaluation value is compared with the preset threshold range corresponding to each working condition level to determine the current working condition level.

[0015] The automatic switching control method for machining process parameters of this invention has the following advantages: This invention uses a multi-channel signal decoupling mechanism to separate real-time acquired multi-channel sensor data into independent cutting force source components and noise source components, eliminating cross-interference at the source. By calculating the signal quality index and dynamically adjusting the noise suppression intensity, details can be preserved when the signal is pure and noise can be eliminated when the signal is noisy, ensuring the authenticity and reliability of the extracted effective cutting force components. Based on this, the effective cutting force components are decomposed into instantaneous and trend components, and a probabilistic iterative correction model is used to output a state vector containing state confidence in real time, improving the accuracy of state estimation. When the detected state deviation exceeds the dynamic deviation threshold dynamically set based on state confidence, the corresponding working condition level is matched according to the distribution characteristics of the state deviation, and the system automatically switches to the corresponding process parameter combination, forming a complete closed-loop control from signal acquisition and state estimation to parameter adaptation. This invention effectively solves the problems of misjudgment and insufficient adaptability caused by signal distortion and fixed parameters in traditional methods, significantly improving the stability and intelligence level of the machining process under varying working conditions. Attached Figure Description

[0016] Figure 1 This is a flowchart of an automatic switching control method for machining process parameters according to the present invention; Figure 2 This is a diagram of the original time-domain waveform of a vibration signal according to the present invention. Figure 3 This is the original time-domain waveform of another vibration signal according to the present invention; Figure 4 This is a diagram of the original time-domain waveform of an acoustic emission signal according to the present invention. Figure 5 This is the original time-domain waveform diagram of another acoustic emission signal according to the present invention; Figure 6 This is a graph showing the amplitude variation of the vibration signal after envelope demodulation. Figure 7 This is a graph showing the amplitude variation of the acoustic emission signal after envelope demodulation. Figure 8 This is a graph showing the effective cutting force components of the present invention. Figure 9 This is a partial schematic diagram of the process planning parameters of the present invention; Figure 10 This is a partial schematic diagram of the tool inventory data of the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0018] In CNC machining, key parameters affecting the machining state are difficult to measure directly online and are usually estimated using indirect signals or offline experience. However, during data sensing, the raw data acquired by sensors is easily overwhelmed by strong noise such as cutting fluid splashes, machine tool vibrations, and electromagnetic interference, resulting in an extremely low signal-to-noise ratio. This makes it difficult to accurately identify state characteristics such as chatter precursors or load changes from such a complex noise background. Furthermore, the ideal measuring point is located at the tool tip or the tool itself. Although force gauges can directly measure cutting forces, they are expensive and affect machine tool rigidity, making them difficult to widely apply in industrial settings. Therefore, most existing methods rely on single signals such as spindle current to indirectly reflect the machining state, or establish static models through offline experiments. This makes it difficult to achieve real-time and accurate perception of the machining state, and the accuracy and response speed of state estimation are difficult to guarantee when machining conditions change.

[0019] This invention acquires multi-channel sensor data in real time, extracts the effective cutting force signal through signal decoupling and adaptive denoising, and decomposes it into instantaneous and trend components. These components are used to update the cutting force, chatter probability, and tool wear degree in the state vector. Based on the updated state vector, the state deviation between the current state vector and the target state vector is calculated. A dynamic deviation threshold is used to determine whether the machining state has deviated. If a deviation is found, the working condition level is matched according to the deviation distribution characteristics, and the corresponding process parameter combination is automatically switched to form a closed-loop control. This invention achieves real-time accurate estimation of the machining state and adaptive switching of process parameters, significantly improving the stability and intelligence level of the machining process under varying working conditions. Example 1

[0020] Reference Figure 1 The automatic switching control method for machining process parameters of the present invention, as detailed in this embodiment, includes the following steps: Step 1: Define and initialize the state vector, where the state variables are cutting force, chatter probability, and tool wear degree. Let the state vector at time k be S. k ={(F k C k,F ), (P k C k,P ), (W k C k,W )}, where F k Let P be the cutting force at time k. k Let W be the flutter probability at time k. k Let C be the degree of tool wear at time k. k,F C k,P C k,WThe state confidence values ​​at time k represent the cutting force, chatter probability, and tool wear degree, respectively. All state variable values ​​are estimated values. State confidence is an index that quantifies the credibility range and reliability level of the current state variable estimate, characterizing the degree to which the estimated value of the corresponding state variable is affected by noise interference or model error. Its value range is [0,1]. A state confidence of 0 indicates completely unreliable, while a state confidence of 1 indicates absolutely reliable. In this embodiment, the tool is an old tool that has been used for 45 minutes, and the initial state vector S0 = {(1900N, 0.72), (0.15, 0.65), (0.375, 0.95)}. The preferred method for initializing the state vector is described in Embodiment 2.

[0021] Step 2: Acquire multi-channel sensor data in real time during the machining process. Use a signal decoupling mechanism to separate the multi-channel sensor data into multiple independent physical source components. These physical source components include at least a cutting force source component and a noise source component. The multi-channel sensor data includes at least two of the following: current signal, vibration signal, and acoustic emission signal. However, vibration signals are sensitive to changes in cutting force and are one of the preferred signals for acquiring the cutting force source component; therefore, multi-channel sensor data typically includes vibration signals. For example... Figures 2 to 7 In this embodiment, the multi-channel sensing data includes vibration signals and acoustic emission signals. Figure 2 This is the original time-domain waveform of the vibration signal at a spindle speed of 300 rpm in this embodiment. Figure 3 The original time-domain waveform of the vibration signal at a spindle speed of 800 rpm. Figure 4 This is the original time-domain waveform of the acoustic emission signal when the cutting depth is 0.5 mm in this embodiment. Figure 5 This is the original time-domain waveform of the acoustic emission signal when the cutting depth is 0.8 mm in this embodiment. Since the original time-domain waveform contains a large number of high-frequency oscillation components, it is easy to introduce interference when used directly for state analysis. In this embodiment, the vibration signal and the acoustic emission signal are respectively subjected to envelope demodulation processing to extract the intensity envelope of the periodic impact event in the signal, which facilitates the quantitative analysis of flutter and wear characteristics in subsequent steps. Figure 6 This is a graph showing the amplitude variation of the vibration signal after envelope demodulation. Figure 7 This is a graph showing the amplitude variation of the acoustic emission signal after envelope demodulation.

[0022] Vibration signals are acquired by an accelerometer (such as the 1A110E type) and its matching acquisition system (such as the DH5922N type), installed on the spindle housing or tool holder near the cutting point. Acoustic emission signals are acquired by an acoustic emission sensor (such as the VS150-R type), installed near the tool holder or workpiece. Vibration signals are sensitive to chatter response, while acoustic emission signals respond earlier to chatter precursors and tool micro-wear. The combination of the two can achieve early warning of chatter and real-time monitoring of tool wear.

[0023] In another embodiment, to reduce hardware costs, the multi-channel sensing data is a combination of vibration and current signals, where the current signal is used to assist in determining the cutting load trend. In yet another embodiment, if vibration sensors cannot be installed due to site conditions (e.g., no installation location for ultra-high-speed spindles), the multi-channel sensing data can use a combination of acoustic emission and current signals. Compared to embodiments where the multi-channel sensing data includes vibration signals, this configuration has a lower accuracy in quantitatively estimating chatter, but it can still reflect the relative trend of cutting force changes. It is suitable for industrial sites where vibration sensors cannot be installed and where machining efficiency requirements are not extreme.

[0024] The signal decoupling mechanism employs a blind source separation algorithm, specifically including: establishing a mixed signal model X=AS, where X represents multi-channel sensor data, S represents physical source components, and A is a mixing matrix. Both X and S are in matrix form, with each row of X corresponding to a sensor channel and each column representing the amplitude of each sensor channel at the same sampling time. Although the mixing matrix A and the physical source components S are unknown, blind estimation can be performed based on the statistical independence between the physical source components. Specifically, by optimizing the separation matrix W, the statistical independence between the components of Y=WX is maximized, thereby estimating the inverse matrix W of the mixing matrix A. At this point, Y approaches the true physical source component S, thus reconstructing the mutually independent physical source components S=WX. After separation, frequency domain analysis is performed on each mutually independent physical source component using a fast Fourier transform. Components whose chatter frequency matches the theoretical cutting frequency (which can be determined based on the current spindle speed and the number of tool teeth) are identified as cutting force source components, while the remaining physical source components are identified as noise source components. The cutting force source component corresponds to an independent signal source related to the spindle load, while the noise source component corresponds to an independent signal source caused by environmental vibration or sensor thermal noise.

[0025] Step 3: Calculate the signal quality index based on the physical source components, adjust the noise suppression intensity according to the signal quality index, and denoise the cutting force source components based on the noise suppression intensity to obtain the effective cutting force components. Calculate the signal-to-noise ratio (SNR) and spectral entropy values ​​of the cutting force source components based on the noise source components, normalize the SNR and spectral entropy values ​​respectively, and then sum them by weight to obtain the signal quality index. When the signal quality index is lower than a preset intensity threshold, increase the noise suppression intensity; when the signal quality index is higher than the preset intensity threshold, decrease the noise suppression intensity.

[0026] Reference Figure 8An adaptive filter (using the cutting force source component as the main input and the noise source component obtained from blind source separation as the reference input) is used to denoise the cutting force source component to obtain the effective cutting force component. The adaptive filter adopts the least mean square (LMS) adaptive filtering algorithm. The noise suppression intensity is used as the step size factor of the adaptive filter to control the balance between the convergence speed and steady-state error of the filter coefficients. The greater the noise suppression intensity, the larger the step size factor, the faster the filtering response speed, and the stronger the noise suppression capability.

[0027] Step 4: Decompose the effective cutting force component into instantaneous and trend components. Using wavelet packet transform or multi-scale decomposition algorithms, the effective cutting force component is decomposed into sub-band signals of different frequency bands. A high-frequency threshold is set based on the typical frequency range of chatter. In this embodiment, the typical frequency range of chatter is determined based on the current spindle speed n and the number of tool teeth z, specifically: chatter dominant frequency f c =nz / 60, the frequency band within ±20% of the dominant flutter frequency is considered the typical frequency range, and the frequency threshold above the upper limit of this typical frequency range is taken as the high-frequency threshold. Based on the high-frequency threshold, the sub-band signal is divided into high-frequency sub-band signal and low-frequency sub-band signal, and then the high-frequency sub-band signal is reconstructed into instantaneous components, and the low-frequency sub-band signal or the sub-band signal after moving average filtering is taken as the trend component.

[0028] The wavelet basis functions used in the wavelet packet transform can be dynamically adjusted according to the characteristics of the cutting material. In this embodiment, the Daubechies 5 wavelet basis function is selected for machining steel, the Coiflet 3 wavelet basis function is selected for machining aluminum alloys, and the Symlet 4 wavelet basis function is selected for machining cast iron. The correspondence between the above wavelet basis functions and the cutting materials is pre-stored in the machine tool CNC system. Before machining, the operator selects the basis function according to the workpiece material, or the system automatically matches it according to the material grade.

[0029] Step 5: Using a probabilistic iterative correction model, update the state variables by combining the effective cutting force component, instantaneous component, and trend component to obtain the state vector containing the state confidence at the current moment. The probabilistic iterative correction model includes a state transition equation and an observation equation. The state transition equation describes the evolution of cutting force, chatter probability, and tool wear, while the observation equation describes the mapping relationship between the effective cutting force component, instantaneous component, trend component, and each state variable. For a specific example of the optimal update method for the state vector, refer to Embodiment 3.

[0030] Using the state vector from the previous moment as the initial basis, the prior estimates of each state variable and the corresponding uncertainty range are predicted through the state transition equation. The effective cutting force component, instantaneous component, and trend component are used as inputs to the observation equation. The corrected weights of each state variable are calculated. The prior estimates are weighted and fused based on the corrected weights to obtain the optimal estimates of each state variable at the current moment. The statistics of the uncertainty range are used as the corresponding state confidence.

[0031] Step 6: Calculate the state deviation between each state variable and the target state variable at the current moment, and adjust the dynamic deviation threshold of each state variable according to the state confidence level. The state deviation is the difference between each current state variable and its corresponding target state variable. The target state variable is the expected state reference value when the machining quality meets the requirements, including the target cutting force F. target Target flutter probability P target and the wear degree of the target tool W target The target cutting force is preset by the process designer based on the workpiece machining accuracy requirements and tool specifications, and stored in the machine tool's CNC system. In this embodiment, the target machining is milling 45Cr steel, and the target cutting force is set to F. target =1500N, target flutter probability set to P target =0.05, the target tool wear level is set to W. target =0.30mm.

[0032] A baseline deviation threshold Δ0 is set. For each state variable, its corresponding state confidence C is extracted, and the dynamic deviation threshold Δ1 = Δ0 + K(1-C) is calculated, where K is the adjustment coefficient and C ∈ [0,1]. In this embodiment, the baseline deviation threshold Δ1 for the cutting force is... 0,F = 200 N, the reference deviation threshold Δ for the flutter probability 0,P = 0.05, the benchmark deviation threshold Δ for tool wear. 0,W = 0.05 mm. When the state confidence decreases, the dynamic deviation threshold is automatically widened to reduce the risk of false triggering.

[0033] Step 7: If any state deviation exceeds the corresponding dynamic deviation threshold, match the corresponding operating condition level according to the distribution characteristics of the current state deviations, switch the process parameters to the process parameter combination corresponding to that operating condition level, and return to Step 2; otherwise, keep the current process parameters unchanged and return to Step 2. The operating condition levels include at least stable machining conditions, slight vibration conditions, chatter critical conditions, and severe tool wear conditions. Each operating condition level corresponds to a set of preset process parameter combinations. Each set of process parameter combinations includes the set values ​​of spindle speed, feed rate, depth of cut, and tool geometry compensation.

[0034] The distribution characteristics of each state deviation amount include the numerical magnitudes and relative proportional relationships of the cutting force deviation amount, the chatter probability deviation amount, and the tool wear deviation amount. Normalize each state deviation amount and convert it into a dimensionless normalized cutting force deviation amount ΔF, a normalized chatter probability deviation amount ΔP, and a normalized tool wear deviation amount ΔW. Then calculate the weighted comprehensive evaluation value Q = w1ΔF + w2ΔP + w3ΔW. Here, w1, w2, and w3 are the weight coefficients of the cutting force deviation amount, the chatter probability deviation amount, and the tool wear deviation amount respectively. The weight coefficients of each state deviation amount are adjusted according to the relative proportional relationship. The relative proportional relationship is the proportion of the normalized value of each state deviation amount in the normalized total. The larger the proportion, the larger the corresponding weight coefficient, making the weighted comprehensive evaluation value tend to the current most significant over-limit state to improve the sensitivity of working condition discrimination. Compare the weighted comprehensive evaluation value with the preset threshold intervals corresponding to each working condition level to determine the current working condition level. In this embodiment, the preset threshold intervals are as follows: when Q ≤ 0.3, it corresponds to a stable machining working condition; when 0.3 < Q ≤ 0.6, it corresponds to a slight vibration working condition; when 0.6 < Q ≤ 0.8, it corresponds to a chatter critical working condition; when 0.8 < Q ≤ 1.0, it corresponds to a serious tool wear working condition. Embodiment 2

[0035] The initialization accuracy of the state vector directly determines the separation purity of the subsequent signal decoupling mechanism, the adjustment of the noise suppression intensity, and the convergence speed of the probability iteration correction model, thus affecting the reliability of the final parameter control. Therefore, this embodiment further discloses an optimized method for initializing the state vector in step 1 to ensure that the cutting force, the chatter probability, the tool wear degree, and their corresponding state confidence levels have benchmark values that conform to physical reality at the beginning.

[0036] Refer to Figure 9 and Figure 10 , before the start of machining (i.e., at time k = 0), obtain the process planning parameters and the tool inventory data. The process planning parameters are a set of standard machining conditions determined by process personnel or CAM software before the start of machining. The tool inventory data is the current available tool information, including the attributes of the tool itself, the life status, and the spare parts situation. The initial state vector S0 = {(F0, C 0,F ), (P0, C 0,P ), (W0, C 0,W )}, at this time, the state vector S0 has not been assigned values and will be calculated separately later.

[0037] Calculation of cutting force F0. The initial estimated value of the cutting force F0 is calculated based on the theoretical cutting force model: F0 = K1a1fcos(θ1-θ2) / [sinθ1cos(θ1+θ3-θ2)], where K1 is the unit cutting force coefficient, a1 is the depth of cut, f is the feed rate, θ1 is the shear angle (derived from cutting force experiments or estimated using empirical formulas), θ2 is the tool rake angle, and θ3 is the friction angle. The unit cutting force coefficient is the cutting force per square millimeter of cutting area (obtainable from a table based on the material to be processed). In this embodiment, the workpiece material is 45Cr, and it is in an unheated state, with a unit cutting force coefficient K1 of 2200 N / mm. 2 The depth of cut a1, feed rate f, and tool rake angle θ2 can be directly obtained from the process planning parameters. The friction angle θ3 is calculated based on the friction coefficient between the carbide tool (used in this embodiment) and the workpiece material (friction coefficient is, for example, 0.58), i.e., θ3 = arctan(0.58). In this embodiment, the shear angle θ1 is estimated based on Merchant's empirical formula, and θ1 is, for example, 34°. The theoretical calculated value of F0 is 1896.7N. Since the input parameters of the theoretical cutting force model itself have slight uncertainties, and the initial state variables do not need to be accurate to the decimal point, for the convenience of engineering application and calculation efficiency, this embodiment rounds the theoretical calculated value of the cutting force F0 to the nearest integer, so F0 is taken as 1900N.

[0038] State confidence C 0,F The calculation of the theoretical cutting force model is performed. The uncertainties of the six input parameters (K1, a1, f, θ1, θ2, θ3) are identified: the uncertainty of K1 stems from the discreteness of material property difference data; the uncertainties of a1 and f stem from the positioning accuracy of the machine tool CNC system and the dynamic following error of the feed axis; the uncertainty of θ1 stems from the tool setting error during installation; the uncertainty of θ2 stems from the approximation error of empirical formulas or back-calculation algorithms under specific working conditions; and the uncertainty of θ3 stems from the applicability deviation of friction angle lookup table data. Using sensitivity analysis, the influence weights of the above six input parameters on F0 are calculated. The influence weights of K1, a1, f, θ1, θ2, and θ3 are 0.35, 0.25, 0.2, 0.1, 0.05, and 0.05, respectively. Based on the influence weights and uncertainties, the total cutting uncertainty is calculated as 0.0415. The total cutting uncertainty is mapped to the [0,1] interval to obtain the state confidence C. 0,F =0.72. If the input parameters are all taken from a high-precision standard library and the model has a high degree of matching, the uncertainty is higher, and the state confidence C is higher. 0,F It also approaches 1; conversely, the state confidence C... k,F The corresponding reduction.

[0039] Calculation of chatter probability P0. The initial value of chatter probability P0 is determined based on the stability lobe diagram of the machine tool-tool system: The initial spindle speed n0 is obtained, and the stable lobe region determination value Δ at that spindle speed is calculated using methods such as semi-discretization or frequency domain methods. n Δ n The sign of Δ indicates the degree to which the current rotational speed deviates from the stable region, and its absolute value reflects the deviation distance. Due to fluctuations in actual cutting conditions (such as uneven material hardness and instantaneous changes in cutting force), even if theoretical calculations determine the system to be stable or unstable, the inherent uncertainty of the system's stability margin must still be considered. Therefore, a critical threshold σ is introduced. σ characterizes the typical fluctuation range of the system's stability margin, typically taken as 5%-10% of the typical fluctuation range, and is used to define the boundaries of the stable region, critical region, and unstable region. If σ ≤ Δ... n If the spindle speed n0 is within the safe and stable blade region, and the chatter probability P0 = 0 is set, then Δ n If -σ ≤ -σ, then the spindle speed n0 is not within the safe and stable blade region. Set P0 = 1. If -σ < Δ n If σ < 0, then the spindle speed n0 is determined to be near the critical boundary, and a smooth transition function P0 = 0.5 - 0.5tanh(Δ) is used. n Calculate the flutter probability P0 using / σ).

[0040] State confidence C 0,P The calculation involves: Calculating the number of validated stable leaf blade samples N1 under similar historical cutting conditions; and then comparing this number with the minimum threshold N required to construct a reliable stability profile. min Perform a comparison and calculate the basic weighting factor w1=min(1, N1 / N). min Then, obtain the goodness-of-fit index (usually the coefficient of determination) of the current system modal parameters (e.g., damping ratio) during the identification process, and use it as the correction coefficient w2. Then, the state confidence C 0,P = w1w2, if there is not a sufficient number of stable leaflet samples (i.e., N1) <N min This can lead to an inability to effectively assess the state confidence level C. 0,P If so, then a conservative, low-confidence baseline value is directly taken as the state confidence level C. 0,P .

[0041] Calculation of tool wear degree W0. For brand new tools, set W0=0 directly. For inventory tools, read the historical usage record T1 and rated total life T2 from the tool management system, and set W0=W max T1 / T2, where W max This represents the maximum allowable wear amount for the corresponding tool model.

[0042] State confidence C0,W The calculation is based on the completeness and accuracy of historical usage records in the tool management system: if the read historical records cover the entire time span and there are no abnormal missing data, then the state confidence level C is set. 0,W The value is 1; if some records are missing, a reduction factor α is introduced based on the proportion of missing data, and the state confidence level C is 1. 0,W =α. Example 3

[0043] This embodiment further discloses a preferred method for updating state variables using a probabilistic iterative correction model in step 5.

[0044] A probabilistic iterative correction model is constructed. This model includes a state transition equation and an observation equation. The state transition equation predicts the state vector at the next moment, while the observation equation correlates the effective cutting force components, instantaneous components, and trend components with the various state variables. The state transition equation S... k|k-1 =DS k-1 +BU k +w k Observation equation Z k =HS k|k-1 +v k , among which, S k|k-1 Let S be the prior state vector at time k. k-1 Let U be the state vector at time k-1, D be the state transition matrix, and U be the state vector at time k-1. k w is an external input parameter during the processing. k Let B be the process noise, H be the control input matrix, and V be the observation matrix. k To observe the noise, Z k It is a theoretical observation vector composed of the theoretical effective cutting force component, the theoretical instantaneous component, and the theoretical trend component.

[0045] Calculate the prior state vector S at time k k|k-1 Based on the state vector of the previous time step (time step k-1), the prior state vector S of the current time step k is generated by deducing through the state transition equation. k|k-1 Then, using the constructed observation equation, the prior state vector S at time k is... k|k-1 As input, the theoretical effective cutting force component, theoretical instantaneous component, and theoretical trend component that should theoretically be observed at the current moment are reconstructed.

[0046] The state vector and state confidence at time k are calculated. The effective cutting force component, instantaneous component, and trend component observed at the current time are extracted, and then the theoretical effective cutting force component, theoretical instantaneous component, and theoretical trend component from the theoretical observation vector are subtracted from each, yielding the observation residuals. The observation residuals are multiplied by the Kalman gain to obtain the state vector correction increment, which is then superimposed on the prior state vector to obtain the updated state vector at time k. The Kalman gain is calculated from the covariance matrix of the prior state estimation and the observation noise covariance matrix. Based on the observation residuals, the state confidence of the corresponding state variables is generated; the smaller the observation residuals, the higher the state confidence; conversely, the larger the observation residuals, the lower the state confidence.

[0047] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for automatically switching and controlling machining process parameters, characterized in that, Includes the following steps: Step 1: Define and initialize the state vector, where the state variables are cutting force, chatter probability, and tool wear degree; Step 2: Acquire multi-channel sensor data in real time during the machining process, and use a signal decoupling mechanism to separate the multi-channel sensor data into multiple independent physical source components. The physical source components include at least cutting force source components and noise source components. Step 3: Calculate the signal quality index based on the physical source components, adjust the noise suppression intensity according to the signal quality index, and denoise the cutting force source components based on the noise suppression intensity to obtain the effective cutting force components; Step 4: Decompose the effective cutting force component into instantaneous component and trend component; Step 5: Use the probabilistic iterative correction model to update the state variables by combining the effective cutting force component, instantaneous component and trend component to obtain the state vector containing the state confidence at the current moment; Step 6: Calculate the state deviation between each state variable and the target state variable at the current time, and adjust the dynamic deviation threshold of each state variable according to the state confidence. Step 7: If any state deviation exceeds the corresponding dynamic deviation threshold, match the corresponding operating condition level according to the distribution characteristics of the current state deviation, switch the process parameters to the process parameter combination corresponding to the operating condition level, and return to Step 2; otherwise, keep the current process parameters unchanged and return to Step 2.

2. The automatic switching control method for machining process parameters according to claim 1, characterized in that, In step 2, the multi-channel sensing data includes at least two of the following: current signal, vibration signal, and acoustic emission signal.

3. The automatic switching control method for machining process parameters according to claim 1, characterized in that, In step 2, the signal decoupling mechanism adopts a blind source separation algorithm, which specifically includes: establishing a mixed signal model X=AS, where X is multi-channel sensing data, S is physical source components, and A is a mixing matrix; estimating the inverse matrix W of the mixing matrix A by maximizing the statistical independence criterion; reconstructing mutually independent physical source components S=WX; the cutting force source component in the physical source components corresponds to an independent signal source related to the spindle load, and the noise source component corresponds to an independent signal source caused by environmental vibration or sensor thermal noise.

4. The automatic switching control method for machining process parameters according to claim 1, characterized in that, In step 3, the signal-to-noise ratio and spectral entropy of the cutting force source component are calculated based on the noise source component. The signal-to-noise ratio and spectral entropy are normalized and weighted and summed to obtain the signal quality index. When the signal quality index is lower than the preset strength threshold, the noise suppression strength is increased; when the signal quality index is higher than the preset strength threshold, the noise suppression strength is decreased.

5. The automatic switching control method for machining process parameters according to claim 1, characterized in that, In step 4, wavelet packet transform or multi-scale decomposition algorithm is used to decompose the effective cutting force component into sub-band signals of different frequency bands. The high-frequency sub-band signals are reconstructed into instantaneous components, and the low-frequency sub-band signals or sub-band signals after moving average filtering are used as trend components.

6. The automatic switching control method for machining process parameters according to claim 1, characterized in that, In step 5, the probabilistic iterative correction model includes a state transition equation and an observation equation. The state transition equation describes the evolution of cutting force, chatter probability, and tool wear, while the observation equation describes the mapping relationship between the effective cutting force component, instantaneous component, and trend component and each state variable.

7. The automatic switching control method for machining process parameters according to claim 6, characterized in that, In step 5, the state vector of the previous moment is used as the initial basis. The prior estimates of each state variable and the corresponding uncertainty range are predicted by the state transition equation. The effective cutting force component, instantaneous component and trend component are used as the input of the observation equation. The corrected weights of each state variable are calculated. The prior estimates are weighted and fused based on the corrected weights to obtain the optimal estimates of each state variable at the current moment. The statistics of the uncertainty range are used as the corresponding state confidence.

8. The automatic switching control method for machining process parameters according to claim 1, characterized in that, In step 6, a baseline deviation threshold Δ0 is set. For each state variable, its corresponding state confidence C is extracted, and the dynamic deviation threshold Δ1 = Δ0 + K(1-C) is calculated, where K is the adjustment coefficient and C ∈ [0,1].

9. The automatic switching control method for machining process parameters according to claim 1, characterized in that, In step 7, the operating condition levels include at least stable machining condition, slight vibration condition, chatter critical condition and severe tool wear condition. Each operating condition level corresponds to a set of preset process parameter combinations. Each set of process parameter combinations includes the set values ​​of spindle speed, feed rate, depth of cut and tool geometric compensation.

10. The automatic switching control method for machining process parameters according to claim 1, characterized in that, In step 7, the distribution characteristics of each state deviation include the magnitude and relative proportion of the cutting force deviation, chatter probability deviation, and tool wear deviation. The weighted comprehensive evaluation value of the cutting force deviation, chatter probability deviation, and tool wear deviation is calculated, wherein the weight coefficient of each state deviation is adjusted according to the relative proportion. The weighted comprehensive evaluation value is compared with the preset threshold range corresponding to each working condition level to determine the current working condition level.

Images