Crankshaft connecting rod grinding force dynamic error compensation method and system

Abstract

The invention relates to the technical field of grinding machining, and particularly discloses a crankshaft connecting rod grinding force dynamic error compensation method and system.The method comprises the steps that a prediction reference signal is collected; a grinding force signal of the grinding wheel spindle is collected based on a piezoelectric sensor; extracting spectrum characteristic parameters according to the predicted reference signal; predicting grinding force fluctuation information according to the spectrum characteristic parameters; a feedforward compensation instruction is generated according to the grinding force fluctuation information; generating a feedback compensation instruction according to the difference between the grinding force signal and the target grinding force of the current task; synthesizing the feedforward compensation instruction and the feedback compensation instruction to generate a compensation control instruction; according to the method, a control strategy combining feedforward and feedback is adopted, the response speed and compensation precision of a system to dynamic changes of grinding force are remarkably improved, accordingly, elastic deformation caused by fluctuation of the grinding force is effectively restrained, and finally the grinding machining precision and surface quality of key parts of the crankshaft connecting rod are improved.

Description

Technical Field

[0001] The present application relates to the technical field of grinding machining, and more specifically, to a method and system for dynamically compensating the error of the grinding force of a crankshaft connecting rod. Background Art

[0002] As high-precision core parts of an automotive engine, the crankshaft and connecting rod need to achieve micron-level dimensional and geometric accuracy through precise control of the interaction force between the grinding wheel and the workpiece during grinding. With the development of grinding processes towards high efficiency and high stability, dynamic error compensation technology has become a key means to improve machining quality by real-time monitoring of the grinding force and adjusting the position of the grinding wheel. In the prior art, a closed-loop control system is constructed using a force sensor and a fast actuator, and significant results have been achieved in dealing with macroscopic wear of the grinding wheel or the dynamic characteristics of the machine tool servo system.

[0003] However, due to factors such as the microscopic hardness difference of the workpiece material and the local change of the grinding wheel abrasive state, the high-frequency fluctuation of the grinding force causes the existing compensation system to be limited by the dynamic response bandwidth and it is difficult to track the rapidly changing force signal in a timely manner. The instantaneous force fluctuation caused by this microscopic non-uniformity will cause dynamic elastic deformation between the grinding wheel and the workpiece, resulting in the actual cutting trajectory deviating from the preset path and forming a microscopic topography error. The traditional feedback compensation mechanism cannot effectively suppress the force fluctuation of the high-frequency component due to signal acquisition lag and actuator response delay, and the compensation accuracy is insufficient, which has become a technical bottleneck restricting the improvement of the machining accuracy of key parts of the crankshaft connecting rod.

[0004] In response to the above problems, there is currently no effective technical solution. Summary of the Invention

[0005] The purpose of the present application is to provide a method and system for dynamically compensating the error of the grinding force of a crankshaft connecting rod, which combines feedforward and feedback for compensation control to suppress the instantaneous force fluctuation caused by microscopic non-uniformity, reduce dynamic elastic deformation, improve microscopic topography accuracy and machining stability, reduce cutting trajectory deviation and microscopic topography error, thereby improving the grinding machining accuracy and surface quality of key parts of the crankshaft connecting rod.

[0006] In a first aspect, the present application provides a method for dynamically compensating the error of the grinding force of a crankshaft connecting rod for compensating the elastic deformation caused by the grinding force. The method includes the following steps: S1. Collect the acoustic emission signal of the current grinding area based on an acoustic emission sensor or collect the current signal of the grinding wheel spindle based on a high-speed spindle current sensor as a prediction reference signal; S2. Collect the grinding force signal of the grinding wheel spindle based on a piezoelectric sensor; S3. Extract spectral characteristic parameters according to the prediction reference signal; S4. Based on the online identification prediction model, predict the grinding force fluctuation information according to the spectral characteristic parameters; S5. Generate a feedforward compensation instruction according to the grinding force fluctuation information; S6. Generate a feedback compensation instruction according to the difference between the grinding force signal and the target grinding force of the current task; S7. Synthesize the feedforward compensation instruction and the feedback compensation instruction to generate a compensation control instruction.

[0007] The method of this application adopts a control strategy combining feedforward and feedback, significantly improving the response speed and compensation accuracy of the system to the dynamic change of grinding force, thus effectively suppressing the elastic deformation caused by grinding force fluctuation, reducing the cutting trajectory deviation and microtopography error, and finally improving the grinding accuracy and surface quality of the key parts of the crankshaft connecting rod.

[0008] In the dynamic error compensation method of the crankshaft connecting rod grinding force, when the prediction reference signal is the acoustic emission signal, the spectral characteristic parameters include the energy change rate or the spectral distribution; The extraction process of the energy change rate includes: S31. Perform a fast Fourier transform on the prediction reference signal to obtain the first spectrum; S32. Divide the first spectrum into multiple frequency bands, calculate the energy of each frequency band, and calculate the change rate of the energy of the same frequency band in adjacent time windows as the energy change rate; The extraction process of the spectral distribution includes: S33. Calculate the mean, variance, skewness and kurtosis of the first spectrum obtained in step S31 to form the spectral distribution.

[0009] Through the above technical solution, when the prediction reference signal is the acoustic emission signal, spectral characteristic parameters that can better reflect the dynamic change of the grinding process can be extracted, thereby improving the prediction accuracy of the grinding force fluctuation information, providing a more reliable basis for subsequent dynamic error compensation, and helping to improve the machining accuracy of the crankshaft connecting rod grinding.

[0010] In the dynamic error compensation method of the crankshaft connecting rod grinding force, when the prediction reference signal is the current signal of the grinding wheel spindle, the spectral characteristic parameters include the amplitude of the high-frequency component of the spindle current; The extraction process of the amplitude of the high-frequency component includes: S31’. Perform a fast Fourier transform on the prediction reference signal to obtain the second spectrum; S32’. According to the preset frequency threshold, screen the high-frequency components in the second spectrum that are higher than the frequency threshold; S33'. Obtain the amplitudes of the high-frequency components above the frequency threshold, and screen the amplitudes of a preset number of high-frequency components with the largest amplitudes as spectral feature parameters.

[0011] The described dynamic error compensation method for the grinding force of a crankshaft connecting rod, wherein step S5 includes: S51. Use the predicted grinding force fluctuation information as the input of a pre-constructed feedforward controller, calculate the obtained feedforward compensation command, the feedforward compensation command being the control signal of the actuator, and the feedforward controller being used to compensate for the dynamic characteristics of the actuator of the grinding equipment.

[0012] The described dynamic error compensation method for the grinding force of a crankshaft connecting rod, wherein step S6 includes: S61. Obtain the target grinding force of the current task, and calculate the deviation between the grinding force signal and the target grinding force to obtain a force deviation signal; S62. Process the force deviation signal using a PID control algorithm to generate a preliminary feedback compensation command; S63. Perform a limiting process on the preliminary feedback compensation command to generate the feedback compensation command.

[0013] The described dynamic error compensation method for the grinding force of a crankshaft connecting rod, wherein step S7 includes: S71. Based on the dynamic energy weight, perform a weighted sum of the feedforward compensation command and the feedback compensation command to obtain the compensation control command.

[0014] The described dynamic error compensation method for the grinding force of a crankshaft connecting rod, wherein step S7 further includes: S72. Perform a moving average filtering process on the compensation control command to obtain the final compensation control command.

[0015] The described dynamic error compensation method for the grinding force of a crankshaft connecting rod, wherein the determination process of the dynamic energy weight includes: S70. Calculate the instantaneous energies of the feedforward compensation command and the feedback compensation command respectively to obtain a feedforward energy value and a feedback energy value, and determine the dynamic energy weight according to the normalization relationship between the feedforward energy value and the feedback energy value.

[0016] The described dynamic error compensation method for the grinding force of a crankshaft connecting rod, wherein the method further includes a step executed between step S2 and step S3: SA. Perform noise reduction and filtering processing on the predicted reference signal and the grinding force signal.

[0017] In a second aspect, the present application also provides a dynamic error compensation system for the grinding force of a crankshaft connecting rod, which is used to compensate for the elastic deformation caused by the grinding force. The system includes: The first acquisition module is used to acquire the acoustic emission signal of the current grinding area based on an acoustic emission sensor or acquire the current signal of the grinding wheel spindle based on a high-speed spindle current sensor as a prediction reference signal; The second acquisition module is used to acquire the grinding force signal of the grinding wheel spindle based on a piezoelectric sensor; The feature extraction module is used to extract spectral feature parameters according to the prediction reference signal; The prediction module is used to predict the grinding force fluctuation information based on an online identification prediction model according to the spectral feature parameters; The first instruction module is used to generate a feedforward compensation instruction according to the grinding force fluctuation information; The second instruction module is used to generate a feedback compensation instruction according to the difference between the grinding force signal and the target grinding force of the current task; The instruction fusion module is used to generate a compensation control instruction by integrating the feedforward compensation instruction and the feedback compensation instruction.

[0018] The system of the present application adopts a control strategy combining feedforward and feedback, significantly improving the response speed and compensation accuracy of the system to the dynamic change of grinding force, thereby effectively suppressing the elastic deformation caused by the grinding force fluctuation, reducing the cutting trajectory deviation and the micro-topography error, and finally improving the grinding accuracy and surface quality of the key parts of the crankshaft connecting rod.

[0019] As can be seen from the above, the present application provides a method and system for compensating the dynamic error of the grinding force of a crankshaft connecting rod. Among them, the method introduces a prediction mechanism for grinding force fluctuation based on prediction reference signals such as acoustic emission or spindle current, and combines an online identification prediction model to generate a feedforward compensation instruction, so as to perform predictive compensation before the actual occurrence of high-frequency fluctuations of the grinding force, achieving the effects of effectively suppressing the instantaneous force fluctuation caused by microscopic non-uniformity, reducing dynamic elastic deformation, and improving the micro-topography accuracy and processing stability. It adopts a control strategy combining feedforward and feedback, significantly improving the response speed and compensation accuracy of the system to the dynamic change of grinding force, thereby effectively suppressing the elastic deformation caused by the grinding force fluctuation, reducing the cutting trajectory deviation and the micro-topography error, and finally improving the grinding accuracy and surface quality of the key parts of the crankshaft connecting rod. Description of the Drawings

[0020] Figure 1 It is a flowchart of a method for compensating the dynamic error of the grinding force of a crankshaft connecting rod provided by an embodiment of the present application.

[0021] Figure 2 It is a schematic structural diagram of a system for compensating the dynamic error of the grinding force of a crankshaft connecting rod provided by an embodiment of the present application.

[0022] Reference numerals: 201, the first acquisition module; 202, the second acquisition module; 203, the feature extraction module; 204, the prediction module; 205, the first instruction module; 206, the second instruction module; 207, the instruction fusion module. Detailed implementation manners

[0023] Next, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Components of the embodiments of the present application described and illustrated herein can be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application claimed, but merely represents selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative efforts fall within the scope of protection of the present application.

[0024] It should be noted that: similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present application, the terms "first", "second", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0025] In a first aspect, please refer to Figure 1 , some embodiments of the present application provide a method for dynamically compensating the grinding force dynamic error of a crankshaft connecting rod, which is used to compensate for the elastic deformation caused by the grinding force. The method includes the following steps: S1. Acquire the acoustic emission signal of the current grinding area based on an acoustic emission sensor or acquire the current signal of the grinding wheel spindle based on a high-speed spindle current sensor as a prediction reference signal; S2. Acquire the grinding force signal of the grinding wheel spindle based on a piezoelectric sensor; S3. Extract spectral feature parameters according to the prediction reference signal; S4. Based on an online identification prediction model, predict the grinding force fluctuation information according to the spectral feature parameters; S5. Generate a feedforward compensation instruction according to the grinding force fluctuation information; S6. Generate a feedback compensation instruction according to the difference between the grinding force signal and the target grinding force of the current task; S7. Synthesize the feedforward compensation instruction and the feedback compensation instruction to generate a compensation control instruction.

[0026] Specifically, the prediction reference signal refers to the signal used to perceive in advance the changes in the state of the grinding area. It can be realized by using the acoustic emission signal collected based on an acoustic emission sensor or the current signal of the grinding wheel spindle collected based on a high-speed spindle current sensor. For example, the ultrasonic signal generated during the grinding process is obtained through an acoustic emission sensor installed near the grinding area, or the grinding load is reflected by monitoring the current change of the grinding wheel spindle motor. Its main function is to reflect microscopic events such as abrasive grain penetration, fragmentation, shedding, and local material hardness changes, and provide prior information for predicting the grinding force fluctuation.

[0027] More specifically, the spectral feature parameter refers to the numerical value or vector extracted from the prediction reference signal that can characterize the frequency components and energy distribution of the signal, and is used to quantify the frequency domain information related to the grinding state in the prediction reference signal.

[0028] More specifically, the online identification prediction model refers to a mathematical model that can update its own parameters in real time during the operation of the system, and is used to predict future grinding force fluctuation information based on the input spectral feature parameters. It can be realized by using a neural network model, a support vector machine model or a time series model trained based on the recursive least squares method, the Kalman filter or other online learning algorithms. For example, a model is constructed that can predict the grinding force fluctuation at the next moment based on historical spectral features and grinding force data, and continuously adjusts the internal parameters of the model according to the error between the actual grinding force and the model prediction value. Its main function is to establish a dynamic mapping relationship between the prediction reference signal and the grinding force fluctuation, and realize the advance prediction of the grinding force fluctuation.

[0029] More specifically, the feedforward compensation instruction refers to the control signal generated according to the predicted grinding force fluctuation information and used to offset part of the expected error in advance, which is used to compensate for the dynamic characteristics of the actuator of the grinding equipment and perform compensation before the actual occurrence of the grinding force fluctuation to reduce the lag.

[0030] More specifically, the feedback compensation instruction refers to the control signal generated according to the deviation between the actual grinding force and the target grinding force and used to eliminate the residual error, which is used to stabilize the actual grinding force near the target value and eliminate the steady-state error and suppress the dynamic error not fully compensated by the feedforward.

[0031] More specifically, the compensation control instruction refers to the control signal generated by integrating the feedforward compensation instruction and the feedback compensation instruction and finally used to control the actuator to adjust the relative position between the grinding wheel and the workpiece, which is used to combine the fast response ability of the feedforward control and the precise tracking ability of the feedback control to generate a more effective and robust control signal. The compensation control instruction acts on the actuator of the grinding equipment to adjust the relative position between the grinding wheel and the workpiece, so as to realize the dynamic compensation of the grinding force.

[0032] More specifically, the spectral features are extracted in step S3 because many microscopic events during the grinding process exhibit specific patterns in the frequency domain. The online identification model is adopted in step S4 because the grinding process is dynamically changing and the model needs to adapt in real time. Feedforward compensation is adopted in step S5 because it can overcome the lag of traditional feedback control and cancel predictable disturbances in advance. Feedback compensation is adopted in step S6 because it can eliminate the residual error caused by prediction errors and unmodeled disturbances and ensure the steady-state accuracy of the system. The feedforward and feedback are integrated in step S7 because the advantages of both can be combined to achieve faster, more accurate, and more robust dynamic error compensation. It is precisely due to this control strategy that combines feedforward based on the prediction of a reference signal and feedback based on the actual force that the system can effectively cope with high-frequency and instantaneous force fluctuations during the grinding process, reduce elastic deformation, and improve machining accuracy.

[0033] The method of the present application introduces a prediction mechanism for grinding force fluctuations based on a reference signal such as acoustic emission or spindle current, and combines an online identification prediction model to generate a feedforward compensation command, so as to perform predictive compensation before the actual occurrence of high-frequency fluctuations in grinding force, achieving the effects of effectively suppressing instantaneous force fluctuations caused by microscopic inhomogeneities, reducing dynamic elastic deformation, improving microscopic topography accuracy, and machining stability. It adopts a control strategy that combines feedforward and feedback, significantly improving the response speed and compensation accuracy of the system to the dynamic changes in grinding force, thus effectively suppressing the elastic deformation caused by grinding force fluctuations, reducing cutting trajectory deviation and microscopic topography error, and ultimately improving the grinding machining accuracy and surface quality of the key parts of the crankshaft connecting rod.

[0034] In some preferred embodiments, when the reference signal is an acoustic emission signal, the spectral feature parameters include the energy change rate or spectral distribution; The extraction process of the energy change rate includes: S31. Perform a fast Fourier transform on the reference signal to obtain the first spectrum; S32. Divide the first spectrum into multiple frequency bands, calculate the energy of each frequency band, and calculate the change rate of the energy of the same frequency band within adjacent time windows as the energy change rate; The extraction process of the spectral distribution includes: S33. Calculate the mean, variance, skewness, and kurtosis of the first spectrum obtained in step S31 to form the spectral distribution.

[0035] Specifically, the fast Fourier transform can convert a time-domain signal into a frequency-domain signal, revealing the components of the signal at different frequencies and their intensities. The design purpose of step S31 is to convert the acoustic emission signal from the time domain to the frequency domain for analyzing its frequency characteristics; the design purpose of step S32 is to conduct a detailed analysis of the energy distribution of the signal within different frequency ranges; energy refers to the intensity measure of the signal within a specific frequency band, usually obtained by calculating the sum of squares or integral of the spectral amplitudes within that frequency band, which can be achieved by squaring and summing the amplitudes at each frequency point within the frequency band, aiming to quantify the activity level of the acoustic emission signal within different frequency ranges; adjacent time windows refer to two data acquisition windows that are temporally adjacent in the continuously acquired acoustic emission signal sequence, and the energy change rate refers to the relative or absolute change in energy within the same frequency band between adjacent time windows, reflecting the dynamic evolution of the energy in that frequency band over time, aiming to capture the energy fluctuations caused by transient events during the grinding process.

[0036] More specifically, the spectral distribution refers to a set of statistics that describe the overall characteristics of a signal in the frequency domain. In this application, it specifically consists of the mean, variance, skewness, and kurtosis, aiming to comprehensively characterize the spectral shape of the acoustic emission signal through multiple statistical dimensions. Among them, the mean reflects the central frequency of the spectrum or the energy concentration trend; the variance reflects the degree of dispersion or bandwidth of the spectrum; the skewness reflects the symmetry of the spectrum, indicating whether the energy is biased towards high frequencies or low frequencies; the kurtosis reflects the sharpness of the spectrum or the presence of prominent frequency components. These statistics can be obtained from the first spectral data using standard statistical calculation methods, which will not be elaborated here.

[0037] Through the above technical solution, when the prediction reference signal is an acoustic emission signal, spectral feature parameters that can better reflect the dynamic changes during the grinding process can be extracted, thereby improving the prediction accuracy of the grinding force fluctuation information and providing a more reliable basis for subsequent dynamic error compensation, which helps to improve the machining accuracy of crankshaft connecting rod grinding.

[0038] In some preferred embodiments, when the spectral feature parameters include the energy change rate, the online identification prediction model is a linear adaptive time series model, satisfying: F(t + 1)=θ T (t)⋅X(t) (1) Wherein, F(t + 1) is the predicted grinding force fluctuation information at time t + 1, θ(t) is the dynamically updated weight vector identified online at time t, T is the matrix transpose symbol, and X(t) is the energy change rate matrix at time t, satisfying: X(t)=[ΔE1 (t),...,ΔEk (t)] T , and ΔEk (t) is the energy change rate of the kth frequency band at time t.

[0039] Specifically, the linear adaptive time series model refers to a linear mathematical model that can process time series data and whose model parameters can be adjusted and updated in real time according to new data. Its purpose is to quickly adapt to the changing pattern of input data over time, thereby improving the accuracy of prediction.

[0040] More specifically, θ(t) is the model weight vector calculated or updated through an online identification algorithm. The elements of this vector reflect the contribution degree of different input features to the prediction result. The purpose of its dynamic update is to enable the model to continuously track the non-steady characteristics of the grinding process.

[0041] More specifically, this linear adaptive time series model uses the energy change rate as the input feature of the model to realize the prediction of grinding force fluctuation information. It uses the current dynamic weight vector θ(t) to perform weighted summation on the input feature X(t), thereby calculating the predicted value F(t + 1) of the grinding force fluctuation at the next moment. Among them, the weight vector θ(t) is updated in real time through an online identification algorithm, which means that the model can continuously adjust its own parameters according to the error between the actually collected grinding force and the model predicted value, so that the model can quickly adapt to the dynamic changes caused by factors such as wheel wear and material inhomogeneity during the grinding process, enabling the prediction model to not only utilize the characteristic information reflecting the process state but also make self-adjustment according to the actual changes of the process. Therefore, it can predict the grinding force fluctuation more accurately and in a more timely manner, providing a reliable basis for subsequent dynamic error compensation, and thus effectively dealing with the high-frequency and non-periodic force fluctuation problems that are difficult to handle in the existing technology.

[0042] More specifically, the structure of the linear model is simple and the calculation efficiency is high, which is conducive to realizing real-time prediction; the adaptive characteristic enables the model to adjust parameters according to the actual changes of the grinding process, improving the accuracy of prediction; the energy change rate feature can effectively reflect the microscopic events during the grinding process, enhancing the sensitivity of prediction. These prediction information can provide a reliable basis for subsequent dynamic error compensation of grinding force, thereby improving the machining accuracy.

[0043] In some preferred embodiments, the update of θ(t) satisfies: θ(t)= θ(t - 1)+K(t)e(t) (2) Where, θ(t - 1) is the dynamic weight vector at time t - 1, K(t) is the Kalman gain matrix at time t, and e(t) is the prediction error, that is, the deviation between the grinding force fluctuation information at time t and the actual situation. Thus, θ(t) can achieve online dynamic update.

[0044] In some preferred embodiments, when the spectral characteristic parameters include spectral distribution, the online identification prediction model is a non-linear feature fusion model, satisfying: (3) Among them, F(t + 1) is the predicted grinding force fluctuation information at time t + 1, N is the total number of support vectors, S i is the i-th support vector, α i is the Lagrange multiplier of S i , b is the bias term, κ(Si, S(t)) is the RBF kernel function, S(t) is the spectral distribution at time t, and it satisfies: S(t) = [μ, σ 2 , γ, κ f T , where μ is the mean, σ 2 is the variance, γ is the skewness, κ f is the kurtosis, and T is the matrix transpose symbol.

[0045] Specifically, the non-linear feature fusion model refers to a model that can handle the non-linear relationship between input features, and its purpose is to effectively capture the complex non-linear relationship between various statistical features in the spectral distribution. The spectral distribution refers to a set of statistics that describe the overall shape and characteristics of the spectrum after performing spectral analysis on the signal. It can be obtained by calculating the mean, variance, skewness, kurtosis, etc. of the spectrum, and its purpose is to quantify the characteristic information of the spectrum.

[0046] More specifically, the non-linear feature fusion model is based on the principle of support vector machines and uses the RBF kernel function to map the spectral distribution S(t) to a high-dimensional space, so as to better capture the non-linear relationship between parameters such as the mean μ, variance σ2, skewness γ, and kurtosis κ f in the spectral distribution. The model calculates the RBF kernel function similarity between the current spectral distribution S(t) and the pre-determined support vector Si, and performs weighted summation on these similarities, and then adds the bias term b to obtain the predicted grinding force fluctuation information F(t + 1) at time t + 1. This way of non-linear mapping and weighted summation enables the model to learn and approximate the complex non-linear mapping relationship between the spectral distribution and the grinding force fluctuation.

[0047] More specifically, since the spectral distribution can reflect the signal characteristics caused by microscopic events such as abrasive grain penetration, fragmentation, shedding, and local hardness changes of the material, and these events are closely related to the high-frequency fluctuations of the grinding force, and there are often non-linear correlations between their characteristic parameters (such as the skewness and kurtosis of the spectrum), the non-linear feature fusion model can be used to more accurately predict the grinding force fluctuation information from the spectral distribution.

[0048] ​Through the above technical solution, when the spectral feature parameter is the spectral distribution, the nonlinear feature fusion model is used to predict the grinding force fluctuation information, which can effectively capture the complex nonlinear relationship between the statistical features in the spectral distribution, thereby improving the prediction accuracy of the grinding force fluctuation information, providing a more reliable basis for feedforward compensation, and further enhancing the overall effect of the dynamic error compensation of the crankshaft connecting rod grinding force, especially when dealing with high-frequency and nonlinear grinding force fluctuations caused by material micro-inhomogeneity or local changes in the grinding wheel state.

[0049] It is worth mentioning that S i and α i are updated based on the online update of the acquisition data by a sliding window, that is, updated based on the new samples selected by the sliding window. This belongs to the conventional parameter update process and will not be elaborated here.

[0050] In some preferred embodiments, when the prediction reference signal is the current signal of the grinding wheel spindle, the spectral feature parameter includes the amplitude of the high-frequency component of the spindle current; The extraction process of the amplitude of the high-frequency component includes: S31’: Perform a fast Fourier transform on the prediction reference signal to obtain the second spectrum; S32’: According to the preset frequency threshold, filter out the high-frequency components in the second spectrum that are higher than the frequency threshold; S33’: Obtain the amplitudes of the high-frequency components higher than the frequency threshold, and filter out the amplitudes of the preset number of high-frequency components with the largest amplitudes as the spectral feature parameters.

[0051] Specifically, the second spectrum refers to the frequency-amplitude or frequency-power spectral density distribution obtained by performing a fast Fourier transform on the prediction reference signal, which can be expressed as a series of discrete frequency points and their corresponding amplitudes or energies. Its purpose is to show the energy or intensity distribution of the signal at different frequencies; the preset frequency threshold refers to a frequency limit preset in spectral analysis for distinguishing the high-frequency and low-frequency components of the signal, which can be determined according to specific grinding processes, machine tool characteristics or experience. Its purpose is to filter out low-frequency components that are irrelevant or interfering with the grinding force fluctuation and focus on high-frequency features; the high-frequency components reflect the rapidly changing components in the signal, and its purpose is to capture the fast dynamic information related to the microscopic events in the grinding process; the amplitude is used to quantify the intensity of different frequency components; the preset number can be adjusted according to actual needs and computing resources. In the embodiments of the present application, it is preferably 3, which can reduce the computational complexity while ensuring the richness of feature information and focusing on the main high-frequency components that have a greater impact on the grinding force fluctuation.

[0052] The amplitudes of these high-frequency components are used as spectral feature parameters and input into an online identification and prediction model to predict the grinding force fluctuation information. This method of feature extraction based on the spindle current signal of the grinding wheel provides another more robust prediction reference signal source compared with the method of feature extraction based on the acoustic emission signal in the basic solution. It effectively overcomes the problem that the acoustic emission sensor is vulnerable to environmental interference, improves the quality of the prediction reference signal and the effectiveness of feature extraction, and further improves the accuracy and stability of the grinding force fluctuation prediction, making the subsequent feedforward compensation instructions more accurate, thus improving the overall effect of the grinding force dynamic error compensation, overcoming the disadvantage that the acoustic emission sensor is vulnerable to interference, simplifying the complexity of sensor installation and debugging, and enhancing the overall effect of the grinding force dynamic error compensation.

[0053] In some preferred embodiments, when the spectral feature parameters include the amplitudes of the high-frequency components of the spindle current, the online identification and prediction model is an adaptive frequency-amplitude weighted model, satisfying: (4) where F(t + 1) is the predicted grinding force fluctuation information at time t + 1, M is the preset number, w j (t) is the online updated adaptive weight of the amplitude of the jth high-frequency component at time t, c is the bias term, A fj (t) is the amplitude of the high-frequency component at the frequency fj at time t, and fj is the frequency of the jth high-frequency component.

[0054] Specifically, the adaptive frequency-amplitude weighted model refers to a prediction model structure that performs weighted summation on the amplitudes of high-frequency components at different frequencies, and these weights are dynamically adjusted over time to distinguish the influence degrees of different frequency components on the grinding force fluctuation.

[0055] More specifically, the adaptive weight refers to a coefficient assigned to the amplitude of each high-frequency component, which is dynamically adjusted over time and can be calculated and updated online according to the prediction error, the gradient descent method or other optimization algorithms, aiming to reflect the importance of different frequency components in contributing to the grinding force fluctuation at the current moment.

[0056] More specifically, the above-mentioned adaptive frequency-amplitude weighted model receives the preset number of amplitudes of high-frequency components extracted from the spindle current signal of the grinding wheel as inputs, assigns an adaptively adjusted weight to each input amplitude of the high-frequency component, multiplies each amplitude of the high-frequency component by its corresponding adaptive weight, then sums all the weighted amplitudes, and adds a dynamic bias term. The model can predict the grinding force fluctuation information at the next moment, can more effectively extract the effective information related to the grinding force fluctuation from the complex current signal, filter out the noise or irrelevant components, thereby improving the accuracy of the prediction and the adaptability to different grinding conditions.

[0057] In some preferred embodiments, the adaptive weight satisfies: w j (t + 1) = w j (t) + △w j (t) (5) where, w j (t + 1) is the online updated adaptive weight of the amplitude of the j-th high-frequency component at time t + 1, and △w j (t) is the weight correction amount of the j-th high-frequency component at time t, and satisfies: △w j (t) = η·e(t)·A fj (t) (6) where, η is the learning rate, which is used to control the weight update step size and can be set according to usage requirements, such as 0.05. Thus, w j (t) can achieve online dynamic update.

[0058] In some preferred embodiments, step S5 includes: S51. Using the predicted grinding force fluctuation information as the input of a pre-constructed feedforward controller, and calculating to obtain a feedforward compensation command.

[0059] Specifically, the feedforward compensation command is a control signal that can be directly used by the actuator of the grinding equipment. The feedforward controller is used to compensate for the dynamic characteristics of the actuator of the grinding equipment. The pre-constructed feedforward controller can be a digital filter, a model-based controller, or a look-up table, such as an FIR or IIR filter, whose transfer function is designed to approximate the discretized form of the inverse dynamics of the actuator. The design purpose of the feedforward controller is to output an adjusted control signal when receiving an input signal to offset the adverse effects brought by the dynamic response of the actuator itself.

[0060] More specifically, the feedforward controller processes the predicted information based on the understanding of the dynamic characteristics of the actuator, generates a feedforward compensation command that can effectively drive the actuator to perform a compensation action, so that the generated feedforward compensation command can more accurately predict the actual response of the actuator, thereby ensuring that the actuator can perform the compensation action in a timely and accurate manner to offset the elastic deformation caused by the predicted grinding force fluctuation.

[0061] In some preferred embodiments, the feedforward controller is designed using the inverse model method based on a pre-constructed transfer function model, and the transfer function model characterizes the dynamic relationship between the input signal and the output displacement of the actuator. The parameters of the transfer function model are obtained through offline experiments or online identification methods.

[0062] Specifically, the inverse model method refers to establishing a mathematical model of the controlled object (actuator), and then designing a controller whose transfer function is equal to the reciprocal (or approximate reciprocal) of the transfer function of the controlled object, so as to eliminate the dynamic influence of the controlled object under ideal conditions. It can be implemented by techniques such as inverse based on transfer function and inverse based on state space model. Its purpose is to offset the dynamic response of the actuator, so that the input signal can more directly control the output displacement.

[0063] More specifically, inputting the predicted grinding force fluctuation information into the feedforward controller designed based on the inverse model can generate a control signal that can compensate for the dynamic response of the actuator in advance or synchronously. This feedforward compensation command is integrated with the feedback compensation command generated according to the actual grinding force signal to jointly control the actuator to adjust the relative position between the grinding wheel and the workpiece. In this way, the feedforward control can effectively compensate for the dynamic characteristics of the actuator of the grinding equipment, making the response of the actuator faster and more accurate, thereby improving the suppression ability of the entire compensation system to high-frequency grinding force fluctuations.

[0064] In some preferred embodiments, step S6 includes: S61. Obtain the target grinding force of the current task, and calculate the deviation between the grinding force signal and the target grinding force to obtain a force deviation signal; S62. Process the force deviation signal with a PID control algorithm to generate a preliminary feedback compensation command; S63. Perform a limiting process on the preliminary feedback compensation command to generate a feedback compensation command.

[0065] Specifically, the PID control algorithm processing refers to a control method that combines proportional, integral, and differential links. It can calculate the control output according to the magnitude, cumulative error, and change rate of the force deviation signal to achieve precise control of the system. The limiting process refers to limiting the value of the signal between a preset maximum value and minimum value, which can prevent the signal from exceeding the allowable range and protect the stable operation of the actuator or system.

[0066] The solution of this application obtains the actual grinding force signal and compares it with the target grinding force to obtain a force deviation signal that reflects the difference between the current grinding state and the desired state. The PID control algorithm can comprehensively consider the current value, historical cumulative value, and change trend of the deviation, so as to generate a preliminary feedback compensation instruction that can quickly respond to the deviation and at the same time suppress overshoot and oscillation. On this basis, a limit processing is performed on the preliminary feedback compensation instruction to ensure that the generated feedback compensation instruction does not exceed the physical or safety limits of the actuator, and to avoid system instability or damage caused by excessive control quantity. The feedback compensation instruction generated in this way can stably and effectively drive the actuator to adjust the relative position between the grinding wheel and the workpiece, and maintain the actual grinding force near the target value. The combination of this optimized feedback control and the prediction-based feedforward compensation, where the feedforward compensation is responsible for handling predictable dynamic fluctuations, and the optimized feedback compensation is responsible for eliminating residual deviations and coping with unpredictable disturbances. The two work together to achieve precise compensation for the complex dynamic errors in the crankshaft connecting rod grinding process, thereby significantly improving the machining accuracy and stability.

[0067] In some preferred embodiments, step S7 includes: S71. Weighted sum the feedforward compensation instruction and the feedback compensation instruction based on the dynamic energy weight to obtain a compensation control instruction.

[0068] Specifically, the dynamic energy weight refers to a parameter whose weight value changes dynamically with time. It can be determined based on the instantaneous amplitude, root mean square value, variance, or other statistics reflecting the signal strength of the feedforward compensation instruction and the feedback compensation instruction. Its purpose is to adaptively adjust their contribution ratios in the final compensation control instruction according to the real-time state or relative importance of the feedforward compensation instruction and the feedback compensation instruction.

[0069] More specifically, the solution of this application realizes the effective fusion of the feedforward compensation instruction and the feedback compensation instruction by weighted summation based on the dynamic energy weight. This dynamic energy weight mechanism can evaluate the signal strengths of the feedforward compensation instruction and the feedback compensation instruction in real time. When the predicted grinding force fluctuation information is accurate and the energy of the feedforward compensation instruction is high, the system assigns a greater weight to the feedforward compensation instruction, enabling it to play a dominant role in the final compensation control instruction, thereby utilizing the fast response ability of the feedforward compensation to offset the predicted grinding force fluctuation in a timely manner. Conversely, when there is a large deviation between the actual grinding force and the target grinding force and the energy of the feedback compensation instruction is high, indicating that a stronger corrective action is required, the system assigns a greater weight to the feedback compensation instruction and uses the accuracy and stability of the feedback compensation to reduce the error and ensure the control accuracy. This dynamically adjusted mechanism makes full use of the rapidity of the feedforward compensation and the accuracy of the feedback compensation to form an adaptive compensation strategy. This strategy is combined with the processes of predicting the grinding force fluctuation information based on the predicted reference signal and generating the feedforward compensation instruction, and generating the feedback compensation instruction based on the actual grinding force in the foregoing steps to construct a complete control loop that can perform predictive compensation and corrective compensation simultaneously, thereby effectively coping with the complex and variable dynamic force fluctuations during the grinding process.

[0070] In some preferred embodiments, step S7 further includes: S72. Perform a moving average filtering process on the compensation control instruction to obtain the final compensation control instruction.

[0071] Specifically, the moving average filtering process refers to a digital signal processing technique that smooths the signal by calculating the average value of the input signal within a certain time window. Specifically, it can be achieved by setting a sliding window with a fixed length, summing the data within the window and dividing by the window length, and then sliding the window forward by one position and repeating the above process. Its purpose is to filter out high-frequency noise and instantaneous mutations in the signal.

[0072] More specifically, the solution of this application performs a moving average filtering process on the compensation control instruction obtained by weighted summation. The moving average filtering effectively suppresses the high-frequency noise or instantaneous mutations that may be introduced during the weighted summation process by averaging the instructions over a period of time. After the filtering process, a smooth final compensation control instruction is obtained, which is used to control the actuator to adjust the relative position between the grinding wheel and the workpiece. This processing method makes the control signal received by the actuator more stable, avoiding frequent minor adjustments, thereby improving the overall stability and control accuracy of the system.

[0073] In some preferred embodiments, the determination process of the dynamic energy weight includes: S70. Calculate the instantaneous energy of the feedforward compensation command and the feedback compensation command respectively to obtain the feedforward energy value and the feedback energy value, and determine the dynamic energy weight according to the normalization relationship between the feedforward energy value and the feedback energy value. The specific calculation process can be as follows: w e = (Efw + ε) / (Efw + Efb + 2ε) (7) where, w e is the dynamic energy weight, Efw is the feedforward energy value, Efb is the feedback energy value, and ε is a very small positive number.

[0074] Specifically, the instantaneous energy refers to the energy magnitude of the signal at a certain moment or within a very short time, which can be realized by calculating the square of the signal amplitude. The normalization relationship refers to the relationship of scaling the value to a specific range, which can be realized by using the proportion of the feedforward energy value in the sum of the feedforward energy value and the feedback energy value as the weight basis. The dynamic energy weight refers to the weight coefficient dynamically adjusted according to the energy of the feedforward and feedback compensation commands, which can be represented by a value between 0 and 1; the very small positive number ε refers to a very small positive value, which can be realized by a floating-point number close to zero.

[0075] More specifically, the solution of the present application can evaluate the relative contributions of feedforward prediction and feedback correction to the current grinding force error compensation in real time by calculating the instantaneous energy of the feedforward compensation command and the feedback compensation command. Formula (7) uses the relative proportion of the feedforward energy value in the total feedforward and feedback energy as the basic basis for the weight, so that the weight can adaptively reflect the strength of the two compensation commands.

[0076] It should be noted that when performing weighted calculation in step S71, w e as the dynamic energy weight of the feedforward energy value, the dynamic energy weight of the feedback energy value is (1 - w e ).

[0077] Through the above technical solution, by using the weight calculation formula based on energy normalization and introducing a very small positive number, the stable determination of the dynamic energy weights of the feedforward compensation command and the feedback compensation command can be realized. This method effectively avoids the sensitivity and instability of weight calculation when the energy value is small or fluctuating, making the fusion of the feedforward and feedback compensation commands smoother and more reasonable. This improves the stability and accuracy of the compensation control command, thereby enhancing the control effect of the grinding force dynamic error compensation.

[0078] In some preferred embodiments, the method further includes a step executed between step S2 and step S3: SA. Denoise and filter the prediction reference signal and the grinding force signal.

[0079] Through the above technical scheme, the prediction reference signal and grinding force signal are subjected to noise reduction and filtering processing, which improves the quality of the signal, makes the subsequent extracted spectral feature parameters more accurate, improves the prediction accuracy of the online identification prediction model, and makes the generated compensation control instructions more accurate, thereby effectively compensating for the dynamic error caused by the grinding force fluctuation and improving the grinding processing accuracy.

[0080] In some preferred embodiments, step SA comprises: SA1, using the wavelet threshold denoising method to perform denoising on the prediction reference signal, and using the Butterworth bandpass filter to filter the denoised prediction reference signal; SA2. Adaptive noise cancellation method is used to reduce the noise of the grinding force signal, and Kalman filter is used to filter the noise-reduced grinding force signal.

[0081] Specifically, for the prediction reference signal, the wavelet threshold denoising method is used for denoising. This method can adaptively select a suitable threshold for denoising according to the frequency characteristics of the signal, so as to remove the noise while retaining the useful information in the signal as much as possible, especially the transient characteristics. Subsequently, the Butterworth bandpass filter is used to filter the denoised prediction reference signal. The Butterworth bandpass filter has a smooth amplitude-frequency response and can effectively filter out high-frequency and low-frequency noise in the prediction reference signal, thereby improving the signal-to-noise ratio of the signal and retaining specific frequency band information reflecting the behavior of the abrasive particles and material changes. By first reducing the noise and then filtering, the influence of noise on the filtering effect can be avoided, and a purer prediction reference signal with key features retained can be obtained.

[0082] More specifically, for the grinding force signal, an adaptive noise cancellation method is used for noise reduction processing. This method can adaptively adjust the parameters of the filter according to the reference noise signal, and effectively cancel the noise in the grinding force signal that is not related to the grinding process, such as machine tool vibration or electromagnetic interference. Then, a Kalman filter is used to filter the denoised grinding force signal. The Kalman filter is an optimal linear filter that can optimally estimate the signal according to the statistical characteristics of the signal, effectively filter out the random noise in the grinding force signal, and obtain a smoother grinding force signal. By first reducing the noise and then filtering, the estimation accuracy of the Kalman filter can be improved, and a more accurate grinding force signal can be obtained. This targeted signal preprocessing method provides a high-quality input signal for subsequent spectral feature extraction, grinding force fluctuation information prediction, and feedback compensation.

[0083] More specifically, the high-quality prediction reference signal enables more accurate extraction of spectral feature parameters, thereby improving the accuracy of the online identification prediction model for predicting the grinding force fluctuation information. The high-quality grinding force signal enables the feedback control to perform deviation calculation and control based on a more accurate actual force signal, improving the effectiveness of the feedback compensation. Therefore, by optimizing the signal preprocessing link, the overall performance and accuracy of the entire crankshaft connecting rod grinding force dynamic error compensation method are significantly improved.

[0084] Through the above technical solution, it is possible to adopt optimized noise reduction and filtering methods according to the different characteristics of the prediction reference signal and the grinding force signal, effectively removing the noise and interference in the signal while retaining the key information in the signal. This improves the signal-to-noise ratio and purity of the signal, providing high-quality input for subsequent spectral feature extraction, grinding force fluctuation information prediction, and compensation control, thereby improving the accuracy and robustness of the entire grinding force dynamic error compensation method.

[0085] In a second aspect, please refer to Figure 2 , some embodiments of the present application further provide a crankshaft connecting rod grinding force dynamic error compensation system for compensating the elastic deformation caused by the grinding force. The system includes: A first acquisition module 201, configured to acquire an acoustic emission signal of the current grinding area based on an acoustic emission sensor or acquire a current signal of the grinding wheel spindle based on a high-speed spindle current sensor as a prediction reference signal; A second acquisition module 202, configured to acquire a grinding force signal of the grinding wheel spindle based on a piezoelectric sensor; A feature extraction module 203, configured to extract spectral feature parameters according to the prediction reference signal; A prediction module 204, configured to predict grinding force fluctuation information based on an online identification prediction model according to the spectral feature parameters; A first instruction module 205, configured to generate a feedforward compensation instruction according to the grinding force fluctuation information; A second instruction module 206, configured to generate a feedback compensation instruction according to the difference between the grinding force signal and the target grinding force of the current task; An instruction fusion module 207, configured to generate a compensation control instruction by integrating the feedforward compensation instruction and the feedback compensation instruction.

[0086] The system of the present application introduces a grinding force fluctuation prediction mechanism based on prediction reference signals such as acoustic emission or spindle current, and combines an online identification prediction model to generate a feedforward compensation instruction, so as to perform predictive compensation before the actual occurrence of high-frequency fluctuations in the grinding force, achieving the effects of effectively suppressing the instantaneous force fluctuations caused by microscopic inhomogeneities, reducing dynamic elastic deformation, improving the microscopic topography accuracy and machining stability. It adopts a control strategy combining feedforward and feedback, significantly improving the response speed and compensation accuracy of the system to the dynamic changes in the grinding force, thus effectively suppressing the elastic deformation caused by the grinding force fluctuations, reducing the cutting trajectory deviation and microscopic topography error, and finally improving the grinding accuracy and surface quality of the key parts of the crankshaft connecting rod.

[0087] In some preferred embodiments, the system further includes: A preprocessing module for denoising and filtering the prediction reference signal and the grinding force signal.

[0088] In addition, the unit described as a separate component may or may not be physically separated, and the component shown as a unit may or may not be a physical unit, that is, it may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0089] Furthermore, in each embodiment of the present application, the various functional modules may be integrated together to form an independent part, or each module may exist alone, or two or more modules may be integrated to form an independent part.

[0090] In this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0091] The above are only the embodiments of the present application and are not used to limit the protection scope of the present application. For those skilled in the art, the present application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A dynamic error compensation method for grinding force of a crankshaft connecting rod, which is used to compensate for the elastic deformation caused by the grinding force, is characterized in that The method includes the following steps: S1. Collect the acoustic emission signal of the current grinding area based on an acoustic emission sensor or collect the current signal of the grinding wheel spindle based on a high-speed spindle current sensor as the prediction reference signal; S2. Collect the grinding force signal of the grinding wheel spindle based on a piezoelectric sensor; S3. Extract spectral characteristic parameters according to the prediction reference signal; S4. Based on an online identification prediction model, predict the grinding force fluctuation information according to the spectral characteristic parameters; S5. Generate a feedforward compensation instruction according to the grinding force fluctuation information; S6. Generate a feedback compensation instruction according to the difference between the grinding force signal and the target grinding force of the current task; S7. Synthesize the feedforward compensation instruction and the feedback compensation instruction to generate a compensation control instruction.

2. The dynamic error compensation method for the grinding force of the crankshaft connecting rod according to claim 1, wherein When the prediction reference signal is the acoustic emission signal, the spectral characteristic parameters include the energy change rate or the spectral distribution; The extraction process of the energy change rate includes: S31. Perform a fast Fourier transform on the prediction reference signal to obtain a first spectrum; S32. Divide the first spectrum into multiple frequency bands, calculate the energy of each frequency band, and calculate the change rate of the energy of the same frequency band within adjacent time windows as the energy change rate; The extraction process of the spectral distribution includes: S33. Calculate the mean, variance, skewness, and kurtosis of the first spectrum obtained in step S31 to form the spectral distribution.

3. The dynamic error compensation method for the grinding force of the crankshaft connecting rod according to claim 1, characterized in that, When the prediction reference signal is the current signal of the grinding wheel spindle, the spectral characteristic parameters include the amplitude of the high-frequency component of the spindle current; The extraction process of the amplitude of the high-frequency component includes: S31’. Perform a fast Fourier transform on the prediction reference signal to obtain a second spectrum; S32’. According to a preset frequency threshold, screen the high-frequency components in the second spectrum that are higher than the frequency threshold; S33’. Obtain the amplitudes of the high-frequency components higher than the frequency threshold, and screen the amplitudes of a preset number of high-frequency components with the largest amplitudes as the spectral characteristic parameters.

4. The dynamic error compensation method for the grinding force of the crankshaft connecting rod according to claim 1, wherein Step S5 includes: S51. Take the predicted grinding force fluctuation information as the input of a pre-constructed feedforward controller, calculate to obtain the feedforward compensation instruction, the feedforward compensation instruction is the control signal of the actuator, and the feedforward controller is used to compensate for the dynamic characteristics of the actuator of the grinding equipment.

5. The dynamic error compensation method for the grinding force of the crankshaft connecting rod according to claim 1, characterized in that Step S6 includes: S61. Obtain the target grinding force of the current task, and calculate the deviation between the grinding force signal and the target grinding force to obtain a force deviation signal; S62. Process the force deviation signal with a PID control algorithm to generate a preliminary feedback compensation instruction; S63. Perform a limiting process on the preliminary feedback compensation instruction to generate the feedback compensation instruction.

6. The dynamic error compensation method for the grinding force of the crankshaft connecting rod according to claim 1, wherein Step S7 includes: S71. Perform a weighted sum of the feedforward compensation instruction and the feedback compensation instruction based on a dynamic energy weight to obtain the compensation control instruction.

7. The dynamic error compensation method for the grinding force of the crankshaft connecting rod according to claim 6, wherein Step S7 further includes: S72. Perform a moving average filtering process on the compensation control instruction to obtain the final compensation control instruction.

8. The dynamic error compensation method for the grinding force of the crankshaft connecting rod according to claim 6, characterized in that The determination process of the dynamic energy weight includes: S70. Calculate the instantaneous energy of the feedforward compensation command and the feedback compensation command respectively to obtain a feedforward energy value and a feedback energy value, and determine a dynamic energy weight according to the normalization relationship between the feedforward energy value and the feedback energy value.

9. The dynamic error compensation method for the grinding force of the crankshaft connecting rod according to claim 1, wherein The method further includes a step executed between step S2 and step S3: SA. Denoise and filter the prediction reference signal and the grinding force signal.

10. A dynamic error compensation system for the grinding force of a crankshaft connecting rod, which is used to compensate for the elastic deformation caused by the grinding force, is characterized in that The system includes: A first acquisition module, configured to acquire an acoustic emission signal of the current grinding area based on an acoustic emission sensor or acquire a current signal of the grinding wheel spindle based on a high-speed spindle current sensor as a prediction reference signal; A second acquisition module, configured to acquire a grinding force signal of the grinding wheel spindle based on a piezoelectric sensor; A feature extraction module, configured to extract spectral feature parameters according to the prediction reference signal; A prediction module, configured to predict grinding force fluctuation information based on an online identification prediction model according to the spectral feature parameters; A first command module, configured to generate a feedforward compensation command according to the grinding force fluctuation information; A second command module, configured to generate a feedback compensation command according to the difference between the grinding force signal and the target grinding force of the current task; A command fusion module, configured to generate a compensation control command by integrating the feedforward compensation command and the feedback compensation command.

Images