Manufacturing and processing system based on artificial intelligence simulation control parameters

Abstract

The invention relates to the technical field of artificial intelligence processing, in particular to a manufacturing and processing system based on artificial intelligence simulation control parameters. According to the manufacturing and processing system, three-dimensional space-time coding is carried out on vibration, temperature and deformation time-varying signals through a multi-mode sensing module; generating a composite feature body including an equipment rigidity distribution matrix, a thermal deformation gradient vector and a material residual stress tensor, and solving the problems of spatial-temporal asynchronization and physical field feature splitting of multi-source data; the parameter resolving module constructs a parameter incidence matrix based on a fractal neural network, combines dynamic constraint conditions of thermal deformation gradient vectors, performs multi-objective optimization on a rigidity distribution matrix and residual stress tensor, generates an initial parameter solution set for eliminating time-lag deviation, and breaks through the nonlinear coupling optimization bottleneck of a traditional static model; and the state dissociation module calls a historical wear feature library to construct a performance attenuation reference surface, and tool accumulated wear and instantaneous thermal deformation disturbance are synchronously inhibited through convolution operation and projection correction of a dynamic compensation factor.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence processing technology, and in particular to a manufacturing processing system based on artificial intelligence simulation control parameters. Background Art

[0002] Traditional manufacturing processing systems face multi-dimensional technical bottlenecks in complex dynamic processing scenarios. First, the fusion accuracy of multi-source heterogeneous data is insufficient: existing systems usually rely on a single type of sensor (such as vibration or temperature sensor) to collect data, and lack a collaborative mechanism for multimodal perception. For example, the waveform data of the vibration sensor array and the discrete temperature measurement values ​​of the temperature field detector are not synchronized in time stamps, resulting in the separation of the spatiotemporal correlation between the thermal deformation gradient vector and the equipment stiffness distribution matrix; this data asynchrony makes it impossible for the parameter solution model to accurately capture the thermomechanical coupling effect of the cutting process, such as the sudden change in residual stress of the material caused by the instantaneous change of the temperature field during high-speed processing. Traditional methods have difficulty in quantifying the interactive effect of heat conduction paths and stiffness attenuation because they do not construct a unified spatiotemporal coordinate system, which ultimately leads to significant time lag in cutting parameter optimization.

[0003] Secondly, the limitations of dynamic parameter optimization and compensation mechanisms are prominent. Existing technologies mostly use linear models for parameter calculation, but in actual processing, equipment stiffness attenuation, tool wear, and material anisotropic deformation show strong nonlinear correlations. For example, the rank change of the material residual stress tensor (revealing the stress concentration area) and the cumulative tool wear rate jointly affect cutting stability. However, because the traditional model does not introduce the adaptive weight adjustment mechanism of fractal neural networks, it cannot dynamically associate the axial attenuation characteristics of the stiffness distribution matrix with the eigenvalue direction of the residual stress, resulting in the solution of cutting speed and feed rate deviating from the Pareto frontier.

[0004] In addition, equipment performance compensation relies on fixed thresholds or mean adjustments: for example, cooling flow distribution does not incorporate the main direction component of the thermal deformation gradient vector, and tool wear compensation is based only on the instantaneous vibration amplitude, lacking progressive modeling of the relationship between historical wear curves and vibration spectrum mapping. Such defects make the machining process susceptible to the dual interference of instantaneous thermal shock (high-frequency disturbance) and low-frequency performance attenuation, exacerbating machining accuracy fluctuations and unexpected tool wear, and significantly increasing maintenance costs and energy consumption. Summary of the Invention

[0005] The purpose of the present invention is to provide a manufacturing and processing system based on artificial intelligence simulation control parameters to solve the problems raised in the above-mentioned background technology. The specific technologies include how to realize the spatiotemporal reference alignment and feature association of multi-source data of vibration, temperature and deformation to solve the parameter solution lag problem caused by the asynchronous data timestamp and spatial feature fragmentation of the traditional method; and how to construct a dynamic compensation factor based on historical wear patterns and real-time thermal deformation gradients to solve the cutting parameter offset problem caused by the traditional compensation mechanism ignoring the correlation between the cumulative wear rate of the tool and the direction of the heat conduction path.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The manufacturing and processing system based on artificial intelligence simulation control parameters includes a multimodal perception module, a parameter solution module and a state dissociation module, among which:

[0008] The multimodal perception module uses a phase difference analysis algorithm to calculate the waveform transmission delay of adjacent vibration sensors, and generates a device stiffness distribution matrix through wavefront reconstruction technology. Its axial dimension dynamically characterizes the attenuation characteristics of the device stiffness during processing, solving the problem of spatiotemporal asynchrony of vibration signals; the discrete temperature measurement data is input into the radial basis function to construct a continuous temperature field model, and the main direction component of the temperature gradient vector is extracted to form a thermal deformation gradient vector. Its modulus quantifies the thermal expansion amplitude, and its direction indicates the heat conduction path, thereby achieving accurate mapping of the temperature field and the thermal deformation direction; the strain waveform is converted through the strain energy density integral method, and the low-frequency residual stress component is extracted by combining time series decomposition to construct the material residual stress tensor. Its rank number reveals the stress concentration area, and the eigenvalue direction corresponds to the anisotropic deformation trend of the material, quantifying the spatial distribution characteristics of the residual stress.

[0009] The multimodal perception module uses wavelet transform to align the time series benchmarks of the stiffness distribution matrix, thermal deformation gradient vector and material residual stress tensor, and implements spatial feature association of multi-source data through convolutional neural network to generate a composite feature body with a unified space-time coordinate system, solving the data fragmentation problem in traditional methods.

[0010] The parameter calculation module analyzes the modulus and direction of the thermal deformation gradient vector, generates a set of boundary conditions including the maximum temperature rise threshold, axial thermal deformation compensation range, and cooling flow adjustment range, and encodes them as dynamic constraint parameters of the optimization problem to ensure that the cutting parameter solution conforms to the actual thermal-mechanical coupling effect.

[0011] The parameter calculation module constructs a parameter association matrix through a fractal neural network. The fractal neural network uses the axial stiffness attenuation rate of the equipment stiffness distribution matrix as the input layer and the eigenvalue direction of the material residual stress tensor as the hidden layer association variable. The node connection weights of the parameter association matrix are adaptively adjusted through the fractal dimension. The fractal iteration depth is dynamically determined by the residual stress tensor, achieving nonlinear dynamic matching of cutting speed, feed rate and cooling flow rate.

[0012] The parameter solution module maps the equipment stiffness distribution matrix to the cutting speed constraint domain, maps the material residual stress tensor to the feed rate adjustment domain, integrates the dynamic constraints of the thermal deformation gradient vector, uses the gradient descent method to solve the Pareto front, and generates an initial parameter solution set that eliminates time lag deviation, ensuring the global optimality of parameter optimization.

[0013] The performance attenuation reference surface construction unit in the state dissociation module retrieves the historical wear feature library of the equipment, phase-matches the historical wear curve with the real-time sensor signal through the time-space alignment algorithm, and constructs a performance attenuation reference surface with the cumulative wear rate as the normal vector and the vibration spectrum eigenvalue as the basis coordinate. Wavelet envelope analysis is used to extract the progressive wear pattern within the tool life cycle and quantify the dynamic relationship between wear and vibration.

[0014] The correction unit in the state dissociation module performs frequency-domain decomposition of the initial parameter solution set and the thermal deformation gradient vector through tensor convolution operations. After removing the high-frequency disturbance components, it generates a dynamic compensation factor containing the cutting parameter compensation coefficient and the thermal deformation correction weight, thus achieving the separation of transient thermal shock and low-frequency performance attenuation.

[0015] The correction unit orthogonally projects the dynamic compensation factor along the normal vector of the cumulative wear rate of the performance degradation reference surface, eliminating parameter offsets caused by tool wear through least-squares fitting. Wavelet threshold constraints are also applied to the thermal deformation gradient vector to suppress abnormal fluctuations caused by transient thermal shock. The corrected compensation factor is then reverse-mapped to the parameter association matrix via the weighted feedback channel of the fractal neural network. The node connection weights of the fractal dimension are dynamically adjusted to form an adaptive closed-loop control circuit, continuously optimizing parameter solution accuracy.

[0016] Compared with the prior art, the present invention has the following beneficial effects:

[0017] The multimodal sensing module performs three-dimensional spatiotemporal encoding on the time-varying signals of vibration, temperature, and deformation data, generating a composite feature volume containing the equipment stiffness distribution matrix, thermal deformation gradient vector, and material residual stress tensor. This enables spatiotemporal alignment of multi-source heterogeneous data and precise modeling of physical features.

[0018] The parameter solution module constructs a parameter correlation matrix based on a fractal neural network, integrates the dynamic constraints of the thermal deformation gradient vector, performs multi-objective optimization on the equipment stiffness distribution matrix and residual stress tensor, and generates an initial parameter solution set that eliminates time lag deviation, thereby improving the real-time performance and global optimality of cutting parameter solution.

[0019] The state dissociation module constructs a convolution operation and projection correction of the performance attenuation reference plane and the dynamic compensation factor to simultaneously suppress the accumulated equipment wear and instantaneous thermal deformation disturbance, forming a closed-loop feedback mechanism. It ultimately outputs an optimized parameter instruction set that simultaneously eliminates time lag deviation and performance attenuation, effectively improving machining accuracy and stability, extending tool life, and reducing maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the overall module of the present invention;

[0021] Figure 2 Schematic diagram of the state dissociation module unit of the present invention.

[0022] In the figure: 100, multimodal perception module; 200, parameter solution module; 300, state dissociation module; 301, performance degradation reference plane construction unit; 302, correction unit. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0024] Next, see Figure 1 The present invention provides a technical solution: a manufacturing processing system based on artificial intelligence simulation control parameters, including a multimodal perception module 100, a parameter solution module 200 and a state dissociation module 300.

[0025] After the multimodal sensing module 100 uses the vibration sensor array to collect the time domain vibration signal of each node, it calculates the waveform transmission delay of adjacent sensors through the phase difference analysis algorithm, and uses the wavefront reconstruction technology to spatially integrate the delay data to generate the equipment stiffness distribution matrix. The axial dimension of the matrix corresponds to the rate of change of the equipment structure stiffness, which is used to characterize the dynamic stiffness attenuation characteristics during the processing process.

[0026] The multimodal sensing module 100 inputs the discrete temperature measurement data of the temperature field detector into the spatial interpolation algorithm, constructs a continuous temperature field model through the radial basis function, and extracts the main direction component of the temperature gradient vector as the thermal deformation gradient vector. Its modulus represents the thermal expansion amplitude, and its direction indicates the heat conduction path, which is used to quantify the coupling effect of the spindle thermal elongation on the processing accuracy.

[0027] The multimodal sensing module 100 converts the strain waveform obtained by the deformation monitoring device through the strain energy density integration method, extracts the low-frequency residual stress component by combining time series decomposition, and constructs the material residual stress tensor. The rank of the material residual stress tensor reveals the stress concentration area inside the workpiece, and the direction of the eigenvalue corresponds to the anisotropic deformation trend of the material, providing material mechanical state input for cutting parameter optimization.

[0028] The multimodal perception module 100 uses a spatiotemporal alignment algorithm to align the stiffness distribution matrix, thermal deformation gradient vector and material residual stress tensor to coordinate systems, adopts wavelet transform to align the time series benchmark, and implements spatial feature association through a convolutional neural network. Finally, a composite feature body with a unified spatiotemporal coordinate system is generated, including the equipment stiffness distribution matrix, thermal deformation gradient vector and material residual stress tensor. This feature body provides a multi-dimensional working condition characterization basis for subsequent parameter solution through the coupled mapping relationship of stiffness-temperature-stress. Its tensor dimension retains the original feature space of each physical field, realizing the fusion characterization of working conditions of multi-source heterogeneous data.

[0029] To address the problem of parameter solution lag caused by asynchronous multi-source data and spatial feature fragmentation in traditional manufacturing systems, the multimodal perception module 100 generates the equipment stiffness distribution matrix (the axial dimension characterizes the dynamic stiffness attenuation characteristics) through the phase difference analysis algorithm and wavefront reconstruction technology of the vibration sensor array, and combines the radial basis function to construct a temperature field model to extract the thermal deformation gradient vector (the modulus quantifies the thermal expansion amplitude, and the direction indicates the heat conduction path), and the strain energy density integral method to extract the material residual stress tensor (the rank number reveals the stress concentration area, and the eigenvalue direction corresponds to the anisotropic deformation trend). It also uses wavelet transform to align the space-time benchmark and convolutional neural network to associate spatial features to generate a composite feature body with a unified space-time coordinate system, solving the problems of multimodal data asynchrony and physical field coupling modeling, and providing high-precision multi-dimensional working condition representation for subsequent parameter optimization.

[0030] After receiving the composite feature body output by the multimodal sensing module 100, the parameter solution module 200 extracts the thermal deformation gradient vector from the composite feature body as a dynamic constraint condition. By analyzing its modulus (characterizing the thermal expansion amplitude) and direction (indicating the heat conduction path), it generates a set of boundary conditions for the spindle thermal elongation, including the maximum temperature rise threshold, the axial thermal deformation compensation range, and the cooling flow adjustment range. The above constraints are encoded as boundary parameters of the optimization problem and integrated into the cutting parameter solution framework as thermal-mechanical coupling dynamic constraints.

[0031] The parameter solution module 200 constructs a hierarchical topology of a fractal neural network based on the axial stiffness attenuation rate (input layer) of the equipment stiffness distribution matrix and the eigenvalue direction of the material residual stress tensor (hidden layer associated variable). The node connection weights of the parameter association matrix are adaptively adjusted through the fractal dimension to dynamically match the nonlinear coupling relationship between cutting speed, feed rate, and cooling flow rate. The fractal iteration depth is determined by the rank of the stress tensor, ensuring that the network structure evolves synchronously with the complexity of the multi-physics field characteristics.

[0032] The parameter solution module 200 maps the stiffness distribution matrix to the cutting speed constraint domain (the correlation function between the stiffness attenuation rate and the cutting chatter threshold), and maps the material residual stress tensor to the feed rate adjustment domain (the balance equation between the material anisotropic deformation and the cutting edge load). At the same time, it integrates the thermo-mechanical coupling dynamic constraint conditions of the thermal deformation gradient vector, uses the gradient descent method to solve the Pareto front, and generates an initial parameter solution set that eliminates the time lag deviation. The elimination of the time lag deviation is achieved through the dynamic iteration of the weights of the fractal neural network, ensuring the real-time synchronization of the cutting parameter optimization and the changes in the equipment working conditions.

[0033] In order to solve the problem that traditional parameter optimization models cannot handle nonlinear coupling relationships, the parameter solution module 200 constructs a dynamic parameter association matrix based on a fractal neural network: the axial stiffness attenuation rate of the stiffness distribution matrix is ​​used as the input layer, and the eigenvalue direction of the material residual stress tensor is used as the hidden layer association variable. The node connection weights are adaptively adjusted through the fractal dimension (the fractal iteration depth is determined by the rank of the stress tensor), the stiffness distribution is mapped to the cutting speed constraint domain, the residual stress tensor is mapped to the feed rate adjustment domain, and the thermomechanical coupling dynamic constraint conditions of the thermal deformation gradient vector (maximum temperature rise threshold, axial compensation range and cooling flow range) are integrated. The gradient descent method is used to solve the Pareto front to generate an initial parameter solution set, thereby realizing nonlinear dynamic matching of cutting speed, feed rate and cooling flow, and eliminating the time lag deviation caused by the traditional static model.

[0034] like Figure 2 As shown, the performance attenuation reference plane construction unit 301 in the state dissociation module 300 retrieves the mapping relationship data between the tool wear amount and the spindle vibration spectrum from the pre-stored equipment historical wear feature library, uses the time-space alignment algorithm to phase-match the historical wear curve with the real-time sensor signal, and constructs a performance attenuation reference plane with the cumulative wear rate as the normal vector and the vibration spectrum eigenvalue as the base coordinate. This performance attenuation reference plane extracts the progressive wear pattern within the tool life cycle through wavelet envelope analysis, providing a quantitative benchmark for dynamic compensation;

[0035] The correction unit 302 in the state dissociation module 300 performs a tensor convolution operation on the initial parameter solution set output by the parameter solution module 200 and the thermal deformation gradient vector in the composite feature volume of the multimodal perception module 100. It removes high-frequency disturbance components through frequency domain decomposition, retains low-frequency components related to equipment stiffness attenuation and thermal-mechanical coupling effects, and generates a dynamic compensation factor containing cutting parameter compensation coefficients and thermal deformation correction weights. The tensor dimension strictly corresponds to the fractal structure of the parameter correlation matrix.

[0036] The correction unit 302 performs orthogonal projection correction on the dynamic compensation factor based on the performance attenuation reference plane: in the reference plane coordinate system, the compensation factor is projected along the direction of the normal vector of the cumulative wear rate, and the parameter offset caused by tool wear is eliminated by least squares fitting; at the same time, a wavelet threshold constraint is imposed on the thermal deformation gradient vector to suppress abnormal fluctuations caused by instantaneous thermal shock, and generate a final parameter instruction set that integrates time lag deviation elimination and progressive wear compensation; in the projection correction process, the correction amount of the compensation factor is reversely mapped to the parameter association matrix of the parameter solution module 200 through the weight feedback channel of the fractal neural network, and the node connection weights of its fractal dimension are dynamically adjusted to form a closed-loop control loop of "parameter optimization → compensation correction → weight iteration" to ensure the synchronous adaptability of the cutting parameter instruction set and the real-time working conditions of the equipment.

[0037] In order to solve the problem that equipment performance attenuation and instantaneous disturbances are difficult to compensate synchronously, the state dissociation module 300 constructs a performance attenuation reference plane by pre-storing a historical wear feature library (with the cumulative wear rate as the normal vector and the vibration spectrum eigenvalue as the base coordinate), and combines the tensor convolution operation to decompose the initial parameter solution set and the thermal deformation gradient vector in the frequency domain, and generates a dynamic compensation factor after stripping off the high-frequency disturbance; the tool wear parameter offset is eliminated by orthogonal projection correction along the normal vector of the performance attenuation reference plane, and wavelet threshold constraints are applied to suppress instantaneous thermal shock fluctuations, and finally a final parameter instruction set that integrates time lag deviation elimination and progressive wear compensation is generated; at the same time, the correction amount is reversely mapped to the parameter association matrix through the fractal neural network weight feedback channel to form a closed-loop control loop to ensure the real-time synchronization of cutting parameters and equipment working conditions, and solve the problem of insufficient stability of traditional compensation mechanisms that rely on single threshold or mean adjustment.

[0038] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A manufacturing system based on artificial intelligence simulation control parameters, characterized in that: The system comprises a multimodal perception module (100), a parameter calculation module (200) and a state dissociation module (300), wherein: The multimodal sensing module (100) is used to perform three-dimensional spatiotemporal encoding on time-varying signals collected by the vibration sensor array, the temperature field detector and the deformation monitoring device, and generate a composite feature body including a device stiffness distribution matrix, a thermal deformation gradient vector and a material residual stress tensor; The parameter calculation module (200) is connected to the output end of the multimodal perception module (100) and extracts a thermal deformation gradient vector from the composite feature body as a dynamic constraint condition; The parameter calculation module (200) constructs a parameter association matrix through a fractal neural network, performs multi-objective optimization of cutting parameters on the equipment stiffness distribution matrix and the material residual stress tensor, and generates an initial parameter solution set that eliminates time lag deviation; The state dissociation module (300) retrieves a pre-stored equipment history wear feature library to construct a performance attenuation reference plane containing a mapping relationship between tool wear and spindle vibration spectrum; performs a convolution operation on the initial parameter solution set and a thermal deformation gradient vector in a composite feature body to generate a dynamic compensation factor; and performs a projection correction on the dynamic compensation factor based on the performance attenuation reference plane to output a final parameter instruction set that simultaneously eliminates time lag deviation and equipment performance attenuation.

2. The manufacturing and processing system based on artificial intelligence simulation control parameters according to claim 1 is characterized in that: The specific process of the multimodal perception module (100) generating a composite feature body including a device stiffness distribution matrix, a thermal deformation gradient vector, and a material residual stress tensor includes: The waveform transmission delay of adjacent vibration sensors is calculated using a phase difference analysis algorithm, and the equipment stiffness distribution matrix is ​​generated through wavefront reconstruction technology. Its axial dimension characterizes the dynamic stiffness attenuation characteristics. The discrete temperature measurement data is input into the radial basis function to construct a continuous temperature field model. The main direction component of the temperature gradient vector is extracted to form the thermal deformation gradient vector. Its modulus represents the thermal expansion amplitude, and its direction indicates the heat conduction path. The strain waveform is converted by the strain energy density integral method, and the low-frequency residual stress component is extracted by combining time series decomposition to construct the material residual stress tensor. The rank of the material residual stress tensor reveals the stress concentration area, and the eigenvalue direction corresponds to the anisotropic deformation trend.

3. The manufacturing and processing system based on artificial intelligence simulation control parameters according to claim 1 is characterized in that: The parameter solving module (200) generates a boundary condition set including a maximum temperature rise threshold, an axial thermal deformation compensation range, and a cooling flow adjustment interval by analyzing the modulus and direction of the thermal deformation gradient vector, and encodes the boundary condition set as a boundary parameter of the optimization problem, which is integrated into the cutting parameter solving framework as a thermal-mechanical coupling dynamic constraint condition.

4. The manufacturing and processing system based on artificial intelligence simulation control parameters according to claim 1 is characterized in that: The fractal neural network uses the axial stiffness attenuation rate of the equipment stiffness distribution matrix as the input layer and the eigenvalue direction of the material residual stress tensor as the hidden layer association variable. The node connection weights of the parameter association matrix are adaptively adjusted through the fractal dimension. The fractal iteration depth is determined by the material residual stress tensor, and is used to achieve dynamic matching of the nonlinear coupling relationship between cutting speed, feed rate and cooling flow rate.

5. The manufacturing and processing system based on artificial intelligence simulation control parameters according to claim 4 is characterized in that: The state dissociation module (300) comprises a performance attenuation reference plane construction unit (301), which uses a time-space alignment algorithm to perform phase matching on a historical wear curve and a real-time sensor signal, constructs a performance attenuation reference plane with a cumulative wear rate as a normal vector and a vibration spectrum eigenvalue as a base coordinate, and extracts a progressive wear pattern within a tool life cycle through wavelet envelope analysis.

6. The manufacturing and processing system based on artificial intelligence simulation control parameters according to claim 1 is characterized in that: The state dissociation module (300) includes a correction unit (302), which performs frequency domain decomposition on the initial parameter solution set and the thermal deformation gradient vector through a tensor convolution operation, and generates a dynamic compensation factor including a cutting parameter compensation coefficient and a thermal deformation correction weight after stripping off the high-frequency disturbance component.

7. The manufacturing and processing system based on artificial intelligence simulation control parameters according to claim 2 is characterized in that: The multimodal perception module (100) uses wavelet transform to align the time series benchmarks of the stiffness distribution matrix, thermal deformation gradient vector and material residual stress tensor, implements spatial feature association through a convolutional neural network, and generates a composite feature body with a unified time-space coordinate system.

8. The manufacturing and processing system based on artificial intelligence simulation control parameters according to claim 4 is characterized in that: The fractal neural network maps the stiffness distribution matrix to the cutting speed constraint domain, maps the material residual stress tensor to the feed rate adjustment domain, integrates the thermo-mechanical coupling dynamic constraint conditions of the thermal deformation gradient vector, and uses the gradient descent method to solve the Pareto front to generate an initial parameter solution set.

9. The manufacturing and processing system based on artificial intelligence simulation control parameters according to claim 1 is characterized in that: The projection correction process includes: The dynamic compensation factor is orthogonally projected along the normal vector of the cumulative wear rate of the performance attenuation reference surface, and the parameter offset caused by tool wear is eliminated by least squares fitting. At the same time, a wavelet threshold constraint is applied to the thermal deformation gradient vector to suppress abnormal fluctuations caused by instantaneous thermal shock.

10. The manufacturing system based on artificial intelligence simulation control parameters according to claim 6, characterized in that: The correction unit (302) reversely maps the correction amount of the compensation factor to the parameter association matrix of the parameter solution module (200) through the weight feedback channel of the fractal neural network, dynamically adjusts the node connection weight of the fractal dimension, and forms a closed-loop control circuit.

Images