Aluminum alloy shell processing adaptive control method and system

Abstract

The application provides an aluminum alloy shell machining adaptive control method and system, relates to the technical field of adaptive control, constructs a digital twin model of the aluminum alloy shell, carries out machining process simulation based on the digital twin model, generates simulation machining data, determines a machining parameter limit interval according to the simulation machining data, collects multiple source sensing signals in real time, extracts a machining feature vector, judges whether an abnormal state occurs in a current machining stage according to the deviation between the machining feature vector and a predicted characteristic value output by the digital twin model, determines an adjustment amount of the machining parameter in response to the judgment result of the abnormal state, and executes adaptive adjustment of the machining parameter according to the adjustment amount. The application can realize accurate identification of an abnormal state in the aluminum alloy shell machining process and self-optimization adjustment of the machining parameter, and improves machining quality and a good product rate.

Description

Technical Field

[0001] This application relates to the field of adaptive control technology, and more specifically, to an adaptive control method and system for aluminum alloy housing processing. Background Technology

[0002] Previously, the machining process control of aluminum alloy shells mainly adopted a method based on offline process planning and fixed machining parameters. Operators set machining parameters such as spindle speed, feed rate, depth of cut, and clamping force according to process manuals or previous trial cut results, and kept these parameters unchanged during actual machining. In recent years, some machining equipment has introduced monitoring methods based on single sensor signals to monitor the machining status in real time and trigger alarms or shutdowns based on preset fixed thresholds. In addition, the application of digital twin technology in the machining field is mainly focused on early-stage process simulation and parameter optimization, used to determine a reasonable range of machining parameters before machining.

[0003] However, the existing technical solutions mentioned above, based on fixed parameters or single signal monitoring, are ill-suited to handle the complex and ever-changing abnormal conditions encountered during aluminum alloy shell machining, such as tool wear, cutting chatter, workpiece deformation, and cutting heat accumulation. Fixed threshold alarm methods cannot distinguish between instantaneous signal spikes and actual machining anomalies, easily leading to false alarms or missed alarms. Therefore, how to accurately identify abnormal states and automatically optimize machining parameters during aluminum alloy shell machining to improve machining quality and yield is a challenge facing the industry. Summary of the Invention

[0004] This application provides an adaptive control method and system for aluminum alloy shell processing, which can accurately identify abnormal states and automatically optimize processing parameters during the processing of aluminum alloy shells, thereby improving processing quality and yield.

[0005] In a first aspect, this application provides an adaptive control method for processing aluminum alloy housings, the adaptive control method comprising the following steps: A digital twin model of an aluminum alloy shell is constructed, and the processing is simulated based on the digital twin model to generate simulation processing data. The limit range of processing parameters is determined based on the simulation processing data. During the processing, multi-source sensor signals are collected in real time, and then processing feature vectors representing the current processing state are extracted. Based on the deviation between the processing feature vector and the predicted feature value output by the digital twin model, it is determined whether an abnormal state has occurred in the current processing stage; In response to the determination result of the abnormal state, the adjustment amount of the processing parameters is determined with the limit range of the processing parameters as the constraint boundary, and the adaptive adjustment of the processing parameters is performed according to the adjustment amount.

[0006] In this embodiment, a digital twin model of the aluminum alloy shell is constructed, and the processing simulation is performed based on the digital twin model to generate simulation processing data, specifically including: Obtain CAD design data and material property data for the aluminum alloy shell, and establish an initial geometric model; The initial geometric model is meshed to construct a digital twin model; Configure the various processing parameters to be optimized in the digital twin model, and set the initial range of each processing parameter; The simulation experiment granularity is set based on the experimental design method, and multiple sets of simulation parameter combinations are generated by discrete sampling within the initial range of each processing parameter. The digital twin model is driven by each set of simulation parameters to simulate the processing process and generate corresponding simulation processing data.

[0007] In this embodiment, the multi-source sensing signals include cutting force signals, tool vibration signals, acoustic emission signals, and workpiece clamping stress signals.

[0008] In this embodiment, determining whether an abnormal state has occurred in the current processing stage based on the deviation between the processing feature vector and the predicted feature value output by the digital twin model specifically includes: The actual processing parameters at the current moment are obtained, and the actual processing parameters are input into a pre-constructed processing proxy model, which drives the processing proxy model to output predicted feature values ​​corresponding to the current processing state; wherein, the processing proxy model is constructed using the response surface methodology based on the simulation data of the digital twin model; The processing feature vector is compared with the predicted feature value to determine the deviation vector; The deviation vectors are weighted and fused to obtain a comprehensive deviation index; Based on the comparison between the comprehensive deviation index and the preset warning threshold, it is determined whether an abnormal state has occurred in the current processing stage.

[0009] In this embodiment, determining whether an abnormal state has occurred in the current processing stage based on the comparison result between the comprehensive deviation index and the preset warning threshold specifically includes: The comprehensive deviation index is stored in a sliding time window of fixed length to form a deviation sequence; Within the sliding time window, short-term trend characteristic quantities and long-term trend characteristic quantities are determined respectively. The short-term trend characteristic quantity reflects the instantaneous rate of change of the most recent few moments, and the long-term trend characteristic quantity reflects the overall drift direction and intensity within the entire window. An abnormal state is determined to have occurred only when the comprehensive deviation index exceeds the warning threshold and the short-term trend feature and the long-term trend feature simultaneously indicate that the deviation is diverging or continuously increasing.

[0010] In this embodiment, the predicted feature values ​​include the predicted cutting force, vibration, acoustic emission, and stress distribution values ​​corresponding to the dimensions of the machining feature vector; the deviation vector includes the component deviation values ​​in each dimension; and the predicted feature values ​​output by the digital twin model have confidence intervals.

[0011] In this embodiment, responding to the determination result of the abnormal state, and using the limit range of the processing parameters as the constraint boundary, determining the adjustment amount of the processing parameters specifically includes: Based on the result of the abnormal status determination, the abnormal type of the current processing stage is determined; According to the type of anomaly, the corresponding adjustment strategy is matched from the preset processing rule base; Based on the adjustment strategy and the limit range of the processing parameters, the adjustment amount of the processing parameters is determined.

[0012] In this embodiment, determining the adjustment amount of the processing parameters based on the adjustment strategy and the limit range of the processing parameters specifically includes: The set of processing parameters that can be adjusted under the current anomaly type is determined based on the adjustment strategy; Based on the adjustment priority determined by the adjustment strategy, a corresponding priority weight is configured for each processing parameter that is allowed to be adjusted; An objective function is constructed with the goal of minimizing the deviation between the processing feature vector and the predicted feature value, and processing parameters with higher priority are adjusted first during the optimization process. Using the limit range of the processing parameters as constraints, local optimization is performed within the limit range of the processing parameters, and the adjustment amount of the processing parameters is output.

[0013] In this embodiment, the adjustment amount of the machining parameters includes at least one of the following: spindle speed adjustment, feed rate adjustment, depth of cut adjustment, and clamping force adjustment.

[0014] Secondly, this application provides an adaptive control system for aluminum alloy casing processing, used to execute an adaptive control method for aluminum alloy casing processing, the adaptive control system comprising: The processing simulation module is used to construct a digital twin model of the aluminum alloy shell, simulate the processing process based on the digital twin model, generate simulated processing data, and determine the limit range of processing parameters based on the simulated processing data. The feature extraction module is used to collect multi-source sensor signals in real time during the processing, and then extract processing feature vectors that represent the current processing state. The anomaly detection module is used to determine whether an abnormal state has occurred in the current processing stage based on the deviation between the processing feature vector and the predicted feature value output by the digital twin model. An abnormal adjustment module is used to respond to the judgment result of the abnormal state, determine the adjustment amount of the processing parameters with the limit range of the processing parameters as the constraint boundary, and perform adaptive adjustment of the processing parameters according to the adjustment amount.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: A digital twin model of an aluminum alloy shell is constructed. Based on this model, a processing simulation is performed to generate simulated processing data. The limit range of processing parameters is determined according to this simulated processing data. During processing, multi-source sensor signals are acquired in real time to extract processing feature vectors representing the current processing state. Based on the deviation between the processing feature vectors and the predicted feature values ​​output by the digital twin model, it is determined whether an abnormal state has occurred in the current processing stage. In response to the abnormal state determination, the adjustment amount of the processing parameters is determined using the limit range of the processing parameters as a constraint boundary. Adaptive adjustment of the processing parameters is then performed based on this adjustment amount.

[0016] Therefore, this application demonstrates the following: First, by constructing a digital twin model of the aluminum alloy shell and simulating the machining process, the processing quality response law under different combinations of processing parameters can be pre-understood before actual machining. Based on this, the limit range of each processing parameter can be determined, providing a safety boundary and constraint range for subsequent online adaptive adjustment. Second, by acquiring multi-source sensor signals in real time during machining and fusing them into a processing feature vector, a multi-dimensional perception of the aluminum alloy shell's machining state can be achieved, providing a data foundation for subsequent anomaly judgment. Third, by comparing the deviation between the actual machining feature vector and the predicted feature value output by the digital twin model, quantitative identification of machining anomalies is achieved, and the deviation vector provides a basis for anomaly type discrimination. Finally, by determining the anomaly type based on the anomaly state judgment result and matching the corresponding adjustment strategy from the preset machining rule library, local optimization is performed within the limit range of machining parameters with the goal of minimizing the deviation to determine the adjustment amount of the machining parameters. Then, the adjustment amount is output to each execution unit of the machine tool through the control system to perform adaptive adjustment, realizing a complete closed loop from anomaly identification to parameter control. This enables the adjusted machining parameters to quickly and accurately suppress the abnormal state and restore the machining process to stability.

[0017] In summary, the technical solution adopted in this application can accurately identify abnormal states and automatically optimize processing parameters during the aluminum alloy shell processing, thereby improving processing quality and yield. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this embodiment of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is an exemplary flowchart of the adaptive control method for aluminum alloy shell processing provided in this application; Figure 2 This is a schematic diagram of the device structure for processing an adaptive control system for an aluminum alloy shell according to the present application; Figure 3 This is a module structure diagram of the adaptive control system for aluminum alloy shell processing provided in this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0021] This application provides an adaptive control method and system for aluminum alloy shell processing. Its core is to construct a digital twin model of the aluminum alloy shell, simulate the processing process based on the digital twin model, generate simulated processing data, and determine the limit range of processing parameters according to the simulated processing data. During processing, multi-source sensor signals are acquired in real time to extract processing feature vectors representing the current processing state. Based on the deviation between the processing feature vectors and the predicted feature values ​​output by the digital twin model, it is determined whether an abnormal state has occurred in the current processing stage. In response to the abnormal state determination result, the adjustment amount of the processing parameters is determined using the limit range of processing parameters as a constraint boundary, and adaptive adjustment of the processing parameters is performed according to the adjustment amount.

[0022] Example 1: To better understand the above technical solution, the following will provide a detailed description of the technical solution in conjunction with the accompanying drawings and specific implementation methods. (Refer to...) Figure 1As shown, this figure is an exemplary flowchart of an adaptive control method for aluminum alloy shell processing according to this embodiment of the present application. The adaptive control method includes the following steps: In step S1, a digital twin model of the aluminum alloy shell is constructed, the processing is simulated based on the digital twin model, simulation processing data is generated, and the limit range of processing parameters is determined based on the simulation processing data.

[0023] In this embodiment, a digital twin model of the aluminum alloy shell is constructed, and the processing is simulated based on the digital twin model to generate simulation processing data. This can be achieved through the following steps: Obtain CAD design data and material property data for the aluminum alloy shell, and establish an initial geometric model; The initial geometric model is meshed to construct a digital twin model; Configure the various processing parameters to be optimized in the digital twin model, and set the initial range of each processing parameter; The simulation experiment granularity is set based on the experimental design method, and multiple sets of simulation parameter combinations are generated by discrete sampling within the initial range of each processing parameter. The digital twin model is driven by each set of simulation parameters to simulate the processing process and generate corresponding simulation processing data.

[0024] In practical implementation, firstly, the CAD design data and material property data of the aluminum alloy shell are acquired to establish an initial geometric model. This involves obtaining the CAD design data of the aluminum alloy shell to be processed, which includes information such as the shell's three-dimensional geometric contour, wall thickness distribution, feature dimensions, and tolerance requirements. Simultaneously, the material property data of the aluminum alloy shell is acquired, including at least: elastic modulus, Poisson's ratio, density, yield strength, tensile strength, coefficient of thermal expansion, thermal conductivity, and specific heat capacity. The CAD design data and material property data are then imported into a finite element preprocessing platform to establish the initial geometric model of the aluminum alloy shell. Secondly, the initial geometric model is meshed to construct a digital twin model. This involves meshing the initial geometric model, considering that aluminum alloy shells are typically thin-walled structures, using multi-layer solid elements or shell elements for discretization in the wall thickness direction, and quadrilateral elements for meshing in the planar direction. For feature areas, such as fillets, reinforcing ribs, and mounting holes, local mesh refinement is performed. After mesh generation, material properties are assigned to each element, boundary conditions are set, and a digital twin model is constructed. Then, the machining parameters to be optimized are configured in the digital twin model, and initial ranges for each parameter are set. These parameters include, but are not limited to, spindle speed, feed rate, depth of cut, and clamping force. The initial ranges for each machining parameter are set based on machine tool performance, tool specifications, and the machining process specifications for the aluminum alloy material.

[0025] Furthermore, in practical implementation, the granularity of simulation experiments is set based on the experimental design method. Discrete sampling is performed within the initial intervals of each machining parameter to generate multiple sets of simulation parameter combinations. That is, the experimental design method can adopt Latin hypercube design, which performs discrete sampling within the initial intervals of each machining parameter. Latin hypercube design can cover the entire parameter space well with fewer experiments and is suitable for high-dimensional parameter scenarios. The number of sampling points is set to generate multiple sets of simulation parameter combinations. The number of sampling points should be determined according to the number of machining parameters and the desired fitting accuracy, generally not less than 10 times the number of parameters. For example, for 4 machining parameters, the number of sampling points is set to 50-100 sets. Finally, the digital twin model is driven by each set of simulation parameter combinations to simulate the machining process and generate corresponding simulated machining data. That is, in the cutting dynamics simulation, by setting the tool-workpiece contact conditions, friction model, and chip separation criteria, the cutting force, torque, and vibration response under different machining parameter combinations are calculated using explicit or implicit finite element solvers. In the thermo-mechanical coupling simulation, the transfer and dissipation of cutting heat among the workpiece, tool, and chips are considered, and the temperature field evolution and thermal deformation of the machining area are calculated. After the simulation is completed, the simulation machining data corresponding to each combination of simulation parameters are recorded. The simulation machining data includes: cutting forces under each combination of simulation parameters, including tangential force, radial force, and axial force; cutting temperature; workpiece deformation; surface residual stress; and predicted surface roughness.

[0026] In practical implementation, the limit range of machining parameters is determined based on simulation machining data. Specifically, based on the design drawings and process requirements of the aluminum alloy shell, the machining quality indicators and their corresponding tolerance ranges are determined. Machining quality indicators include at least: surface roughness, dimensional accuracy, and thin-wall deformation. Data fitting is performed on the simulation-generated machining data to establish a surrogate model of the machining process between machining parameters and machining quality indicators, i.e., a response surface model. The regression relationship between each quality indicator and machining parameters is established using the response surface method. Using spindle speed, feed rate, depth of cut, and clamping force as independent variables, and surface roughness, dimensional error, and maximum deformation as response variables, second-order polynomial response surface models of each quality indicator with respect to each machining parameter are established. These models include constant terms, first-order terms of each parameter, second-order terms of each parameter, and interaction terms between parameters. The least squares method is used to perform regression fitting on the simulation data, and the coefficient of determination and adjusted coefficient of determination are used to evaluate the model fitting accuracy. When the coefficient of determination is greater than 0.95 and the adjusted coefficient of determination is greater than 0.90, the response surface model is considered to accurately describe the mapping relationship between each machining parameter and quality indicator and can be used for subsequent feasible region analysis. Based on the response surface models of each quality index, feasible subspaces satisfying a single quality constraint are solved in the parameter space. Taking surface roughness constraint as an example: the constraint condition is set as surface roughness not exceeding the maximum allowable value. All parameter combinations satisfying this constraint are searched within the initial parameter interval to form the feasible subspace for surface roughness. Using the Monte Carlo sampling method, a large number of parameter points are randomly generated in the parameter space. The surface roughness value of each parameter point is predicted by the response surface model. Parameter points satisfying the constraint condition are marked as feasible points, and the boundary surface of the feasible subspace for surface roughness is fitted to form the boundary surface. Similarly, feasible subspaces satisfying dimensional accuracy constraints and deformation constraints are solved separately. The intersection of each single-constraint feasible subspace is taken to obtain the comprehensive feasible region that simultaneously satisfies all machining quality constraints. The projection boundary of the comprehensive feasible region in the parameter space constitutes the limit interval of each machining parameter. To ensure machining safety, a safety margin is set by further shrinking the boundaries of the comprehensive feasible region. This margin shrinks the boundaries of each parameter within the comprehensive feasible region by a predetermined percentage, such as 5% to 10%, as the final determined limit range of machining parameters. The safety margin is used to compensate for uncertainties such as model errors, material property fluctuations, and tool wear that exist in actual machining.

[0027] In step S2, during the processing, multi-source sensor signals are collected in real time, and then processing feature vectors that characterize the current processing state are extracted.

[0028] In practical implementation, during the machining process, multi-source sensor signals are acquired in real time, and then machining feature vectors characterizing the current machining state are extracted. Specifically, during the machining of the aluminum alloy shell, multiple types of sensors arranged on the machine tool and workpiece are used to acquire multi-source sensor signals in real time. These multi-source sensor signals include cutting force signals, tool vibration signals, acoustic emission signals, and workpiece clamping stress signals. The cutting force signal is acquired through a three-dimensional force sensor mounted on the worktable. This sensor can simultaneously acquire the dynamic cutting force components of the tool in the feed direction, depth of cut direction, and radial direction. The tool vibration signal is acquired through a piezoelectric accelerometer mounted on the spindle bearing housing. This accelerometer acquires the vibration acceleration signals of the spindle and tool along three orthogonal directions during machining. The acoustic emission signal is acquired through an acoustic emission sensor mounted on the spindle housing. This sensor senses the elastic stress waves generated in the machining area due to material shearing, tool wear, and chip collision. The workpiece clamping stress signal is acquired through pressure sensors positioned on the surface of the aluminum alloy shell in contact with the fixture. Thin-film pressure sensors are installed at the contact surfaces between the aluminum alloy shell edge and each clamping unit to collect the contact pressure values ​​at each clamping point in real time. The acquisition clocks of all these sensors are synchronized using a unified synchronous controller to ensure strict alignment of the multi-source sensor signals in the time dimension. The acquired multi-source sensor signals are then preprocessed to remove noise interference and outliers. This preprocessing includes at least denoising and filtering.

[0029] In addition, in practical implementation, the preprocessed sensor signals are converted from time-domain signals to time-frequency domain signals. Wavelet packet transform is used to perform time-frequency decomposition on each preprocessed signal, specifically selecting the db5 wavelet from the Daubechies wavelet family as the mother wavelet, with the decomposition level set to three to five levels, to obtain the wavelet packet coefficients of each sensor signal in multiple frequency bands. In the time-frequency domain, multi-dimensional feature parameters are extracted from each frequency band component of each sensor signal, including energy features, peak features, and spectral features. Energy features are characterized by calculating the sum of squares of the wavelet packet coefficients in each frequency band to represent the total energy of the signal within each frequency band. Energy features reflect the power distribution of the signal in different frequency bands. Peak features are characterized by extracting the maximum absolute value of the wavelet packet coefficients in each frequency band, reflecting the instantaneous impact intensity of the signal. Spectral features are characterized by calculating the power spectral density distribution and spectral centroid within each frequency band, with the spectral centroid reflecting the concentrated frequency position of the signal's spectral energy. Z-score normalization is used to normalize each feature component. Specifically, for each feature dimension, its historical mean is subtracted and then divided by its historical standard deviation. The historical mean and standard deviation are obtained statistically from stable cutting segment data within the first preset time window after machining begins. All normalized feature components are concatenated in a preset order to form an initial fused feature vector. The dimension of the initial fused feature vector is equal to the sum of the number of feature components extracted from each sensor signal. Principal component analysis is then performed on the initial fused feature vector to reduce the feature dimension to a preset target dimension. The dimensionality-reduced feature vector is the machining feature vector representing the current machining state.

[0030] In step S3, based on the deviation between the processing feature vector and the predicted feature value output by the digital twin model, it is determined whether an abnormal state has occurred in the current processing stage.

[0031] In this embodiment, determining whether an abnormal state has occurred in the current processing stage based on the deviation between the processing feature vector and the predicted feature value output by the digital twin model can be achieved through the following steps: The actual processing parameters at the current moment are obtained, and the actual processing parameters are input into a pre-constructed processing proxy model, which drives the processing proxy model to output predicted feature values ​​corresponding to the current processing state; wherein, the processing proxy model is constructed using the response surface methodology based on the simulation data of the digital twin model; The processing feature vector is compared with the predicted feature value to determine the deviation vector; The deviation vectors are weighted and fused to obtain a comprehensive deviation index; Based on the comparison between the comprehensive deviation index and the preset warning threshold, it is determined whether an abnormal state has occurred in the current processing stage.

[0032] In practical implementation, firstly, the actual machining parameters at the current moment are obtained and input into a pre-built machining process proxy model. This model drives the machining process proxy model to output predicted feature values ​​corresponding to the current machining state. That is, the machining process proxy model is constructed using response surface methodology based on simulation data from the digital twin model. The machining parameters actually being used in the machine tool's operation at the current moment are obtained. These parameters include at least the current spindle speed, feed rate, depth of cut, and clamping force. The actual machining parameters are then input into the pre-built machining process proxy model, driving it to perform a forward prediction calculation and output predicted feature values ​​corresponding to the current machining state. These predicted feature values ​​include predicted cutting force, vibration, acoustic emission, and stress distribution values ​​corresponding to each dimension of the machining feature vector. These are the expected sensing features predicted by the digital twin model under the current machining parameter conditions, serving as a benchmark for assessing whether the actual machining state deviates from the expected state. Secondly, the machining feature vector is compared with the predicted feature values ​​to determine the deviation vector. That is, each component of the deviation vector is equal to the difference between each dimension component of the machining feature vector and the corresponding dimension component of the predicted feature value. The deviation vector reflects the degree of deviation between the actual machining state and the predicted state of the digital twin model. If all components of the deviation vector are close to zero, it indicates that the actual machining state is consistent with the predicted state, and the current machining is normal. If some components deviate significantly, it indicates that there may be abnormal factors in the actual machining process. The sign of each component's deviation value also reflects the direction of deviation; for example, a positive deviation indicates that the actual value is higher than the predicted value, and a negative deviation indicates that the actual value is lower than the predicted value. Then, the deviation vector is weighted and fused to obtain a comprehensive deviation index, that is, each component of the deviation vector is assigned a weight coefficient and then weighted and summed. The weight coefficients are set according to the following principles: for key signal dimensions that have a greater impact on machining quality under the current machining state, a larger weight is assigned; for signal dimensions that have a smaller impact, a smaller weight is assigned. All weight coefficients are non-negative, and the sum of the weight coefficients of all dimensions is 1. The weight coefficients are dynamically adjusted according to the current machining stage; that is, based on the current tool position and the surface accuracy requirements of the corresponding machining area in the CAD design data, the current machining stage is automatically identified as either roughing or finishing. When identified as roughing, the weights of cutting force deviation and vibration deviation are increased. When the machining stage is identified, the weights of acoustic emission deviation and stress distribution deviation are increased. Specific values ​​can be set by expert experience. For example, in the roughing stage, the weights for cutting force deviation are 0.35, vibration deviation is 0.30, acoustic emission deviation is 0.20, and stress distribution deviation is 0.15; in the finishing stage, these weights are adjusted to 0.15, 0.15, 0.35, and 0.35 respectively. Finally, based on the comparison between the comprehensive deviation index and the preset warning threshold, it is determined whether an abnormal state has occurred in the current machining stage. This process can be described in detail in the following steps.

[0033] It should be noted that the predicted feature values ​​output by the digital twin model include confidence intervals, which reflect the range of uncertainty in the predicted values ​​of the surrogate model for the processing procedure. When constructing a surrogate model for the processing procedure using the response surface methodology, the simulation data itself contains numerical noise, and the surrogate model is an approximate fit to the simulation data. Therefore, the predicted values ​​of the surrogate model at each parameter point have a certain degree of uncertainty, which is quantified by the confidence intervals of the predicted feature values. During the construction of the surrogate model, for each combination of simulation parameters, the corresponding simulation processing data is recorded. After completing the regression fitting of the response surface model, for any given processing parameter point, the response surface model outputs the predicted feature value at that point. Simultaneously, the confidence interval of the predicted feature value is calculated using the statistical theory of regression analysis. That is, the upper and lower limits of the confidence interval for each predicted response value are calculated based on the regression residuals, the design matrix, and a given confidence level, such as 95%.

[0034] It should be noted that when comparing the processing feature vector with the predicted feature value, each component of the processing feature vector, i.e., the measured feature value, is also compared with the predicted feature value to determine whether the measured feature value falls within the confidence interval of the corresponding predicted feature value. If the measured feature values ​​of each dimension fall within the confidence interval of the corresponding predicted feature value, the current processing state is determined to be normal. When the measured feature value of a certain dimension exceeds the confidence interval of the corresponding predicted feature value, but the comprehensive deviation index has not exceeded the preset warning threshold, the current processing state is determined to be potentially abnormal. The current processing trajectory point is marked and the subsequent sampling frequency is increased to achieve close tracking of the processing state in that area, while avoiding premature triggering of unnecessary adjustments. The basis for the determination is that exceeding the confidence interval means that the actual processing state deviates from the normal fluctuation range that the processing process proxy model can explain in a statistical sense, but the degree of deviation is not enough to trigger the warning threshold. Therefore, it is marked as a potentially abnormal state and monitoring is strengthened.

[0035] It should be noted that the confidence interval is specifically applied in the abnormal state determination process as follows: The actual extracted processing feature vector at the current moment is compared with the predicted feature value output by the surrogate model. If all components of the processing feature vector fall within the confidence interval of the corresponding predicted feature value, the current processing state is determined to be normal, the current processing parameters remain unchanged, and no warning or adjustment operation is triggered. If at least one component of the processing feature vector exceeds the confidence interval of the corresponding predicted feature value, but the comprehensive deviation index has not yet exceeded the preset warning threshold, the current processing state is determined to be potentially abnormal. When responding to the determination of a potential abnormal state: Mark the current processing trajectory point, record the current processing parameters, processing feature vector, deviation vector, and timestamp in the potential abnormality log, and add a mark at the corresponding position in the digital twin model of the processing trajectory; increase the sampling frequency of signal acquisition and processing from the first sampling frequency in normal mode to the second sampling frequency in enhanced mode, where the second sampling frequency is higher than the first sampling frequency. For example, the first sampling frequency is 1 kHz, and the second sampling frequency is 2 kHz. Simultaneously, start the potential abnormality timer. During enhanced sampling monitoring, if the comprehensive deviation index falls back to the confidence interval for several consecutive sampling times, and the short-term trend characteristic turns negative, it is determined that the potential abnormality has been eliminated, exiting the potential abnormality state and returning to the normal state, with the sampling frequency returning to the first sampling frequency. If the comprehensive deviation index continues to increase and exceeds the warning threshold, it is determined that the potential abnormality has developed into a confirmed abnormality, entering the abnormal state determination process and triggering subsequent adjustment amount determination and execution steps. If the potential abnormal state continues to exceed the preset second duration threshold and the comprehensive deviation index still does not exceed the warning threshold but also does not fall back to the confidence interval, a warning prompt is triggered to notify the operator. When the comprehensive deviation index exceeds the warning threshold, and both the short-term trend characteristic and the long-term trend characteristic indicate that the deviation is diverging or continuously increasing, it is determined that an anomaly has occurred in the current processing stage, and the subsequent adjustment amount determination and execution process begins.

[0036] In this embodiment, based on the comparison result between the comprehensive deviation index and the preset warning threshold, it is determined whether an abnormal state has occurred in the current processing stage. Specifically, this can be achieved through the following steps: The comprehensive deviation index is stored in a sliding time window of fixed length to form a deviation sequence; Within the sliding time window, short-term trend characteristic quantities and long-term trend characteristic quantities are determined respectively. The short-term trend characteristic quantity reflects the instantaneous rate of change of the most recent few moments, and the long-term trend characteristic quantity reflects the overall drift direction and intensity within the entire window. An abnormal state is determined to have occurred only when the comprehensive deviation index exceeds the warning threshold and the short-term trend feature and the long-term trend feature simultaneously indicate that the deviation is diverging or continuously increasing.

[0037] In practical implementation, firstly, the comprehensive deviation index is stored in a fixed-length sliding time window to form a deviation sequence. That is, a fixed-length sliding time window is preset, the length of which is measured by the number of sampling moments; for example, the window length is set to the most recent fifty sampling moments. After calculating the comprehensive deviation index at each sampling moment, the comprehensive deviation index is stored in the sliding time window to form a deviation sequence. Then, within the sliding time window, short-term trend features and long-term trend features are determined. The short-term trend features reflect the instantaneous rate of change of the most recent few moments, while the long-term trend features reflect the overall drift direction and intensity within the entire window. That is, the comprehensive deviation index of the most recent few moments in the sliding time window is taken, where the number of data points used to calculate the short-term trend is less than the length of the entire sliding time window. For example, the sliding time window length is fifty moments, and the comprehensive deviation index of the most recent five moments is used to calculate the short-term trend. The least squares method is used to perform linear fitting on the data points used for short-term trend calculation to obtain the slope of the fitted line, which is the short-term trend feature. When the short-term trend characteristic is positive and large, it indicates that the deviation has been rising rapidly in recent times, meaning the deviation is increasing rapidly. When the short-term trend characteristic is negative, it indicates that the deviation has been falling in recent times, meaning the deviation is narrowing. When the short-term trend characteristic is close to zero, it indicates that the deviation has remained stable recently. Next, the comprehensive deviation index for all moments in the sliding time window is taken, and a least squares method is used to perform a linear fit on all data points. The slope of the fitted line is the long-term trend characteristic. When the long-term trend characteristic is positive, it indicates that the comprehensive deviation index has been rising overall over the entire window length, meaning there is a long-term drift with continuously increasing deviation. When the long-term trend characteristic is negative, it indicates that the deviation has been falling overall. When the long-term trend characteristic is close to zero, it indicates that there is no significant long-term drift in deviation over the entire window.

[0038] Furthermore, in practical implementation, an abnormal state is determined only when the comprehensive deviation index exceeds the warning threshold, and both the short-term and long-term trend characteristics indicate that the deviation is diverging or continuously increasing. That is, after calculating the short-term and long-term trend characteristics, a joint determination of the abnormal state is performed. The current comprehensive deviation index exceeds the preset warning threshold, the short-term trend characteristic is greater than the short-term trend judgment threshold, and the long-term trend characteristic is greater than the long-term trend judgment threshold. Both the short-term and long-term trend judgment thresholds are preset positive numbers. An abnormal state is determined only when both conditions are met simultaneously—that is, when the current deviation level exceeds the threshold and shows a continuously increasing trend on both the recent and long-term time scales. Specifically, the short-term trend characteristic is determined as follows: if the short-term trend characteristic is greater than the first preset threshold, the short-term trend characteristic deviation is determined to be diverging or continuously increasing; otherwise, the short-term divergence condition is not met. The first preset threshold is a positive number, preferably ranging from one-thousandth to five-thousandths of the full-scale comprehensive deviation index, and can be specifically set based on experimental data analysis. The criteria for determining the long-term trend characteristic are set as follows: if the long-term trend characteristic is greater than the second preset threshold, the long-term trend indicator deviation is determined to be divergent or continuously increasing; if the long-term trend characteristic is less than or equal to the second preset threshold, the long-term trend is determined not to meet the divergence condition. The second preset threshold is a positive number, and its value ranges from one ten-thousandth to one thousandth of the full scale of the comprehensive deviation index, which can be specifically set by experimental data analysis.

[0039] In step S4, in response to the determination result of the abnormal state, the adjustment amount of the processing parameters is determined with the limit range of the processing parameters as the constraint boundary, and the adaptive adjustment of the processing parameters is performed according to the adjustment amount.

[0040] In this embodiment, in response to the determination result of the abnormal state, the adjustment amount of the processing parameters is determined using the limit range of the processing parameters as the constraint boundary. Specifically, this can be achieved through the following steps: Based on the result of the abnormal status determination, the abnormal type of the current processing stage is determined; According to the type of anomaly, the corresponding adjustment strategy is matched from the preset processing rule base; Based on the adjustment strategy and the limit range of the processing parameters, the adjustment amount of the processing parameters is determined.

[0041] In practical implementation, based on the abnormal state determination results, the abnormal type of the current machining stage is determined. These abnormal types include: tool wear abnormality, cutting vibration abnormality, workpiece deformation abnormality, and cutting heat accumulation abnormality. The deviation vector is input into a pre-trained abnormality classifier. The abnormality classifier uses the deviation values ​​of each dimension component of the deviation vector and their changing trends as input features, and outputs the type label of the current abnormality and its corresponding confidence level. For example, for tool wear abnormality, since tool wear leads to a gradual increase in cutting force and an increase in acoustic emission signal intensity, the corresponding deviation vector shows a continuous positive increase in cutting force deviation, and a synchronous increase in acoustic emission deviation. For cutting vibration abnormality, since the vibration amplitude increases sharply when chatter occurs in the machining system, the corresponding deviation vector shows a significant increase in vibration deviation components accompanied by high-frequency components. For workpiece deformation abnormality, since the thin-walled aluminum alloy shell undergoes elastic deformation under the action of cutting force, the corresponding deviation vector shows an increase in stress distribution deviation, and the direction of cutting force deviation is consistent with the direction of deformation. For abnormal heat accumulation during cutting, the material softens due to increased cutting temperature, resulting in an abnormality in both acoustic emission and stress distribution. Then, based on the anomaly type, a corresponding adjustment strategy is matched from a pre-defined machining rule base. For example, when the anomaly type is abnormal tool wear, the corresponding adjustment strategy is: allow adjustment of feed rate and depth of cut, prohibit adjustment of spindle speed; adjustment priority is feed rate over depth of cut; adjustment direction constraint is feed rate decreasing and depth of cut decreasing. When the anomaly type is abnormal cutting vibration, the corresponding adjustment strategy is: allow adjustment of spindle speed, prohibit adjustment of depth of cut; adjustment priority is spindle speed as the only adjustment parameter; adjustment direction constraint is spindle speed adjusting to avoid the resonant frequency range. When the anomaly type is abnormal workpiece deformation, the corresponding adjustment strategy is: allow adjustment of clamping force and feed rate, prohibit adjustment of depth of cut; adjustment priority is clamping force over feed rate; adjustment direction constraint is clamping force increasing and feed rate decreasing. When the anomaly type is cutting heat accumulation anomaly, the corresponding adjustment strategy is as follows: feed rate and spindle speed can be adjusted, but depth of cut adjustment is prohibited; the adjustment priority is feed rate over spindle speed; the adjustment direction constraint is feed rate to increase, and spindle speed to increase or decrease based on the heat source characteristics. The machining rule base is an updatable knowledge base. After each machining operation, the anomaly type, the adopted adjustment strategy, and the adjusted machining quality data are associated and stored in the machining rule base.When a certain type of anomaly occurs repeatedly, the historical processing records of that anomaly type are retrieved from the processing rule base. If the success rate of a certain adjustment strategy in the historical records reaches a preset threshold, such as a success rate of more than 80%, then the strategy will be matched first when the same anomaly type is encountered in the future. Finally, based on the adjustment strategy, the adjustment amount of the processing parameters is determined. This process can be described in detail by the following steps.

[0042] In this embodiment, the adjustment amount of the processing parameters is determined based on the adjustment strategy and the limit range of the processing parameters. Specifically, this can be achieved through the following steps: The set of processing parameters that can be adjusted under the current anomaly type is determined based on the adjustment strategy; Based on the adjustment priority determined by the adjustment strategy, a corresponding priority weight is configured for each processing parameter that is allowed to be adjusted; An objective function is constructed with the goal of minimizing the deviation between the processing feature vector and the predicted feature value, and processing parameters with higher priority are adjusted first during the optimization process. Using the limit range of the processing parameters as constraints, local optimization is performed within the limit range of the processing parameters, and the adjustment amount of the processing parameters is output.

[0043] In practical implementation, firstly, the set of machining parameters that can be adjusted under the current anomaly type is determined according to the adjustment strategy. Specifically, when the anomaly type is tool wear anomaly, the adjustable machining parameters are feed rate and depth of cut; adjusting the spindle speed is prohibited because changing the spindle speed may alter the dynamic characteristics of the cutting system, thus exacerbating tool wear. When the anomaly type is cutting vibration anomaly, the adjustable machining parameter is spindle speed; adjusting the depth of cut is prohibited because changing the depth of cut directly affects the cutting force amplitude and may induce vibrations in other modes, while adjusting the spindle speed can effectively avoid the resonant frequency range. When the anomaly type is workpiece deformation anomaly, the adjustable machining parameters are clamping force and feed rate; adjusting the depth of cut is prohibited to avoid changing the magnitude of the cutting force and thus exacerbating deformation. When the anomaly type is cutting heat accumulation anomaly, the adjustable machining parameters are feed rate and spindle speed, while depth of cut adjustment is prohibited to reduce cutting temperature while avoiding changes in cutting force amplitude. Secondly, based on the adjustment priority determined by the adjustment strategy, corresponding priority weights are assigned to each adjustable machining parameter. That is, the adjustment strategy predefines the priority order of adjustable parameters under each anomaly type. Priority weights are used to control the adjustment order and magnitude of each parameter during subsequent optimization. Specifically, the parameter with the highest priority is given the greatest degree of adjustment freedom during optimization, and this parameter will be adjusted first to suppress deviations. Lower priority parameters are only allowed to participate in adjustment if, after adjusting higher priority parameters to their limit range boundaries, the objective function still cannot be reduced below a preset threshold. For example, when the anomaly type is cutting vibration anomaly, the adjustable machining parameter is spindle speed, which is configured as the highest priority. When the anomaly type is tool wear anomaly, the allowed machining parameters to be adjusted are feed rate and depth of cut, with feed rate configured as the first priority and depth of cut as the second priority. This means that reducing the feed rate is the preferred method to reduce the cutting load; only when reducing the feed rate to the lower limit of its limit range is insufficient to suppress the deviation should the depth of cut be reduced. When the anomaly type is workpiece deformation anomaly, the allowed machining parameters to be adjusted are clamping force and feed rate, with clamping force configured as the first priority and feed rate as the second priority.

[0044] Furthermore, in practical implementation, the objective function is constructed with minimizing the deviation between the processed feature vector and the predicted feature value as the optimization objective. During the optimization process, high-priority processing parameters are adjusted first; that is, the number of adjustable processing parameters is set to k. The deviation vector between the current processed feature vector and the predicted feature value includes components such as cutting force deviation, vibration deviation, acoustic emission deviation, and stress distribution deviation. The objective function is the weighted sum of squares of the deviation values ​​of each dimension's components. This is achieved by squaring the deviation value of each dimension, multiplying it by the corresponding weight coefficient, and then summing the results for all dimensions to obtain the objective function value. The weight coefficients for each dimension are dynamically set based on whether the current processing stage is roughing or finishing. During roughing, the weight coefficients corresponding to cutting force deviation and vibration deviation are set to larger values; during finishing, the weight coefficients corresponding to acoustic emission deviation and stress distribution deviation are set to larger values. During the optimization process, priority weights constrain and guide the optimization process in the following ways: For high-priority processing parameters, in each step of the iterative optimization, the parameter is adjusted first within the current limit interval in a direction that reduces the objective function; only when a high-priority parameter has been adjusted to the boundary of its limit interval, and the objective function has not yet fallen below a preset threshold, is the next priority processing parameter allowed to participate in the adjustment; finally, using the limit interval of the processing parameters as a constraint, local optimization is performed within the limit interval, and the adjustment amount of the processing parameters is output. That is, for each processing parameter that is allowed to be adjusted, its value range is constrained by the upper and lower limits of the limit interval. For processing parameters that are prohibited from being adjusted, their values ​​are fixed at the current value and do not participate in the optimization process. During the optimization process, an iterative adjustment method based on sensitivity analysis is used for local optimization, that is, the surrogate model is used to calculate the sensitivity coefficient of the objective function to each allowable adjustable processing parameter at the current processing parameter point. The adjustment direction of each adjustable processing parameter is determined by the sign of the sensitivity coefficient. A positive sensitivity coefficient indicates that increasing the parameter will increase the objective function, and the adjustment should be made in the direction of decreasing. A negative sensitivity coefficient indicates that decreasing the parameter will increase the objective function, and the adjustment should be made in the direction of increasing. The adjustment step size is allocated based on the absolute value of the sensitivity coefficient and the priority weight of each parameter, with higher priority parameters receiving greater weight in the step size allocation. After each iteration, the new processing parameters are input into the surrogate model for prediction verification. If the overall prediction deviation decreases, the iteration continues; otherwise, the step size is reduced and readjusted until the convergence condition is met. During the optimization process, if the highest priority processing parameter has been adjusted to its limit interval boundary, but the objective function has not yet fallen below the preset threshold, the next priority processing parameter is included in the adjustment range, and iterative optimization continues. This process continues until the objective function falls below the preset threshold, or all adjustable processing parameters have been adjusted to their limit interval boundaries, or the preset maximum number of iterations is reached.When the iteration terminates, the difference between the current adjustable processing parameter value and the initial value is output as the corresponding adjustment amount. The adjustment amount for processing parameters that are prohibited from adjustment is zero.

[0045] It should be noted that during the optimization process, if the highest priority processing parameter has been adjusted to its limit range boundary, but the objective function has not yet fallen below the preset threshold, then the next priority processing parameter is included in the adjustment range, and iterative optimization continues. This process continues until either of the following termination conditions is met: Termination Condition 1: The objective function value falls below the preset acceptable deviation threshold. In this case, optimization is considered successful, and the current adjustment amount is output. Termination Condition 2: All adjustable processing parameters have been adjusted to their respective limit range boundaries, but the objective function value is still higher than the acceptable deviation threshold. In this case, the current anomaly is determined to be unavoidable through parameter adjustment, triggering the system saturation handling mechanism. The system saturation handling mechanism includes: recording the current anomaly type, the current adjustment amount of each processing parameter, and the objective function value; generating a saturation status report; and issuing a warning signal through the human-machine interface to alert the operator that the current anomaly exceeds the system's autonomous adjustment capability. Based on the current anomaly type and processing stage, it is determined whether to allow entry into a degraded operation mode. If the current machining stage or critical dimension machining area is in a degraded operation phase, which could lead to workpiece scrap, then degraded operation is rejected, and a safety shutdown is executed. The spindle speed and feed rate are slowly reduced to zero, while the clamping force remains constant to prevent workpiece loosening. If the current machining stage or non-critical area is in a degraded operation phase, and the current deviation, although not reduced to an acceptable level, has not continued to worsen (i.e., the long-term trend characteristic is less than half of the long-term trend judgment threshold), then degraded operation mode can be entered. Machining continues with the limit parameter values ​​within the current allowable range, while the monitoring frequency is increased to twice the normal sampling frequency to continuously monitor the deviation change trend. If, in degraded operation mode, the deviation gradually falls back below the acceptable level, then degraded operation mode is exited, and normal adaptive control is restored. If, in degraded operation mode, the deviation continues to worsen and exceeds the safety threshold, an emergency shutdown is immediately executed.

[0046] In practical implementation, adaptive adjustments to machining parameters are performed based on the adjustment amounts. Specifically, the adjustment amounts of each machining parameter determined in the preceding steps are output to the machine tool's execution unit via the control system. The corresponding execution mechanism then performs the adaptive adjustment of the machining parameters. Machining parameters include at least one of spindle speed, feed rate, depth of cut, and clamping force. Specifically, spindle speed adjustment is achieved by outputting a speed command signal to the machine tool spindle driver; this speed command signal is a digital signal. Feed rate adjustment is achieved by outputting a feed rate adjustment signal to the machine tool servo driver. This signal is superimposed on the feed command of the current machining program, changing the feed rate of the tool relative to the workpiece in real time. Depth of cut adjustment is achieved by outputting a position offset command to the machine tool's axial feed servo system, changing the tool's radial or axial depth of cut. Clamping force adjustment is achieved through multiple independently controllable clamping units located on the edge of the aluminum alloy housing. Each clamping unit is equipped with a pressure sensor and a displacement sensor to provide real-time feedback of the actual pressure and displacement values ​​at each clamping point. When the adjustment of machining parameters includes the adjustment of clamping force, the control system outputs pressure adjustment commands to the drive devices of each clamping unit. Each clamping unit independently adjusts its output pressure according to the received pressure adjustment commands. The pressure adjustment commands of each clamping unit are independent of each other, so that the clamping force at each clamping point can be adjusted according to its respective target value. After each adaptive adjustment is completed, the complete data of this adjustment is recorded as a data sample. The data sample includes: machining parameters before adjustment, machining parameters after adjustment, sensor signal data before and after adjustment, comprehensive deviation index before and after adjustment, and final machining quality inspection results. The data sample is stored in the machining database. When the accumulated number of data samples reaches a preset number, these data samples are used to incrementally learn and update the surrogate model. Incremental learning refers to using new data samples to locally correct the model parameters without retraining the entire model, so that the model can gradually adapt to new working condition changes that occur in actual machining. When the comprehensive deviation index after adjustment is significantly lower than before adjustment, and the final machining quality inspection results meet the requirements, the combination of anomaly type-adjustment strategy-adjustment amount corresponding to this adjustment is stored as a successful case in the machining rule base. When the same or similar anomaly type occurs later, the adjustment strategy and adjustment amount corresponding to the successful case will be matched directly from the rule base to speed up the adjustment decision-making process.

[0047] like Figure 2The diagram shows a schematic of the device structure of an adaptive control system for aluminum alloy shell processing provided in this application embodiment. The system includes a digital twin server, a multi-source sensor group, a signal acquisition and processor, an industrial control computer, and a machine tool execution unit. The digital twin server is used to construct a digital twin model of the aluminum alloy shell and simulate the processing process to generate simulated processing data and determine the limit range of processing parameters. The multi-source sensor group is used to collect multi-source sensor signals such as cutting force, vibration, acoustic emission, and clamping stress in real time during processing. The signal acquisition and processor is used to perform time-frequency domain analysis and feature extraction on the multi-source sensor signals to generate a processing feature vector. The industrial control computer is used to input the current actual processing parameters into the proxy model in the digital twin server to obtain predicted feature values, compare the processing feature vector with the predicted feature values ​​to determine whether an abnormal state has occurred, and determine the adjustment amount of the processing parameters when an abnormal state is determined. The machine tool execution unit is used to perform adaptive adjustment of the processing parameters according to the adjustment amount.

[0048] Therefore, this application demonstrates the following: First, by constructing a digital twin model of the aluminum alloy shell and simulating the machining process, the processing quality response law under different combinations of processing parameters can be pre-understood before actual machining. Based on this, the limit range of each processing parameter can be determined, providing a safety boundary and constraint range for subsequent online adaptive adjustment. Second, by acquiring multi-source sensor signals in real time during machining and fusing them into a processing feature vector, a multi-dimensional perception of the aluminum alloy shell's machining state can be achieved, providing a data foundation for subsequent anomaly judgment. Third, by comparing the deviation between the actual machining feature vector and the predicted feature value output by the digital twin model, quantitative identification of machining anomalies is achieved, and the deviation vector provides a basis for anomaly type discrimination. Finally, by determining the anomaly type based on the anomaly state judgment result and matching the corresponding adjustment strategy from the preset machining rule library, local optimization is performed within the limit range of machining parameters with the goal of minimizing the deviation to determine the adjustment amount of the machining parameters. Then, the adjustment amount is output to each execution unit of the machine tool through the control system to perform adaptive adjustment, realizing a complete closed loop from anomaly identification to parameter control. This enables the adjusted machining parameters to quickly and accurately suppress the abnormal state and restore the machining process to stability.

[0049] In summary, the technical solution adopted in this application can accurately identify abnormal states and automatically optimize processing parameters during the aluminum alloy shell processing, thereby improving processing quality and yield.

[0050] Example 2: This application provides an adaptive control system for aluminum alloy casing processing, referencing... Figure 3 As shown in the figure, this is a block structure diagram of the adaptive control system for aluminum alloy shell processing according to this embodiment of the present application. The adaptive control system includes: The processing simulation module 100 is used to construct a digital twin model of the aluminum alloy shell, simulate the processing process based on the digital twin model, generate simulated processing data, and determine the limit range of processing parameters based on the simulated processing data. The feature extraction module 200 is used to collect multi-source sensor signals in real time during the processing, and then extract processing feature vectors that represent the current processing state. The anomaly detection module 300 is used to determine whether an abnormal state has occurred in the current processing stage based on the deviation between the processing feature vector and the predicted feature value output by the digital twin model. The abnormal adjustment module 400 is used to respond to the judgment result of the abnormal state, determine the adjustment amount of the processing parameters with the limit range of the processing parameters as the constraint boundary, and perform adaptive adjustment of the processing parameters according to the adjustment amount.

[0051] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0052] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compactdisc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0053] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

Claims

1. An adaptive control method for machining aluminum alloy shells, characterized in that, The adaptive control method includes the following steps: A digital twin model of an aluminum alloy shell is constructed, and the processing is simulated based on the digital twin model to generate simulation processing data. The limit range of processing parameters is determined based on the simulation processing data. During the processing, multi-source sensor signals are collected in real time, and then processing feature vectors representing the current processing state are extracted. Based on the deviation between the processing feature vector and the predicted feature value output by the digital twin model, it is determined whether an abnormal state has occurred in the current processing stage; In response to the determination result of the abnormal state, the adjustment amount of the processing parameters is determined with the limit range of the processing parameters as the constraint boundary, and the adaptive adjustment of the processing parameters is performed according to the adjustment amount.

2. The adaptive control method for aluminum alloy shell processing as described in claim 1, characterized in that, Constructing a digital twin model of the aluminum alloy shell, simulating the manufacturing process based on the digital twin model, and generating simulation processing data specifically includes: Obtain CAD design data and material property data for the aluminum alloy shell, and establish an initial geometric model; The initial geometric model is meshed to construct a digital twin model; Configure the various processing parameters to be optimized in the digital twin model, and set the initial range of each processing parameter; The simulation experiment granularity is set based on the experimental design method, and multiple sets of simulation parameter combinations are generated by discrete sampling within the initial range of each processing parameter. The digital twin model is driven by each set of simulation parameters to simulate the processing process and generate corresponding simulation processing data.

3. The adaptive control method for aluminum alloy shell processing as described in claim 1, characterized in that, The multi-source sensing signals include cutting force signals, tool vibration signals, acoustic emission signals, and workpiece clamping stress signals.

4. The adaptive control method for aluminum alloy shell processing as described in claim 1, characterized in that, Determining whether an abnormal state has occurred in the current processing stage based on the deviation between the processing feature vector and the predicted feature value output by the digital twin model specifically includes: The actual processing parameters at the current moment are obtained, and the actual processing parameters are input into a pre-constructed processing proxy model, which drives the processing proxy model to output predicted feature values ​​corresponding to the current processing state; wherein, the processing proxy model is constructed using the response surface methodology based on the simulation data of the digital twin model; The processing feature vector is compared with the predicted feature value to determine the deviation vector; The deviation vectors are weighted and fused to obtain a comprehensive deviation index; Based on the comparison between the comprehensive deviation index and the preset warning threshold, it is determined whether an abnormal state has occurred in the current processing stage.

5. The adaptive control method for aluminum alloy shell processing as described in claim 4, characterized in that, Based on the comparison between the comprehensive deviation index and the preset warning threshold, determining whether an abnormal state has occurred in the current processing stage specifically includes: The comprehensive deviation index is stored in a sliding time window of fixed length to form a deviation sequence; Within the sliding time window, short-term trend characteristic quantities and long-term trend characteristic quantities are determined respectively. The short-term trend characteristic quantity reflects the instantaneous rate of change of the most recent few moments, and the long-term trend characteristic quantity reflects the overall drift direction and intensity within the entire window. An abnormal state is determined to have occurred only when the comprehensive deviation index exceeds the warning threshold and the short-term trend feature and the long-term trend feature simultaneously indicate that the deviation is diverging or continuously increasing.

6. The adaptive control method for aluminum alloy shell processing as described in claim 4, characterized in that, The predicted feature values ​​include the predicted cutting force, vibration, acoustic emission, and stress distribution values ​​corresponding to the dimensions of the machining feature vector; the deviation vector includes the component deviation values ​​in each dimension; and the predicted feature values ​​output by the digital twin model have confidence intervals.

7. The adaptive control method for aluminum alloy shell processing as described in claim 1, characterized in that, In response to the determination result of the abnormal state, and using the limit range of the processing parameters as the constraint boundary, the specific adjustment amount of the processing parameters is determined as follows: Based on the result of the abnormal status determination, the abnormal type of the current processing stage is determined; According to the type of anomaly, the corresponding adjustment strategy is matched from the preset processing rule base; Based on the adjustment strategy and the limit range of the processing parameters, the adjustment amount of the processing parameters is determined.

8. The adaptive control method for aluminum alloy shell processing as described in claim 7, characterized in that, Based on the adjustment strategy and the limit range of the processing parameters, determining the adjustment amount of the processing parameters specifically includes: The set of processing parameters that can be adjusted under the current anomaly type is determined based on the adjustment strategy; Based on the adjustment priority determined by the adjustment strategy, a corresponding priority weight is configured for each processing parameter that is allowed to be adjusted; An objective function is constructed with the goal of minimizing the deviation between the processing feature vector and the predicted feature value, and processing parameters with higher priority are adjusted first during the optimization process. Using the limit range of the processing parameters as constraints, local optimization is performed within the limit range of the processing parameters, and the adjustment amount of the processing parameters is output.

9. The adaptive control method for aluminum alloy shell processing as described in claim 1, characterized in that, The adjustment amount of the machining parameters includes at least one of the following: spindle speed adjustment, feed rate adjustment, depth of cut adjustment, and clamping force adjustment.

10. An adaptive control system for aluminum alloy casing processing, used to execute the adaptive control method for aluminum alloy casing processing as described in any one of claims 1 to 9, characterized in that, The adaptive control system includes: The processing simulation module is used to construct a digital twin model of the aluminum alloy shell, simulate the processing process based on the digital twin model, generate simulated processing data, and determine the limit range of processing parameters based on the simulated processing data. The feature extraction module is used to collect multi-source sensor signals in real time during the processing, and then extract processing feature vectors that represent the current processing state. The anomaly detection module is used to determine whether an abnormal state has occurred in the current processing stage based on the deviation between the processing feature vector and the predicted feature value output by the digital twin model. An abnormal adjustment module is used to respond to the judgment result of the abnormal state, determine the adjustment amount of the processing parameters with the limit range of the processing parameters as the constraint boundary, and perform adaptive adjustment of the processing parameters according to the adjustment amount.

Images