Slotting machine rigidity dynamic adjustment system and method based on material image recognition

Abstract

The application discloses a slotting machine rigidity dynamic adjustment system and method based on material image recognition, relates to the technical field of slotting machine rigidity dynamic adjustment, and comprises the following steps: acquiring a material surface and a subsurface image sequence of a machining area of a slotting machine; inputting the image sequence into a pre-trained deep visual analysis model to output a material characteristic parameter set in real time; converting the material characteristic parameter set into a target rigidity value based on a rigidity demand mapping model; obtaining an initial rigidity adjustment amount according to the target rigidity value and generating a rigidity adjustment instruction to be sent to a rigidity adjustment execution mechanism for dynamic adjustment within a preset control window period before a tool contacts a target machining area; and fusing feedback data and image analysis results to correct the initial rigidity adjustment amount during the cutting process. The application realizes dynamic, accurate and predictable active adjustment of the rigidity of the slotting machine, thereby effectively improving the machining adaptability, intelligence and stability of the slotting machine.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of slotter rigidity dynamic adjustment, and more particularly to a slotter rigidity dynamic adjustment system and method based on material image recognition. BACKGROUND

[0002] Slotting is a key link in the process of mechanical manufacturing and material forming, and its processing quality and efficiency directly affect the structural performance and manufacturing cost of products. In actual industrial applications, the reasonable matching of the system rigidity of the slotter when processing different materials or materials with different structural characteristics is of great significance to ensure processing stability, suppress tool chatter and prolong equipment life. Existing slotting equipment mostly adopts fixed rigidity structure or manual preset rigidity adjustment scheme, lacks the ability to perceive and respond to the material state of the processing object, and is difficult to realize dynamic optimization of rigidity for different material states, especially in the face of complex working conditions such as uneven material hardness, strong structural heterogeneity or containing micro-defects, which often leads to the following problems:

[0003] The traditional system relies on static setting of rigidity parameters and fails to make predictive adjustments based on the actual pre-cutting working condition information of the tool, resulting in a lag in rigidity response at the moment of tool contact, which easily causes initial cutting impact and transient instability. Moreover, existing methods mostly infer the material state based on a single sensor or mechanical feedback, which cannot realize the collaborative identification of material macro-texture characteristics and micro-defects, limiting the accuracy and pertinence of rigidity adjustment. In view of the above problems, the present application provides a solution. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a slotter rigidity dynamic adjustment system and method based on material image recognition, which realizes dynamic, accurate and predictive active adjustment of the rigidity of the slotter, thereby effectively improving the processing adaptability, intelligence and stability of the slotter.

[0005] To achieve the above object, the present application provides the following technical scheme:

[0006] The present application provides a slotter rigidity dynamic adjustment method based on material image recognition, which comprises: acquiring a material surface and subsurface image sequence of a processing area of a slotter; inputting the image sequence into a pre-trained deep visual analysis model to output a material characteristic parameter set in real time; converting the material characteristic parameter set into a target rigidity value based on a rigidity demand mapping model; obtaining an initial rigidity adjustment amount according to the target rigidity value and generating a rigidity adjustment instruction to be sent to a rigidity adjustment execution mechanism for dynamic adjustment within a preset control window period before the tool contacts the target processing area; and fusing feedback data and image analysis results to correct the initial rigidity adjustment amount during the cutting process.

[0007] In one of the embodiments, the image sequence is input into a pre-trained deep visual analysis model to output a material characteristic parameter set in real time, specifically: the image sequence is preprocessed to obtain standardized input images; the input images are input into a first branch module in the deep visual analysis model to extract a macro-texture feature map; the input images are input into a second branch module in the deep visual analysis model to extract a micro-defect feature map; the macro-texture feature map is input into an offset prediction module to generate an offset; the micro-defect feature map is spatially aligned and calibrated according to the offset to obtain a calibrated micro-feature map that is aligned with the macro-texture feature map in the spatial dimension; and the calibrated micro-feature map and the macro-texture feature map are spliced in the channel dimension based on a fusion layer module to obtain a fusion feature map.

[0008] In one of the embodiments, a material characteristic parameter set is output in real time, specifically: the fusion feature map is compressed in the channel and reorganized to obtain a material visual characteristic representation map; and the material visual characteristic representation map is input into a pre-set prediction module to perform corresponding task reasoning according to a target task type to output the material characteristic parameter set.

[0009] In one of the embodiments, the deep visual analysis model is a multi-scale feature fusion network, including a first branch module, a second branch module, and a fusion layer module: the first branch module includes a plurality of stacked hollow convolution layers with different hole rates; the second branch module includes a residual network structure composed of a plurality of residual units, and an attention mechanism module is embedded in each residual unit; and the fusion layer module splices the macro-texture feature and the micro-defect feature in the channel to output the material characteristic parameter set.

[0010] In one of the embodiments, the material characteristic parameter set is converted into a target stiffness value based on a stiffness demand mapping model, specifically: the material characteristic parameter set is processed in cascade to construct a joint feature vector; the joint feature vector is nonlinearly transformed based on a residual connection multi-layer perception network to extract a final residual expression vector; an output regression layer is constructed based on the final residual expression vector to map it into the target stiffness value, and the model is trained to obtain the stiffness demand mapping model; and the material characteristic parameter set is input into the stiffness demand mapping model for joint, residual perception, and regression calculation to output a predicted target stiffness value in real time.

[0011] In one of the embodiments, in a preset regulation and control window period before the tool contacts the target machining area, specifically: path planning data is obtained; the minimum distance between the tool and the target machining area is obtained based on the path planning data; and when the minimum distance is less than a set start threshold, a preset stiffness adjustment and control window period is started.

[0012] In one of the embodiments, an initial stiffness adjustment amount is obtained according to the target stiffness value, and a stiffness adjustment instruction is generated and sent to a stiffness adjustment execution mechanism for dynamic adjustment. Specifically, in the regulation window period, the stiffness value of the system and the target stiffness value and the difference therebetween are obtained to obtain an initial stiffness adjustment amount, and a multi-stage stiffness adjustment instruction sequence is generated. In the regulation window period, the execution mechanism gradually implements stiffness adjustment actions according to the stiffness adjustment instruction sequence, so that the system stiffness value dynamically transitions to the target stiffness value. The last time slice when the tool is about to contact the target machining area is obtained, and in the last time slice, the current stiffness value is compared with the target stiffness value. If the error is within the preset tolerance range, it is confirmed that the stiffness adjustment is completed, and the tool is allowed to contact the target machining area.

[0013] In one of the embodiments, in the cutting process, the fusion of the feedback data and the image analysis result is obtained to correct the initial stiffness adjustment amount. Specifically, key feature parameters representing the process state are obtained. The key feature parameters are normalized, and a multi-modal feature fusion method based on principal component analysis and mutual information weight allocation is used to construct a unified cutting state perception vector. The cutting state perception vector and the initial stiffness adjustment amount are fused to obtain a stiffness offset correction amount under the current cutting state. The stiffness offset correction amount is added to the initial stiffness adjustment amount to obtain a second stiffness adjustment amount after dynamic correction. A stability petal diagram is constructed based on system working condition parameters, the second stiffness adjustment amount is substituted into the stability petal diagram for feasibility verification to obtain a critical cutting depth under the current corrected stiffness, and the critical cutting depth is compared with the actual cutting depth. If the critical cutting depth is greater than the actual cutting depth, it is determined that the second stiffness adjustment amount obtained by correction is effective. Otherwise, it is determined that the current second stiffness adjustment amount is not satisfied. Under the condition that the correction is verified to be feasible, the second stiffness adjustment amount is taken as the final stiffness adjustment amount, a second stiffness adjustment instruction is generated and sent to the stiffness adjustment execution mechanism.

[0014] The one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0015] 1. By fusing multi-modal high-resolution imaging and deep visual analysis network, multi-scale visual features of materials are extracted from surface and subsurface image sequences, and macro texture and micro defect are cooperatively modeled based on a double-branch structure with cavity convolution and attention enhancement, and the feature space is aligned and calibrated by deformable convolution, which significantly improves the extraction accuracy and interpretability of material characteristic parameters; further, a multi-layer perception mechanism with residual perception is introduced to build a stiffness demand mapping model, which realizes high-precision and nonlinear mapping between material micro characteristics and machining stiffness, effectively supports adaptive stiffness scheduling and dynamic control of slotting machines and other machining equipment, and has high real-time performance, scalability and industrial practical value.

[0016] 2. By introducing the "preset control window period" mechanism, the stiffness dynamic pre-adjustment is started before the tool contacts the target machining area. Through path planning, early prediction and response are achieved to ensure that the stiffness is adjusted to the target value before key machining, significantly improving machining stability and precision. At the same time, the dynamic correction of stiffness deviation is realized by combining multi-source perception (force, vibration, image) and deep learning model (LSTM), and the stability lobe diagram is used to verify the working condition adaptability of stiffness adjustment, and a stiffness adjustment system with forward-looking and closed-loop adaptive characteristics is constructed, which improves the real-time response ability of the system to chatter and machining quality abnormalities and the intelligent level of the machining process. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A flowchart of a slotting machine stiffness dynamic adjustment method based on material image recognition is provided for the embodiments of the present application.

[0018] Figure 2 A slotting machine stiffness dynamic adjustment system structure diagram based on material image recognition is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0020] In an exemplary embodiment, Figure 1 A flowchart of a slotting machine stiffness dynamic adjustment method based on material image recognition is provided for the embodiments of the present application, including the following steps:

[0021] S1, obtaining a material surface and subsurface image sequence of a slotting machine machining area.

[0022] The material surface image sequence refers to a set of images of a material bare surface continuously captured by a high-resolution imaging device (such as an industrial camera) before or during material processing, including surface texture (rolling marks, processing marks, grain structure), surface defects (cracks, oxidation, peeling, rust), and surface structural heterogeneity (different grain regions, heat treatment marks). The material subsurface image sequence refers to a sequence of structural images in the range of microns to millimeters below the material surface indirectly obtained by imaging means (including but not limited to structured light, polarization imaging, short-wave infrared, super-resolution reconstruction, etc.), including local hardening zones, annealing residual layers, micro-crack initiation zones, or crack tip propagation directions. By obtaining the material surface and subsurface image sequence, high-dimensional and interpretable input feature sources are provided for stiffness control, processing parameter optimization, and intelligent prediction during the processing.

[0023] S2, inputting the image sequence into a pre-trained deep visual analysis model to output a material characteristic parameter set in real time.

[0024] Optionally, the deep visual analysis model can realize end-to-end mapping of image pixels to material mechanical characteristics by fusing macro-texture and micro-defect features.

[0025] The deep visual analysis model is a multi-scale feature fusion network, including a first branch module, a second branch module, and a fusion layer module.

[0026] The first branch module includes a plurality of stacked dilated convolution layers with different dilation rates;

[0027] Optionally, the stacked dilated convolution layers with different dilation rates can be understood as stacking a plurality of dilated convolution layers, and the dilation rate ri of each layer is different (such as r1=1, r2=2, r3=4), so as to realize texture feature extraction under different scales of perception, and enable the model to simultaneously perceive local details and large-scale structures, wherein r1, r2, and r3 represent the dilation rates of different dilated convolution layers.

[0028] The second branch module includes a residual network structure composed of a plurality of residual units, and an attention mechanism module is embedded in each residual unit to enhance the expression ability of key micro-defect regions in the image. The attention mechanism module is a joint mechanism of channel attention and spatial attention.

[0029] The fusion layer module concatenates the macro-texture features and the micro-defect features in the channel to output the material characteristic parameter set.

[0030] The image sequence is input into the pre-trained deep visual analysis model to output the material characteristic parameter set in real time, specifically:

[0031] The image sequence is subjected to a preprocessing operation to obtain a standardized input image;

[0032] The preprocessing operation includes size normalization, pixel channel standardization and image enhancement processing, which is used to improve the adaptability and robustness of the model to different material image sequences.

[0033] The input image is input into a first branch module in the deep visual analysis model to extract a macro-texture feature map.

[0034] The macro-texture feature map refers to a feature tensor generated by extracting the large-scale texture distribution, structural periodicity and context organization information of the input image through the first branch with multiple stacked hollow convolution layers (with different hollow rates), and is mainly used to represent the texture trend, arrangement pattern or organizational structure of the material at the macro scale.

[0035] The input image is input into a second branch module in the deep visual analysis model to extract a micro-defect feature map.

[0036] The micro-defect feature map refers to a feature tensor generated by extracting local small defect areas, low-contrast foreign matter, edge damage and other micro-objects in the image through a residual network structure embedded with an attention mechanism (CBAM or SEblock) in the second branch, which emphasizes the representation ability of fine-grained areas and improves the sensitivity and accuracy of defect detection.

[0037] The macro-texture feature map is input into an offset prediction module to generate an offset, and the offset prediction module is composed of multiple 3x3 convolutions and nonlinear activation units, so that the offset can adaptively learn the geometric deviation of macro-texture and micro-defect at the structural boundary.

[0038] The micro-defect feature map is spatially aligned and calibrated according to the offset to obtain a calibrated micro-feature map aligned with the macro-texture feature map in the spatial dimension, and the spatial alignment and calibration is specifically:

[0039] The offset is input into a deformable convolution module as a sampling position correction parameter to perform convolution sampling on the micro-defect feature map, so that it is aligned to the macro-texture feature map in the feature geometric structure.

[0040] Based on the fusion layer module, the calibrated micro-feature map and the macro-texture feature map are subjected to channel dimension splicing operation to obtain a fusion feature map, which is used to represent the comprehensive feature expression of the input image at the macro and micro scales.

[0041] It should be noted that the nonlinear activation unit refers to a nonlinear function module introduced after each layer of convolution (or full connection) operation, which has the core function of giving the neural network the ability to fit nonlinear mapping, so as to enhance the ability of the model to express complex patterns (such as boundary change, texture mutation, etc.). In the offset prediction module, the nonlinear activation unit is usually composed of the following structure: after linear convolution transformation of the input macro-texture feature map, it is input to the nonlinear activation function Where W is the convolution kernel parameter, b is the bias term, Y is the offset, and the nonlinear activation function refers to the LeakyReLU activation function.

[0042] The real-time output material characteristic parameter set is as follows:

[0043] The fusion feature map is compressed and reorganized by a convolution operator group containing at least one 1x1 or 3x3 convolution kernel, realizing the fusion of cross-scale information and the suppression of redundant features, and obtaining a material visual characteristic representation map.

[0044] Wherein, the material visual characteristic representation map refers to a high-dimensional feature map with discriminative ability formed by multi-scale feature extraction, fusion and reorganization of the input image or image sequence by the deep visual analysis model, used to represent the visual attributes of the material to be detected in terms of macro-texture structure, micro-defect distribution, edge morphology, etc.

[0045] The material visual characteristic representation map is input into a preset prediction module, and corresponding task reasoning is performed according to the target task type, outputting a material characteristic parameter set, the task reasoning including parameter estimation, classification identification or defect detection operation, the material characteristic parameter set including material local hardness distribution, defect category and material structure parameter, the prediction module including at least one fully connected layer or convolution prediction head, constructing a network structure for performing task-related reasoning on the input feature map, and outputting the corresponding material characteristic parameter set according to the target task type.

[0046] Wherein, the prediction module refers to a network sub-module that realizes numerical regression, classification identification or spatial positioning of specific physical / semantic attributes through a structured neural network calculation path based on the fusion feature map output by the deep visual analysis model. Its structure design is determined according to the task attribute, which can be a classifier, regressor, detection head or semantic segmentation head, etc.

[0047] S3, based on the stiffness requirement mapping model, converting the material characteristic parameter set into a target stiffness value.

[0048] ​The stiffness requirement mapping model is a functional model that uses a set of material property parameters as input variables and predicts the required stiffness value through a mapping function or learning model. In grooving, different material regions have varying hardness and defect distributions. If a fixed stiffness parameter is used uniformly, problems such as tool runout and uneven machining can easily occur. The stiffness requirement mapping model can output the optimal clamping stiffness or structural stiffness required at a given point based on real-time or pre-detected material properties, providing feedback control to the actuator for adjustment, thus achieving a dynamic, precise, and low-loss machining strategy.

[0049] Based on the stiffness requirement mapping model, the set of material characteristic parameters is converted into target stiffness values, specifically as follows:

[0050] The set of material characteristic parameters is cascaded to construct a joint feature vector;

[0051] A multilayer perceptron network based on residual connections is used to perform a nonlinear transformation on the joint feature vector to extract the final residual representation vector, specifically:

[0052] Configure an L-layer MLP, and define the output of each layer. Defined in residual form as:

[0053]

[0054] in, =0,1,...,L-1, For activation function, This is the weight matrix. Here, L represents the bias term of the MLP, and L is the total number of layers in the multilayer perceptron. For the first The input joint feature vector of the layer, For the first The output feature vector of the layer;

[0055] The initial input is a joint feature vector, and the final residual representation vector after feature compression is obtained. ;

[0056] An output regression layer is constructed based on the final residual representation vector, which is then mapped to the target stiffness value. The model is then trained to obtain the stiffness requirement mapping model.

[0057] Based on the stiffness requirement mapping model, the input material characteristic parameter set is used for joint, residual perception and regression calculation, and the predicted target stiffness value is output in real time.

[0058] The target stiffness value is used to guide the optimal stiffness value for equipment processing or structural response. This stiffness value reflects the stiffness response capability that the grooving machine should have under the current material condition, so as to ensure the stability, accuracy and safety of the processing.

[0059] For example, assuming that there are uneven hardness and local micro-cracks on the surface of a workpiece, if the device stiffness is too high or too low, it may cause tool jumping, machining surface tearing or local collapse. At this time, input these microscopic information into the model, the predicted target stiffness value is 180 N / mm, and the device can adjust the clamping mechanism or support stiffness according to this value to realize the adaptive response of the machining area and ensure the forming quality.

[0060] The stiffness requirement mapping model has the following specific calculation formula:

[0061]

[0062] Wherein, is the target stiffness value, is the regression layer activation function, is the regression layer weight, is the regression layer bias, is the final residual expression vector.

[0063] It should be noted that by introducing the deep nonlinear mapping network with residual mechanism, the problems of gradient disappearance and feature degradation in traditional deep MLP are effectively alleviated, ensuring stable transmission and sufficient extraction of feature expression in deep network. Finally, the regression layer with Softplus activation outputs the target stiffness value, realizing high-precision mapping between material microscopic characteristics and macroscopic stiffness requirement. By applying the residual connection mechanism to the material stiffness modeling task, combining nonlinear activation and structure regularization, the training stability and generalization performance of the model are effectively improved; the advantage is that the accuracy and robustness of stiffness prediction can be significantly improved, providing reliable stiffness prediction support for adaptive parameter scheduling of intelligent machining equipment such as slotting machine.

[0064] S4, in the preset control window period before the tool contacts the target machining area, an initial stiffness adjustment amount is obtained according to the target stiffness value, and a stiffness adjustment instruction is generated and sent to a stiffness adjustment execution mechanism for dynamic adjustment.

[0065] Wherein, by introducing a preset control window period before the tool contacts the target machining area, the tool position is predicted in real time based on the path planning data, and the stiffness adjustment process is started in advance before approaching the target area, realizing dynamic pre-adjustment of the stiffness of the machining system. This scheme ensures that the system stiffness has been adjusted to the preset target value before the tool enters the key machining area, thereby effectively suppressing the structural vibration in the cutting process, improving the machining stability and precision, and ensuring the continuity and quality of the machining process.

[0066] In the preset control window period before the tool contacts the target machining area, specifically:

[0067] Obtain path planning data;

[0068] obtaining a minimum distance between the tool and the target machining area based on the path planning data;

[0069] when the minimum distance is less than a set start threshold, starting a preset rigidity adjustment control window period.

[0070] The path planning data is a clear trajectory of the tool, accurately describing the trajectory path of the tool moving from the current state to the target machining area, including three-dimensional coordinate points, curve forms (such as spline curves or segmented line segments) of the cutting path, feed speed, acceleration, current position of the tool and other dynamic motion parameters. Based on these data, the system can calculate the minimum spatial distance between the current position of the tool and the boundary of the target machining area.

[0071] According to the target rigidity value, an initial rigidity adjustment amount is obtained, and a rigidity adjustment instruction is generated and sent to a rigidity adjustment execution mechanism for dynamic adjustment, specifically:

[0072] In the control window period, the rigidity value and the target rigidity value of the system and the difference between the two are obtained to obtain an initial rigidity adjustment amount, and a multi-stage rigidity adjustment instruction sequence is generated according to a preset adjustment strategy, wherein each stage of adjustment instruction includes a target rigidity setting value, an action time segment and an execution priority identifier;

[0073] Optionally, the multi-stage rigidity adjustment instruction sequence is generated according to the preset adjustment strategy, which can be understood as discretizing the entire rigidity adjustment process into multiple control stages according to the predefined adjustment strategy (derived from process rules, model prediction, optimization control or historical experience, etc.), each stage forming a rigidity setting instruction to form a control sequence with time sequence and execution logic, and being issued to the rigidity adjustment execution mechanism to gradually complete the rigidity transition;

[0074] The initial rigidity adjustment amount refers to the numerical difference between the current actual rigidity value of the system and the preset target rigidity value, which is used to quantify the rigidity change range that the system needs to complete in the control window period, and is the direct driving amount and control basis of the rigidity adjustment behavior;

[0075] The rigidity adjustment instruction sequence is sent to the rigidity adjustment execution mechanism, and the execution mechanism analyzes and prepares to execute the rigidity adjustment action according to the time sequence and priority of the instruction;

[0076] The stiffness adjusting actuator (such as a magneto-rheological damper, an intelligent clamping device) completes the dynamic adjustment of the equivalent stiffness of at least one of the spindle system, the guide rail system or the tool clamping system before the tool starts cutting. After receiving the stiffness adjustment instruction sequence, the corresponding physical structure adjustment action (such as pre-tightening force adjustment, structure support stiffness change, etc.) is prepared for execution according to the time sequence and priority order. The actuator can respond and analyze the priority instructions in real time to achieve multi-level dynamic adjustment of the stiffness.

[0077] During the regulation window period, the actuator gradually implements stiffness adjustment to dynamically transition the system stiffness value to the target stiffness value. This process has the following control features:

[0078] The response time control is completed within the constrained response time for each adjustment to avoid excitation overshoot. The smoothness optimization is achieved by using an interpolation function or an S-shaped curve planning to make the stiffness change continuous and smooth, avoiding system vibration excitation.

[0079] The last time slice when the tool is about to contact the target machining area is obtained. In the last time slice, the system collects the current stiffness value and compares it with the target stiffness value. If the error is within the preset tolerance range, it is confirmed that the stiffness adjustment is complete, and the tool is allowed to contact the target area.

[0080] If the error is within the preset tolerance range, it can be understood that the difference between the two is greater than the preset minimum tolerance value and less than the preset maximum tolerance value.

[0081] S5, in the cutting process, the fusion of feedback data and image analysis results is obtained to correct the initial stiffness adjustment amount.

[0082] By correcting the initial stiffness adjustment amount, high-precision dynamic closed-loop control of the stiffness adjustment in the cutting process is achieved, which can sense the change of the machining state in real time and actively correct the stiffness adjustment strategy, significantly improving the response ability of the system to chatter and surface quality abnormalities.

[0083] In the cutting process, the fusion of feedback data and image analysis results is obtained to correct the initial stiffness adjustment amount, specifically:

[0084] A multi-source sensor data acquisition model of a cutting machining system is established to obtain key characteristic parameters representing a process state through the model, the model comprising: a three-dimensional force sensor for real-time acquisition of cutting force, a three-axis vibration sensor for capturing vibration characteristics of a spindle and a workpiece, and an industrial image acquisition system for monitoring surface morphology of a cutting area; the three types of data are synchronously acquired and time-aligned to construct a multi-source feature input sequence in a unified time domain, the key characteristic parameters including cutting force mean, pulsation amplitude and frequency extracted from the force sensor data, dominant frequency mode, amplitude envelope and order response extracted from the vibration sensor, and a workpiece surface roughness texture feature (which can indirectly reflect the influence of machining parameters on surface quality and identify whether there are scratches, burrs and other problems) and a boundary contour integrity index (which is used to identify machining contour abnormalities caused by vibration, tool offset or unstable workpiece stiffness) extracted from the image data through a deep convolutional neural network;

[0085] The key characteristic parameters are normalized and a multi-modal feature fusion method based on principal component analysis and mutual information weight allocation is used to construct a unified cutting state perception vector;

[0086] The cutting state perception vector and an initial stiffness adjustment amount are fused and input into a preset stiffness offset prediction model to obtain a stiffness offset correction amount under the current cutting state, the stiffness offset prediction model being constructed based on an LSTM (Long Short-Term Memory neural network);

[0087] The stiffness offset correction amount is calculated according to the following formula:

[0088]

[0089] wherein, is the stiffness offset correction amount, is the cutting state perception vector, is the initial stiffness adjustment amount, is a long short-term memory network module, is a multi-layer perceptron output layer, which performs nonlinear mapping on the LSTM output to generate the final stiffness correction amount.

[0090] The stiffness offset correction amount is added to the initial stiffness adjustment amount to obtain a second stiffness adjustment amount after dynamic correction;

[0091] Optionally, adding the stiffness offset correction amount to the initial stiffness adjustment amount can be understood as the sum of the stiffness offset correction amount and the initial stiffness adjustment amount.

[0092] constructing a stability blade diagram based on system working condition parameters, wherein the system working condition parameters include tool rotation speed, axial cutting depth, and material damping ratio, the stability blade diagram is used to represent the system chatter stability boundary at different stiffnesses, and a mapping function SLD:(K,ω)→a is generated, wherein K is the second stiffness adjustment amount, ω is the spindle speed, and a is the critical cutting depth;

[0093] substituting the second stiffness adjustment amount into the stability blade diagram for feasibility verification, obtaining the critical cutting depth under the current corrected stiffness according to the mapping function, and comparing the critical cutting depth with the current actual cutting depth;

[0094] if the critical cutting depth is greater than the actual cutting depth, it is determined that the corrected stiffness adjustment amount is effective;

[0095] otherwise, it is determined that the current second stiffness adjustment amount does not meet the dynamic stability requirement;

[0096] under the condition of verifying the correction feasibility, the second stiffness adjustment amount is taken as the final stiffness adjustment amount, a second stiffness adjustment instruction is generated and sent to a stiffness adjustment execution mechanism, and the dynamic closed-loop regulation and control of the stiffness during the cutting process is completed.

[0097] Embodiment 2, Figure 2 The structure schematic diagram of the slotting machine stiffness dynamic adjustment system based on material image recognition provided by the embodiment of the application includes a pattern acquisition module, a pattern analysis module, a target stiffness value generation module, a stiffness adjustment module, and a stiffness adjustment correction module, and there is a connection between the modules:

[0098] The pattern acquisition module is used to acquire a material surface and subsurface image sequence of a slotting machine processing area;

[0099] The pattern analysis module is used to input the image sequence into a pre-trained deep visual analysis model and output a material characteristic parameter set in real time;

[0100] The target stiffness value generation module is used to convert the material characteristic parameter set into a target stiffness value based on a stiffness demand mapping model;

[0101] The stiffness adjustment module is used to obtain an initial stiffness adjustment amount according to the target stiffness value and generate a stiffness adjustment instruction to be sent to a stiffness adjustment execution mechanism for dynamic adjustment within a preset regulation and control window period before the tool contacts the target processing area;

[0102] The stiffness adjustment correction module is used to fuse feedback data and image analysis results to correct the initial stiffness adjustment amount during the cutting process.

[0103] The above embodiments can be realized in whole or in part by software, hardware, firmware, or any other combination. When realized by software, the above embodiments can be realized in whole or in part in the form of a computer program product.

[0104] Those skilled in the art can understand that the modules and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized by hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0105] In addition, each functional module in each embodiment of the present application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0106] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0107] Finally: the above is only the preferred embodiment of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for dynamic adjustment of the stiffness of a grooving machine based on material image recognition, characterized in that, Includes the following steps: Acquire image sequences of the material surface and subsurface in the grooving machine processing area; The image sequence is input into a pre-trained deep visual analysis model, which outputs a set of material characteristic parameters in real time. Based on the stiffness requirement mapping model, the set of material characteristic parameters is converted into target stiffness values; Within a preset control window period before the cutting tool contacts the target machining area, the initial stiffness adjustment amount is obtained based on the target stiffness value, and a stiffness adjustment command is generated and sent to the stiffness adjustment actuator for dynamic adjustment. During the cutting process, the feedback data and image analysis results are fused together to correct the initial stiffness adjustment.

2. The method for dynamic adjustment of grooving machine stiffness based on material image recognition according to claim 1, characterized in that, The image sequence is input into a pre-trained deep visual analysis model, which outputs a set of material characteristic parameters in real time, specifically: The image sequence is preprocessed to obtain a standardized input image; The input image is fed into the first branch module of the deep visual analysis model to extract macroscopic texture feature maps; The input image is fed into the second branch module of the deep visual analysis model to extract microscopic defect feature maps; The macroscopic texture feature map is input into the offset prediction module to generate the offset; Based on the offset, the micro-defect feature map is spatially aligned and calibrated to obtain a calibrated micro-feature map that is spatially aligned with the macro-texture feature map. Based on the fusion layer module, the calibrated microscopic feature map and the macroscopic texture feature map are spliced ​​together along the channel dimension to obtain the fused feature map.

3. The method for dynamic adjustment of grooving machine stiffness based on material image recognition according to claim 2, characterized in that, The real-time output set of material characteristic parameters is specifically as follows: The fused feature map is subjected to channel compression and feature reshaping to obtain a visual characteristic representation map of the material. The material visual characteristic representation map is input into a preset prediction module, and the corresponding task reasoning is performed according to the target task type to output a set of material characteristic parameters.

4. The method for dynamic adjustment of grooving machine stiffness based on material image recognition according to claim 3, characterized in that, The deep visual analysis model is a multi-scale feature fusion network, including a first branch module, a second branch module, and a fusion layer module: The first branch module includes multiple stacked dilated convolutional layers with different void ratios; The second branch module includes a residual network structure composed of multiple residual units, and an attention mechanism module is embedded in each residual unit; The fusion layer module performs channel splicing of macroscopic texture features and microscopic defect features to output the material characteristic parameter set.

5. The method for dynamic adjustment of grooving machine stiffness based on material image recognition according to claim 1, characterized in that, The stiffness requirement mapping model converts the set of material characteristic parameters into target stiffness values, specifically as follows: The material property parameter set is cascaded to construct a joint feature vector; A multilayer perceptron network with residual connections is used to perform a nonlinear transformation on the joint feature vector to extract the final residual representation vector. An output regression layer is constructed based on the final residual representation vector, which is then mapped to the target stiffness value. The model is then trained to obtain the stiffness requirement mapping model. Based on the stiffness requirement mapping model, the input material characteristic parameter set is used for joint, residual perception and regression calculation, and the predicted target stiffness value is output in real time.

6. The method for dynamic adjustment of grooving machine stiffness based on material image recognition according to claim 1, characterized in that, The preset control window period before the tool contacts the target machining area specifically includes: Obtain path planning data; The minimum distance between the tool and the target machining area is obtained based on the path planning data; When the minimum distance is less than the set activation threshold, the preset stiffness adjustment control window period is activated.

7. The method for dynamic adjustment of grooving machine stiffness based on material image recognition according to claim 6, characterized in that, The step of obtaining an initial stiffness adjustment amount based on the target stiffness value and generating a stiffness adjustment command to be sent to the stiffness adjustment actuator for dynamic adjustment specifically involves: During the control window period, the system stiffness value and target stiffness value, as well as the difference between them, are obtained to obtain the initial stiffness adjustment amount and generate a multi-level stiffness adjustment command sequence. During the control window period, the actuator gradually implements stiffness adjustment actions according to the stiffness adjustment command sequence, so that the system stiffness value dynamically transitions to the target stiffness value; The last time slice before the tool is about to contact the target machining area is obtained. During the last time slice, the current stiffness value is collected and compared with the target stiffness value. If the error is within the preset tolerance range, the stiffness adjustment is confirmed to be complete, and the tool is allowed to contact the target machining area.

8. The method for dynamic adjustment of grooving machine stiffness based on material image recognition according to claim 7, characterized in that, The process of fusing feedback data and image analysis results during cutting to correct the initial stiffness adjustment amount specifically involves: Obtain key characteristic parameters that characterize the process state; The key feature parameters are normalized, and a unified cutting state perception vector is constructed by using a multimodal feature fusion method based on principal component analysis and mutual information weight allocation. The cutting state perception vector is fused with the initial stiffness adjustment amount to obtain the stiffness offset correction amount under the current cutting state; The stiffness offset correction is added to the initial stiffness adjustment to obtain the second stiffness adjustment after dynamic correction. A stability lobe diagram is constructed based on system operating parameters. The second stiffness adjustment is substituted into the stability lobe diagram to verify its feasibility. The critical cutting depth under the current modified stiffness is obtained and compared with the current actual cutting depth. If the critical cutting depth is greater than the actual cutting depth, then the second stiffness adjustment amount obtained by the correction is deemed valid; Otherwise, the current second stiffness adjustment amount is determined to be insufficient; If the correction is verified to be feasible, the second stiffness adjustment amount is taken as the final stiffness adjustment amount, a second stiffness adjustment command is generated and sent to the stiffness adjustment actuator.

9. A system using the material image recognition-based dynamic stiffness adjustment method for a grooving machine as described in any one of claims 1-8, characterized in that, It includes a graphics acquisition module, a graphics parsing module, a target stiffness value generation module, a stiffness adjustment module, and a stiffness adjustment correction module. These modules are interconnected. The image acquisition module is used to acquire image sequences of the material surface and subsurface in the grooving machine processing area; The image parsing module is used to input the image sequence into a pre-trained deep visual parsing model and output a set of material characteristic parameters in real time. The target stiffness value generation module is used to convert the set of material characteristic parameters into target stiffness values ​​based on the stiffness requirement mapping model. The stiffness adjustment module is used to obtain the initial stiffness adjustment amount based on the target stiffness value within a preset control window period before the tool contacts the target machining area, and generate a stiffness adjustment command to send to the stiffness adjustment actuator for dynamic adjustment. The stiffness adjustment and correction module is used to fuse feedback data and image analysis results during the cutting process to correct the initial stiffness adjustment amount.

Images