An auxiliary control system and method for manufacturing and assembly line abnormal intervention

Abstract

This invention discloses an auxiliary control system and method for anomaly intervention in manufacturing and assembly lines, relating to the field of auxiliary equipment and physical execution control technology in intelligent manufacturing workshops. The system includes: a multi-dimensional physical state acquisition unit for generating a multi-dimensional physical state dataset; a dynamic mapping and anomaly detection unit for real-time anomaly detection; a multi-process linkage intervention unit for generating flexible intervention control commands; an auxiliary control unit for executing intervention actions; and a closed-loop optimization unit for constructing feedback comparisons of intervention action effects and determining residual attitude errors, forming an adaptive correction closed loop for the auxiliary device. This invention enhances the flexibility and stability of workshop auxiliary mechanical devices, featuring non-intrusive installation, modular expansion, and rapid multi-process adaptation. It is particularly suitable for auxiliary control and precise intervention of multi-variety, variable-batch workpieces in complex assembly scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent manufacturing workshop auxiliary equipment and its physical execution control technology, and particularly to an auxiliary control system and method for abnormal intervention in manufacturing and assembly production lines. Background Art

[0002] With the continuous improvement of product complexity, batch differences, and flexible production requirements in the manufacturing industry, the workshop assembly production line faces challenges in higher workpiece docking accuracy, process beat stability, and multi-task dynamic adaptability. Especially in the assembly scenario of small batches of multiple varieties, the dimensional tolerances, assembly fitting methods, and process connection beats of different workpieces are different, often resulting in small deviations in workpiece posture or insufficient stiffness of support alignment during the process execution, affecting the final assembly success rate. Traditional production lines usually rely on static clamping equipment, fixed-beat logistics buffers, and single support platforms in multi-process parallel operations, lacking a flexible buffer and real-time compensation mechanism for new workpieces or new tasks, and it is difficult to meet the rapid adaptation requirements for instantaneous fitting offsets in the assembly process. This limitation is particularly obvious in complex assembly, variable batch switching, or multi-station linkage operations, often directly affecting the overall beat continuity, physical action matching, and quality consistency of the final product of the production line.

[0003] Currently, most of the auxiliary device structures commonly used in workshops are mainly single-process rigid fixtures or single-axis supports, lacking multi-dimensional flexible adjustment capabilities for multi-process complex collaboration. Although sensors have been introduced in some scenarios for basic monitoring of clamping force or support angle, these monitors are often only used for safety limit alarms and cannot support the dynamic linkage of multi-process physical actions. Specifically, when there is a deviation between the rigid curve of the clamp clamping force and the buffer interval of the support arm, the system lacks an effective physical adjustment or fine-tuning channel, resulting in interference or insufficient support during fluctuations in workpiece geometry or changes in assembly posture. In addition, the logistics buffer rollers usually use fixed beats or simple beat controls and cannot achieve dynamic buffering according to the actual assembly synchronization beat, resulting in beat conflicts or buffer overflow phenomena easily occurring during multi-process concurrent execution. These problems are particularly prominent in the case of multi-task parallelism and rapid workpiece switching in the production line, becoming an important obstacle restricting the flexible upgrade of the production line and the safe matching of processes.

[0004] Especially in multi-product mixed-line assembly production, traditional rigid fixtures and logistics devices lack the ability to quickly respond to the diversity of workpieces, the variability of processes, and dynamic cycle time fluctuations. This makes it difficult to support the efficient assembly adaptation of intelligent flexible production lines to changes in workpiece size and shape. Faced with complex assembly steps and rapid switching between multiple tasks, existing workshop auxiliary devices typically cannot provide dynamic buffering, flexible cycle time absorption, and intelligent compensation for physical movements at the level of mechanized motion. This leads to problems such as assembly mismatch, process delays, and cycle time disorder in actual production line operation. Currently, there is a lack of a comprehensive control system that integrates multi-degree-of-freedom fixtures, flexible support arms, and adjustable logistics buffer rollers, along with real-time physical state perception, flexible motion adjustment, and multi-process motion adaptive compensation capabilities. This system is insufficient to meet the actual assembly needs of intelligent manufacturing workshops facing high-frequency, variable-batch production and process cycle time differences, and also restricts the safety assurance capabilities and flexible production adaptation level of workshop-level production lines. Summary of the Invention

[0005] To address the issues of fragmented physical coordination and insufficient flexible adaptation capabilities in multi-process production lines of assembly workshops caused by problems such as assembly cycle fluctuations, unstable clamping force, mismatched support platform posture, and logistics blockages, this invention proposes an auxiliary control system and method for abnormal intervention in manufacturing and assembly production lines, ensuring adaptive productivity and continuous optimization capabilities under dynamic switching of multiple tasks in intelligent manufacturing production lines.

[0006] On the one hand, to achieve the above objectives, the present invention provides an auxiliary control system for intervention in manufacturing and assembly line anomalies, comprising:

[0007] Multi-dimensional physical state acquisition unit: used to collect the operating status of equipment in the production line, the clamping force of fixtures, the workpiece arrival status of logistics path and the workpiece assembly angle error, and generate a multi-dimensional physical state dataset covering process-level actions.

[0008] Dynamic mapping and anomaly detection unit: used to construct a multidimensional device-based mapping model containing multi-source physical state mapping based on the multidimensional physical state dataset, and to detect residual alignment error of workpiece in real time, wherein the residual alignment error of workpiece includes clamping force overload, support arm posture angle drift or logistics cycle deviation.

[0009] Multi-process linkage intervention unit: used to automatically generate flexible intervention control commands when an abnormal state is detected, triggering dynamic correction of clamping force, multi-degree-of-freedom flexible fine adjustment of support arm, and adaptive adjustment of logistics buffer roller transmission rate;

[0010] Auxiliary control unit: used for instruction decomposition and modular action issuance, refining system control instructions into mechanical low-level execution parameters, and executing intervention actions;

[0011] Closed-loop optimization unit: Based on the real-time acquisition of the clamping force flexible loading curve, support arm posture compensation amplitude and logistics buffer state changes by the sensor array, it constructs feedback comparison of intervention action effect and residual posture error judgment to form an adaptive correction closed loop of the auxiliary device.

[0012] Preferably, the multidimensional physical state dataset includes:

[0013] Clamping force signal of fixture, multi-dimensional attitude angle vector of support arm, conveying speed of logistics roller and temperature and humidity of workstation.

[0014] Preferably, the multidimensional physical state acquisition unit generates the multidimensional physical state dataset covering process-level actions through an adjustable weighting coefficient matrix and a nonlinear mapping activation function, specifically:

[0015] ,

[0016] In the formula, It is a nonlinear mapping activation function. To cover the multidimensional physical state dataset of process-level actions, Both are weight matrices. For clamping force, For the multi-dimensional attitude angle vector of the support arm, For logistics roller conveyor speed, This refers to the temperature and humidity at the workstation.

[0017] Preferably, the multi-source physical state mapping includes a clamp module, a flexible support arm buffer structure, and a logistics buffer device.

[0018] The clamping module has a built-in servo electric cylinder and force sensor, which are used to adjust the clamping force of the clamp according to the real-time physical state of the workpiece.

[0019] The flexible support arm buffer structure includes a multi-degree-of-freedom rotary joint, a guide rail slider, and an electric actuator, which is used to adaptively adjust the spatial posture of the support arm in real time according to the workpiece assembly posture to compensate for the posture deviation of the workpiece during the insertion process.

[0020] The logistics buffer device consists of an electric roller drive module, a buffer support frame, a built-in encoder, and a buffer area sensor array. It is used to adjust the transmission speed of the buffer roller in real time according to the changes in the assembly task cycle and the dynamic accumulation of logistics.

[0021] Preferably, the clamping force of the clamp in the clamp module is adjusted in real time through a clamping control model, wherein the clamping control model is:

[0022] ;

[0023] In the formula, For clamping control model, This is the stiffness coefficient. The damping coefficient is... This represents the instantaneous rate of change of the workpiece docking error.

[0024] Preferably, the transmission speed of the buffer roller is calculated based on a dynamic adjustment model, specifically as follows:

[0025] ;

[0026] In the formula, To buffer the roller transfer speed, The reference speed of the drum. This is the cache flexibility adjustment coefficient.

[0027] Preferably, the dynamic mapping and anomaly detection unit integrates a pose sensor to detect residual alignment error of the workpiece in real time, wherein the residual alignment error of the workpiece is:

[0028] ;

[0029] In the formula, This refers to the residual alignment error of the workpiece. Assemble the pose of the desired target. This is the inspection pose after the actual assembly is completed. It is an L2 norm.

[0030] Preferably, when the multi-process linkage intervention unit detects that the residual alignment error exceeds the set safety tolerance, it triggers multi-dimensional flexible adaptive compensation through the auxiliary control unit, automatically reduces the upper limit of the clamping force of the fixture, extends the logistics buffer cycle, widens the production line buffer zone, and adjusts the support arm fine-tuning angle buffer zone in real time.

[0031] Preferably, the system further includes:

[0032] The modular quick-release structure allows for the rapid replacement of clamp modules, flexible support arm buffer structures, and logistics buffer devices via standardized interfaces;

[0033] Human-machine interface and visual operation interface: used to display the clamping force curve, the support arm posture buffer zone and logistics cycle fluctuation in real time, and supports manual fine-tuning and safety warning.

[0034] On the other hand, to achieve the above objectives, the present invention also provides an auxiliary control method for intervening in manufacturing and assembly line anomalies, comprising:

[0035] The multi-dimensional physical state acquisition unit collects the operating status of equipment in the production line, the clamping force of fixtures, the workpiece arrival status of logistics path, and the workpiece assembly angle error, generating a multi-dimensional physical state dataset covering process-level actions.

[0036] Based on the multidimensional physical state dataset, a multidimensional device mapping model containing multi-source physical state mapping is constructed, and real-time detection of fixture clamping force overload, support arm attitude angle drift, or logistics cycle deviation is performed.

[0037] Anomaly detection is performed through a multi-process linkage intervention unit. When an abnormal state is detected, a flexible intervention control command is automatically generated, triggering dynamic correction of clamping force, multi-degree-of-freedom flexible fine adjustment of support arm, and adaptive adjustment of logistics buffer roller transmission rate.

[0038] By decomposing instructions and issuing modular actions through the auxiliary control unit, the system control instructions are refined into mechanical low-level execution parameters to execute intervention actions;

[0039] Based on the real-time acquisition of the clamping force flexible loading curve, support arm posture compensation amplitude, and logistics buffer status changes by the sensor array, a feedback comparison of the intervention action effect and a determination of residual posture error are constructed to form an adaptive correction closed loop for the auxiliary device.

[0040] Compared with the prior art, the present invention has the following advantages and technical effects:

[0041] (1) This invention integrates multi-dimensional physical sensing data in real time, covering core dimensions of mechanical actions such as clamp loading force, support arm fine-tuning posture, and logistics roller cycle time, to form a unified physical state representation model for auxiliary devices. This representation breaks through the limitations of the traditional workshop's fragmented and single-point configuration of "clamp-support arm-logistic roller", and can achieve high-dimensional unification of multi-modal physical characteristics in complex assembly task switching, laying a data foundation for the flexible matching and dynamic adaptation of subsequent mechanical auxiliary actions, and significantly improving the physical adaptability and data-driven capability of workshop-level assembly multi-process workstations.

[0042] (2) This invention innovatively introduces a dynamic physical compensation control mode, which combines the flexible loading curve of clamping force, the buffer posture range of support arm and the flexible absorption band of logistics buffer beat to form a "real-time perception-buffering adjustment-safe correction" mechanism for multi-dimensional physical actions. This control mode can dynamically correct the stiffness range of clamp loading force, the buffer angle of support arm and the floating range of logistics roller beat when assembling new workpieces and performing multiple tasks in parallel, so as to avoid assembly mismatch caused by differences in workpiece size, rigidity or beat fluctuation, ensure the stability of the assembly process and the physical matching performance under high beat, and further improve the safety margin and dynamic balance capability of the workshop flexible device.

[0043] (3) Based on modular device configuration and multi-condition map, this invention establishes a similar working condition matching mechanism for physical states and a dynamic evolution mechanism for working condition nodes enhanced by graph neural networks, realizing rapid configuration of fixture-support arm-logistics buffer modules and efficient adaptation between multiple tasks. Through the linkage matching of physical states of each device in the multi-condition map and the fusion of multiple strategies driven by similarity, the system can quickly generate reasonable initial configuration of mechanical actions for the complex physical adaptation requirements of new tasks, and perform real-time fine-tuning and closed-loop correction during task execution. This mechanism not only improves the safety and process stability of the workshop assembly process, but also significantly enhances the flexibility of the device level in the face of diversified processes, cycle time fluctuations and workpiece changes, ensuring the adaptive productivity and continuous optimization capability of intelligent manufacturing production lines under dynamic switching of multiple tasks. Attached Figure Description

[0044] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0045] Figure 1 This is a schematic diagram of an auxiliary control system structure for abnormal intervention in manufacturing and assembly production lines, according to an embodiment of the present invention.

[0046] Figure 2 This is a flowchart of an auxiliary control method for intervening in manufacturing and assembly production line anomalies, according to an embodiment of the present invention. Detailed Implementation

[0047] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0048] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0049] This embodiment proposes an auxiliary control system for intervening in manufacturing and assembly line anomalies, such as... Figure 1 ,include:

[0050] Multi-dimensional physical state acquisition unit: used to collect the operating status of equipment in the production line, the clamping force of fixtures, the workpiece arrival status of logistics path and the workpiece assembly angle error, and generate a multi-dimensional physical state dataset covering process-level actions.

[0051] Dynamic mapping and anomaly detection unit: used to construct a multidimensional device-based mapping model containing multi-source physical state mapping based on the multidimensional physical state dataset, and to detect residual alignment error of workpiece in real time, wherein the residual alignment error of workpiece includes clamping force overload, support arm posture angle drift or logistics cycle deviation.

[0052] Multi-process linkage intervention unit: used to automatically generate flexible intervention control commands when an abnormal state is detected, triggering dynamic correction of clamping force, multi-degree-of-freedom flexible fine adjustment of support arm, and adaptive adjustment of logistics buffer roller transmission rate;

[0053] Auxiliary control unit: used for instruction decomposition and modular action issuance, refining system control instructions into mechanical low-level execution parameters, and executing intervention actions;

[0054] Closed-loop optimization unit: Based on the real-time acquisition of the clamping force flexible loading curve, support arm posture compensation amplitude and logistics buffer state changes by the sensor array, it constructs feedback comparison of intervention action effect and residual posture error judgment to form an adaptive correction closed loop of the auxiliary device.

[0055] Specifically, this embodiment enhances the physical adaptability of workshop-level assembly intervention and the safety robustness of process rhythm through the integrated response and action adaptive evolution mechanism of the multi-functional mechanized auxiliary device. It is particularly suitable for the physical flexible auxiliary control and real-time assembly quality assurance of multi-batch, mixed-line assembly processes, and meets the requirements of rapid adaptation, rhythm balance and dynamic safety of multi-task assembly in intelligent manufacturing workshops.

[0056] Furthermore, the multidimensional physical state dataset includes:

[0057] Clamping force signal of fixture, multi-dimensional attitude angle vector of support arm, conveying speed of logistics roller and temperature and humidity of workstation.

[0058] Specifically, by integrating multiple types of sensor arrays (such as force sensors, rotary encoders, photoelectric speed sensors, and temperature and humidity probes) to collect the physical working status of each auxiliary device in real time, the system fuses multimodal signals into a comprehensive physical state vector, i.e., a multidimensional physical state dataset, which is used to accurately characterize the instantaneous working conditions of the mechanized auxiliary actions at the workstation.

[0059] Furthermore, an adjustable weighting coefficient matrix and a nonlinear mapping activation function are used to generate the multidimensional physical state dataset covering the process-level actions.

[0060] Specifically, the fusion process employs an adjustable weighting coefficient matrix and a nonlinear mapping activation function, and the fusion expression is as follows:

[0061] ,

[0062] In the formula, It is a nonlinear mapping activation function. To cover the multidimensional physical state dataset of process-level actions, Both are weight matrices. For clamping force, For the multi-dimensional attitude angle vector of the support arm, For logistics roller conveyor speed, This refers to the temperature and humidity at the workstation.

[0063] Specifically, in this embodiment, Swish, Tanh, or ReLU can be selected to enhance the fusion representation capability under nonlinear mapping, ensuring the sensitivity and robustness of the multidimensional physical state dataset X covering process-level actions to changes in physical state under multi-process assembly cycle time, and supporting subsequent dynamic adaptive adjustment and flexible compensation action generation.

[0064] Furthermore, the multi-source physical state mapping is a multi-source physical state mapping that includes the clamp module, the flexible support arm buffer structure, and the logistics buffer device;

[0065] The clamping module has a built-in servo electric cylinder and force sensor, which are used to adjust the clamping force of the clamp according to the real-time physical state of the workpiece.

[0066] The flexible support arm buffer structure includes a multi-degree-of-freedom rotary joint, a guide rail slider, and an electric actuator, which is used to adaptively adjust the spatial posture of the support arm in real time according to the workpiece assembly posture to compensate for the posture deviation of the workpiece during the insertion process.

[0067] The logistics buffer device consists of an electric roller drive module, a buffer support frame, a built-in encoder, and a buffer area sensor array. It is used to adjust the buffer roller transmission speed in real time according to the changes in the assembly task cycle and the dynamic accumulation of logistics.

[0068] Specifically, the clamping force of the clamp in the clamping module is adjusted in real time through a clamping control model, which is as follows:

[0069] ,

[0070] In the formula, For clamping control model, This is the stiffness coefficient. The damping coefficient is... This represents the instantaneous rate of change of the workpiece docking error.

[0071] The aforementioned control mode enables the fixture to quickly adjust the clamping force according to the real-time physical state of the workpiece during multi-process assembly cycle switching and complex workpiece rotation, thereby improving the flexibility and adaptability to different workpiece stiffness, size and fit states, and ensuring physical safety and stability during the assembly process.

[0072] Specifically, the flexible support arm buffer structure can adaptively adjust the spatial posture of the support arm in real time according to the workpiece assembly posture, compensate for the posture deviation of the workpiece during the insertion process, and avoid assembly failure caused by posture mismatch.

[0073] Especially in complex assembly processes or multi-task parallel operation conditions, the flexible buffering action of the support arm can be matched with the cycle fluctuation of the logistics buffer roller, realizing the coupling and adaptation of mechanized actions, and enhancing the collaborative flexibility and process stability of the station-level mechanical auxiliary device.

[0074] Specifically, the buffer roller transfer speed is calculated based on a dynamic adjustment model, as follows:

[0075] ,

[0076] In the formula, To buffer the roller transfer speed, The reference speed of the drum. The buffer flexibility adjustment coefficient is adaptively adjusted in real time based on the stacking height of workpieces in the buffer area, the fluctuation of the process cycle time, and the safety margin of the target buffer capacity, so as to prevent cycle time mismatch caused by material flow interruption or excessive stacking.

[0077] Logistics buffer devices possess rapid response and flexible buffer management capabilities, making them key mechanical auxiliary devices for workshop-level cycle time balancing and process cycle time safety.

[0078] Furthermore, after each round of intervention actions, the system integrates pose sensors (such as laser rangefinders and industrial camera vision sensors) to detect residual alignment errors of the workpiece in real time. This is used to quantify assembly accuracy and the stability of physical motion matching. The residual alignment error of the workpiece is:

[0079] ,

[0080] In the formula, This refers to the residual alignment error of the workpiece. Assemble the pose of the desired target. This is the inspection pose after the actual assembly is completed. It is the L2 norm, used to characterize the overall error in the multidimensional attitude space.

[0081] Specifically, this indicator serves as the core basis for closed-loop verification and subsequent adaptive correction of mechanized assisted actions, supporting multi-dimensional quality assurance of assembly safety and cycle time robustness.

[0082] Furthermore, when the multi-process linkage intervention unit detects that the residual alignment error exceeds the set safety tolerance, it triggers multi-dimensional flexible adaptive compensation, automatically reduces the upper limit of the clamping force of the fixture, extends the logistics buffer cycle, widens the buffer zone between production lines, and adjusts the support arm fine-tuning angle buffer zone in real time.

[0083] Specifically, when residual alignment error of the workpiece is detected p exceeds the set safety tolerance At that time, multi-dimensional flexible adaptive compensation is triggered through the auxiliary control unit:

[0084] Automatically reduce the upper limit of clamping force to reduce the risk of rigid interference, extend the logistics buffer cycle to widen the production line buffer zone, and adjust the support arm fine-tuning angle buffer zone in real time to ensure that the workpiece is accurately aligned in the next assembly cycle, reducing cycle time impact and assembly failure rate.

[0085] The aforementioned comprehensive flexible adjustment mechanism forms a dynamic self-balancing and safety correction path for mechanical actions, which is particularly suitable for highly flexible assembly conditions such as rapid fluctuations in cycle time, diverse workpieces, and complex assembly, supporting the safety margin and execution consistency of the workshop assembly process.

[0086] Furthermore, the system also includes:

[0087] The modular quick-release structure allows for the rapid replacement of clamp modules, flexible support arm buffer structures, and logistics buffer devices via standardized interfaces;

[0088] Human-machine interface and visual operation interface: used to display the clamping force curve, the support arm posture buffer zone and logistics cycle fluctuation in real time, and supports manual fine-tuning and safety warning.

[0089] Specifically, the system is particularly suitable for rapid switching in multi-variety, variable-batch production scenarios. The auxiliary devices all adopt a modular quick-disassembly structure, including replaceable clamp modules, flexible support arm assemblies, and logistics buffer roller modules. Each module is connected through standardized mechanical interfaces and quick-connect pneumatic or hydraulic locking pairs, and has the function of quick disassembly and reassembly.

[0090] The modular quick-release structure not only supports the modular combination and mechanical flexible reconstruction of auxiliary devices when different batches or diverse workpieces are put on the line, but also effectively reduces the time for workstation adjustment and equipment debugging cycle, avoids the bottleneck of auxiliary devices caused by changeover, and ensures the continuity and consistency of assembly actions during process changeover.

[0091] Specifically, the fixture module can be quickly replaced or finely adjusted for different workpiece geometries, the support arm assembly can be replaced with a multi-degree-of-freedom attitude buffer structure to adapt to different mating states, and the logistics roller module can quickly increase or decrease the buffer capacity according to the fluctuation of the logistics cycle, realizing "plug and play - quick switching - high flexibility adaptation" mechanized auxiliary control, comprehensively improving the flexibility level and high efficiency adaptability of multi-process collaborative assembly production lines.

[0092] Specifically, the human-machine interface and visual operation interface intuitively display key physical states such as the real-time curve of the clamping force, the buffer zone of the support arm posture, and the fluctuation of the logistics roller buffer cycle. Combined with threshold safety warnings and motion limit alarms, it ensures that operators can monitor the operating status of the production line machinery in real time.

[0093] The interface allows operators to quickly make minor adjustments to auxiliary actions, correct the upper limit of clamping force, or switch to manual operation mode based on prompts, enhancing the workshop's flexible scheduling capabilities and adapting to sudden flexible compensation needs in multi-process assembly.

[0094] The system simultaneously records multi-dimensional sensor data and auxiliary device response logs during intervention execution, forming a complete execution history and safe operation record, which facilitates subsequent quality traceability and production line maintenance.

[0095] In addition, this modular visual interactive design supports rapid information sharing and collaborative adjustment between workstations, enabling dynamic safety assurance and flexible upgrading of production lines through multi-process linkage. It is particularly suitable for comprehensive assurance and continuous optimization of multi-process production line assembly safety, intelligent mechanical movements, and human-machine collaboration flexibility.

[0096] This embodiment also provides an auxiliary control method for intervening in manufacturing and assembly line anomalies, including:

[0097] The multi-dimensional physical state acquisition unit collects the operating status of equipment in the production line, the clamping force of fixtures, the workpiece arrival status of logistics path, and the workpiece assembly angle error, generating a multi-dimensional physical state dataset covering process-level actions.

[0098] Based on the multidimensional physical state dataset, a multidimensional device mapping model containing multi-source physical state mapping is constructed, and real-time detection of fixture clamping force overload, support arm attitude angle drift or logistics cycle deviation is performed.

[0099] Anomaly detection is performed through a multi-process linkage intervention unit. When an abnormal state is detected, a flexible intervention control command is automatically generated, triggering dynamic correction of clamping force, multi-degree-of-freedom flexible fine adjustment of support arm, and adaptive adjustment of logistics buffer roller transmission rate.

[0100] By decomposing instructions and issuing modular actions through the auxiliary control unit, the system control instructions are refined into mechanical low-level execution parameters to execute intervention actions;

[0101] Based on the real-time acquisition of the clamping force flexible loading curve, support arm posture compensation amplitude, and logistics buffer status changes by the sensor array, a feedback comparison of the intervention action effect and a determination of residual posture error are constructed to form an adaptive correction closed loop for the auxiliary device.

[0102] To more clearly illustrate the technical solution of the present invention, specific embodiments are provided below for description:

[0103] An auxiliary control method for intervening in manufacturing and assembly line anomalies, such as... Figure 2 ,include:

[0104] S1. In the multi-process production line of the assembly workshop, the system takes mechanical auxiliary devices as the core, fully collects and analyzes multi-source physical state data involving fixtures, support arms and logistics buffers, and constructs auxiliary motion perception input with multi-modal and hierarchical characteristics.

[0105] First, the clamping force of the fixture It is a key physical parameter for auxiliary actions. The system integrates a multi-point force sensor array at the gripping end of the fixture to monitor the evolution curve of the clamping force over time in real time, ensuring that the assembled workpiece has stable clamping matching performance under different stiffness and shape conditions, and avoiding assembly failure due to workpiece deformation or material brittleness.

[0106] Secondly, the multi-dimensional attitude angle vector of the support arm By using high-precision angle sensors, IMUs, and limit detection devices, the fine-tuning angle, buffer stroke, and force coupling characteristics are collected in real time to ensure that the slight posture difference during the assembly process can be dynamically compensated, thereby achieving flexible docking and precise alignment of the workpiece.

[0107] In addition, the speed of the logistics roller conveyor Relying on a multi-point rotary encoder, a speed-measuring laser module, and a buffer roller self-balancing monitoring unit, the linear speed and cycle response rate of the logistics roller are collected in real time, and the status of cycle fluctuation, buffer capacity saturation, and cycle flexible matching range are identified.

[0108] Based on environmental parameters, temperature and humidity sensors, vibration and acoustic monitoring modules are deployed around the workstation to analyze the potential impact of workshop environmental disturbances on the action response of auxiliary devices, thereby assisting in achieving refined dynamic compensation.

[0109] The system normalizes the aforementioned multidimensional mechanical motion and operating condition data into equipment state vectors. (Fixture mechanical data), process execution vector (Support arm and auxiliary structure execution status), logistics status vector (Buffer cycle time and logistics attitude), environmental parameter vector (Fluctuations in work environment).

[0110] To adapt to the changing working conditions in highly flexible assembly scenarios, the system introduces the following nonlinear fusion mapping to form an input state vector that comprehensively reflects the multidimensional dynamic characteristics of auxiliary actions. :

[0111] ,

[0112] in, Both are weight matrices. It is a nonlinear mapping activation function, which improves the discrimination and robustness of the fused state under complex working conditions.

[0113] After state fusion, the system can optionally introduce sparse principal component analysis or multimodal variational autoencoder to reduce the dimensionality and compress the fused vector, reduce redundant channels, and highlight the characteristics of mechanized actions, ensuring that the input state vector is accurate even in complex multi-process scenarios. It possesses high-resolution and highly generalized mechanical-assisted action expression capabilities.

[0114] Through this mechanized sensing and fusion mapping step, the system not only constructs a multi-dimensional dynamic state input that is closely coupled with the production line conditions, but also provides a solid physical sensing foundation for subsequent multi-objective action strategy learning and assembly action adaptive control.

[0115] S2, Based on the input state vector obtained in S1 The system enters the learning phase of multi-objective physical-assisted motion strategies. The core is to train mechanical motion strategies that can adapt to changing working conditions and improve the collaborative efficiency and adaptability of mechanical-assisted motions on the production line.

[0116] At this stage, the system constructs a state-action-result triplet decision model centered on mechanically assisted actions. The core objective is to simultaneously optimize the rigidity / flexibility matching of the clamping fixture, the stability of the support arm's posture compensation, and the smooth absorption capability of the logistics buffer cycle time. The "state-action-result" triplet decision model uses the input state vector... Given the state input, combined with the current assembly task objective, output the optimal combination of physical action strategies (clamping force adjustment, posture compensation, and rhythm adjustment).

[0117] Specifically, the following instant reward function was designed. Taking into account the performance of multi-target physical actions:

[0118] ,

[0119] In the formula, This indicates the rhythmic coordination efficiency between the logistics buffer and the support arm's movements; It measures the energy consumption of the support arm and clamp movements per unit time, and evaluates execution efficiency and energy saving level. This represents the yield rate of workpiece assembly after mechanical assistance actions. The instantaneous attitude deviation modulus of the support arm during physical compensation reflects the stability of flexible compensation. The probability of failure in logistics buffering or clamping operations (such as jamming, crushing of workpieces, etc.); These are all weighting coefficients, which can be flexibly adjusted according to the needs of workshop flexibility, energy saving, and high-reliability assembly.

[0120] This reward function achieves a comprehensive balance of multi-dimensional auxiliary motion performance by introducing mechanical physical motion constraint indicators, ensuring that the movements of the clamp and support arm meet the requirements of flexible adaptation while also being executed efficiently.

[0121] Subsequently, a policy gradient-based reinforcement learning framework, such as proximal policy optimization or deterministic policy gradient, is employed to iteratively update the policy network within the mechanical action decision space. .

[0122] The updated gradient formula is as follows:

[0123] ,

[0124] In the formula, Indicates the state The probability distribution of mechanical actions under the following conditions. To consider long-term returns, The gradient direction of the policy parameters. For strategy parameters The optimization objective gradient direction, In the current strategy The expected value operator for state-action pairs, In a given state Next, action Gradient information relative to the current policy parameters.

[0125] During training, a dynamic weight adjustment mechanism for mechanical motion is introduced. Based on physical layer feedback such as clamping force / support arm buffer response rate and clamping overload trigger rate, the weight coefficients of the reward function are dynamically updated. This enables flexible and adaptive compensation for the policy network.

[0126] To adapt to high-frequency, multi-task switching conditions, the training strategy also considers the multi-channel physical motion coupling of fixtures, supports, and logistics buffers. Specifically, the mutual influence characteristics of these three elements are introduced into the reward function and policy updates to ensure the consistency of multiple auxiliary actions and the robustness of mechanical coupling.

[0127] Ultimately, the S2 stage not only forms a multi-objective physical motion optimization strategy network for mechanical auxiliary devices, but also achieves adaptive intelligent control of complex assembly scenarios in the workshop through dynamic adjustment and multi-objective balancing, providing a reliable physical motion strategy foundation for subsequent strategy matching, fusion, and instruction issuance.

[0128] S3. When the production line is ready to launch a new task or switch between multiple tasks, the system enters the stage of matching similar working conditions and integrating multiple strategies for mechanical auxiliary motion strategies to ensure the flexible adaptation and rapid execution of mechanical motion configuration under the new task.

[0129] This stage first maps the physical auxiliary motion status of the current working condition, including the clamping force distribution of the fixture, the flexible posture of the support arm, the dynamic cycle time of the logistics buffer, and the adaptability of the workstation environment, to the historical working condition database for retrieval and matching.

[0130] During the matching process, the distance function By combining numerical similarity and physical motion pattern similarity, not only considering vector space distance, but also introducing physical layer matching tags such as "clamp flexible absorption curve similarity" and "matching rate between dynamic buffer zones of support arms", the search results can truly reflect the safety and adaptability at the mechanical motion level.

[0131] The optimal working condition matched, and the corresponding mechanical auxiliary action configuration, includes the clamp loading force range, support arm compensation bandwidth, and dynamic safety margin of logistics buffer, which is the initial mechanical action configuration for the new task.

[0132] However, when a new task introduces a completely new process segment or assembly path (such as first assembly, new configuration modules, etc.), the historical working conditions may not be fully matched. Therefore, the system introduces a multi-strategy fusion mechanism.

[0133] Select the n most similar operating condition configurations from the historical operating condition database. , , And based on similarity dynamic weights Perform physical motion fusion:

[0134] ,

[0135] In the formula, The parameters after fusion. This is the set of physical motion parameters executed under the i-th historical working condition, including the optimal configuration values ​​of key control quantities such as clamping force, support arm attitude angle, and logistics roller speed in that working condition. is the dynamic weight for similarity, and n is the number of historical similar working conditions used for fusion.

[0136] During the integration process, the clamping force of the fixture not only considers the loading stiffness under different working conditions, but also incorporates the flexible section of the fixture opening and closing curve to avoid mechanical fatigue and workpiece brittleness; the support arm movement integrates the fine adjustment range and dynamic response speed of each working condition to ensure that it still has flexible adaptability under new assembly alignment deviations; the logistics buffer cycle zone integrates the buffer step length and roller speed distribution under multiple working conditions to ensure the impact resistance and absorption stability of the production line logistics under the new cycle.

[0137] After integration, the system features a specially designed self-checking mechanism for mechanical motion safety and a buffer strategy correction mechanism. Specifically, this includes:

[0138] The system detects the maximum / minimum force limits for the clamping force and automatically adds a flexible buffer section based on the workpiece's brittleness tolerance.

[0139] For the fine-tuning range of the support arm, calculate the dynamic compensation limit. If there is a risk of overshoot or undercompensation, automatically increase the torque stability range of the buffer arm.

[0140] For logistics buffer configuration, analyze the coupling relationship between cycle time fluctuation range and process cycle time. If there is a risk of buffer overflow, the system adds a "buffer flexibility adjustment coefficient". "Dynamic smoothing of roller acceleration and deceleration curves."

[0141] This process not only ensures that the initial configuration of the mechanically assisted actions for new tasks is safe, reliable, and reasonably matched mechanically, but also significantly reduces the time and probability of errors in manual configuration.

[0142] After generating the fusion results, the operating terminal provides an intuitive multi-dimensional motion curve visualization interface. Operators can view the fitting degree and safety margin of the clamping force curve, support arm buffer curve, and logistics cycle absorption curve in real time, and can make manual fine adjustments when necessary, forming an efficient mechanical motion configuration adaptation process of human-machine integration and system self-adaptation.

[0143] This step achieves a closed loop throughout the entire process, from "historical experience - multi-strategy integration - physical layer safety correction - manual visual fine-tuning" to "initial configuration of mechanical auxiliary actions," ensuring that the flexible execution of fixtures, support arms, and logistics auxiliary actions has high robustness and high intelligence under new working conditions and new assembly modes.

[0144] S4. After completing the initial configuration of auxiliary actions and the fusion of multiple strategies under the new task, the process enters the stage of physical action instruction generation, conflict detection and execution, ensuring the safety, flexibility and adaptability of mechanical auxiliary actions and their dynamic collaboration capabilities in a multi-process environment in the workshop.

[0145] First, for the action configurations such as clamping, support arm buffering, and logistics buffering, the fused parameters are... This is converted into a standardized multidimensional instruction structure, with the following format:

[0146] ,

[0147] in, Indicates the workstation or equipment number in which the action is performed; It is further refined into specific physical actions such as "dynamic range adjustment of clamp loading force", "flexible buffer adjustment of support arm posture", and "absorption of logistics roller beat". Core parameters of physical motion, such as "upper limit of clamping force". "Support arm fine-tuning angle" "Cache tick size" "wait; This is the acceptable deviation range for the action, ensuring a safety margin for the action; Set the target value for the physical action to be performed; A structure for executing instructions for standardized physical actions.

[0148] After the instruction is generated, the system first performs multi-process mechanized motion coordination and conflict detection. In this detection, the system dynamically analyzes the physical coupling relationship between the fixture, support arm, and material buffer in the instruction based on a working condition correlation model, including:

[0149] Clamping force and support arm buffer conflict detection: Real-time analysis of the maximum support load range of the support arm. If the clamping force in the command exceeds the maximum support load range, the system will trigger "support buffer overload warning" and automatically adjust the clamping loading curve to prevent overload jamming or workpiece damage.

[0150] Fixture / support and logistics buffer cycle sequence matching: if the logistics buffer roller acceleration If the assembly synchronization cycle is exceeded, a "flexible buffer factor" is automatically added to dynamically slow down the roller cycle and prevent the material from overflowing before the clamp / support arm is in place.

[0151] Concurrent execution safety detection: For scenarios where multiple processes are executed simultaneously, based on the dependency links in the process diagram, ensure that the physical execution sequence and cycle interval of actions such as fixtures, support arms, and logistics buffers are reasonable, so as not to cause action conflicts or hard collisions.

[0152] To further enhance the adaptability of mechanized action issuance, a dynamic safety buffer and fine-tuning compensation mechanism has been introduced. Specifically, the system automatically generates buffer segments for the instruction set based on the workshop's flexibility requirements and the equipment's rigidity limits, dynamically adjusting the following:

[0153] Clamping force buffer section When the fixture performs loading, an elastic loading curve is introduced to ensure that the clamping force can climb flexibly during the unstable period of workpiece assembly, avoiding excessive rigidity of one-time loading.

[0154] Support arm fine-tuning attitude buffer zone If the support arm needs to absorb differences in workpiece position during assembly, the system allows for dynamic expansion of the support arm's fine-tuning range, improving the flexibility of assembly docking.

[0155] Logistics buffer cycle flexible area In the peak assembly cycle zone, the system allows for dynamic fluctuations in the roller linear speed, forming a flexible absorption zone to ensure smooth coupling of logistics buffers and upstream and downstream processes.

[0156] During the instruction issuance process, the system relies on workshop edge computing nodes to convert the instruction set into standard industrial protocol data packets (OPC UA, Modbus TCP, CANopen, etc.), and pushes them to the production line's underlying controllers (such as PLCs, embedded motion controllers, intelligent support arm modules, etc.) in real time. The instruction sending process supports the dynamic allocation of high-priority and low-priority actions, forming a dynamic priority execution logic for multi-process mechanized actions, preventing process delays or physical interference caused by incorrect action coupling sequence.

[0157] To ensure the accurate execution of mechanical motion commands, real-time physical motion curve monitoring is initiated simultaneously with the issuance of commands, monitoring the following physical data:

[0158] Real-time variation curve of clamping force Analyze the slope of the clamping force increase / decrease to detect the presence of mechanical hysteresis or oscillation;

[0159] Support arm attitude angle response curve : Detect the dynamic response of the support arm's fine-tuning compensation to prevent coordination deviations caused by motion stagnation;

[0160] Logistics buffer roller cycle response To determine whether the roller linear speed meets the requirements of flexible cycle absorption, and to avoid assembly cycle instability.

[0161] If a deviation in the physical execution curve is detected (such as instantaneous fluctuations in clamping force) If the tolerance bandwidth is exceeded, the system immediately triggers a "fine-tuning secondary compensation" action to make small, flexible corrections within the physical motion space, such as fine-tuning the flexible loading curve of the clamping force and widening the buffer attitude bandwidth of the support arm, to ensure the safety, continuity and accuracy of the mechanical auxiliary actions during execution.

[0162] The operating terminal is equipped with a motion curve and safety margin visualization panel, which displays the changes of key physical indicators such as clamping force-time curve, support arm compensation angle-time curve, and logistics buffer roller linear speed-time curve in real time. This allows operators to dynamically view, manually fine-tune, and conduct process safety assessments during production line operation, forming a dual safety guarantee of "human-machine integration" and "physical motion self-adaptation".

[0163] This phase enables a closed-loop mechanized action system that spans the entire chain, from "physical motion configuration - flexible conflict detection - safety instruction issuance - physical execution monitoring - fine-tuning and correction". This strengthens the physical adaptability, execution safety, and production line collaboration efficiency of mechanical auxiliary actions in multi-process and complex assembly scenarios.

[0164] S5. As production lines continuously switch tasks and introduce new assembly processes, the system faces the challenge of quickly adapting mechanical auxiliary motion configurations to new working conditions, ensuring the flexibility of mechanical motions and the continuity of processes.

[0165] To this end, a new task launch strategy initialization and self-evolutionary dynamic adjustment mechanism was introduced in the S5 phase to ensure that mechanical auxiliary actions have the ability to adapt instantly and have the potential for continuous optimization when new tasks are launched.

[0166] First, when a new task is launched, based on the fused state vector... In the historical working condition map Node matching is performed in the process. Indicates historical operating conditions, edge This represents the transferable path of mechanically assisted actions. Dynamic updates of node features are achieved through a graph neural network.

[0167] ,

[0168] In the formula, For embedding nodes in layer l, For the set of neighboring nodes, This is the weight matrix. For bias, It is a nonlinear mapping activation function. For the embedding of nodes in layer l+1, For historical working condition diagrams and nodes The adjacent first The embedding representation vector of the layer neighbor node, where u is the adjacent historical working condition node that has a migration edge with the current target node.

[0169] This process not only considers the historical mechanical experience of the clamp loading stiffness curve and the dynamic compensation characteristics of the support arm, but also takes into account the dynamic adaptability of the logistics buffer cycle flexible section under multi-process cycle coupling, and generates a recommendation for the optimal initial mechanical motion configuration for the new task.

[0170] Matched working condition nodes Mechanical motion configuration in A multi-layered safety self-check is performed based on the physical characteristics of the new task, including:

[0171] Clamping force safety assessment: Considering the material brittleness of the new workpiece and the upper and lower limits of the clamping force, prevent overloading or insufficient clamping in the initial configuration;

[0172] Dynamic compensation limit test of support arm: Considering the geometric deviation of the new assembly, ensure the maximum fine adjustment angle of the support arm. Sufficient to cover actual assembly and fit errors;

[0173] Logistics buffer cycle time dynamic absorption evaluation: detecting buffer cycle time step size Whether the range meets the safety margin of the process buffer capacity, and avoid logistics interruption or overflow under sudden fluctuations in cycle time.

[0174] For new tasks with entirely new combinations of operating parameters (such as parallel first-time assembly of new modules, complex assembly sequences, etc.), historical operating conditions may not be able to directly provide a highly compatible physical configuration.

[0175] To address this, a multi-strategy fusion compensation mechanism is introduced, starting with the strategy with the highest similarity. Extract mechanical motion configuration from each working condition And merge them according to the dynamic weight of similarity:

[0176] ,

[0177] In the formula, The parameters after fusion. This is the set of physical motion parameters executed under the i-th historical working condition, including the optimal configuration values ​​of key control quantities such as clamping force, support arm attitude angle, and logistics roller speed in that working condition. is the dynamic weight for similarity, and n is the number of historical similar working conditions used for fusion.

[0178] During the integration process, physical parameters such as the flexible absorption range of the fixture, the buffer width of the support arm, and the flexible cycle curve of the logistics buffer work together to form a comprehensive configuration that is more in line with the physical adaptability of the new task.

[0179] Meanwhile, adaptive buffering for mechanical motion safety is introduced, such as automatically inserting a flexible ramp curve in the clamping force segment to avoid excessive instantaneous impact when clamping new workpieces; and introducing a dynamic buffer correction band in the support arm movement to ensure the tolerance and impact resistance of the support movement flexibility range in new tasks.

[0180] After the new task is launched, real-time monitoring of multi-dimensional physical actions, including the real-time change curve of the clamping force, is used. Support arm attitude angle response curve Logistics buffer roller cycle response This generates real-time execution feedback.

[0181] If detected:

[0182] The clamping force fluctuates beyond the threshold (e.g., instantaneous fluctuations in clamping force). Safety threshold for clamping force variation );

[0183] The support arm exhibits delayed or unstable attitude response.

[0184] Logistics buffer overshoot caused overflow;

[0185] The system immediately triggers secondary physical buffer compensation, dynamically adjusting the stiffness of the clamp, the buffer section of the support arm, and the flexible step size of the logistics buffer, forming a closed-loop correction mechanism of "secondary buffer - real-time compensation - dynamic rebalancing".

[0186] Furthermore, real-time physical feedback data is used as self-evolving update samples for the reinforcement learning module, continuously enriching the policy network's multi-dimensional physical condition perception and multi-task assembly adaptation capabilities. Specifically, after each round of assisted assembly actions, the system summarizes multi-dimensional physical feedback information, such as the fixture loading force curve, support arm angle buffer amplitude, logistics buffer roller cycle adjustment results, and workpiece residual alignment error, into dynamic sample pairs (state-action-result), forming an incremental self-learning data pool for reinforcement learning. By comparing the difference between actual feedback and expected action effects, the system automatically identifies the flexible matching characteristics of the current assembly condition, extracts assembly action correction rules under new physical scenarios, and further enriches the dynamic adaptation space of policy parameters.

[0187] During the update process, a "freeze-fine-tune" structured update strategy is adopted: the underlying physical motion feature extraction module is frozen to maintain a basic understanding of physical signals and long-term experience accumulation, while only the high-level strategy generation part is finely adjusted. This ensures that the flexible matching and physical motion adaptive knowledge under historical working conditions are not lost, while enhancing the robustness of motion generation and assembly safety stability under new working conditions, new cycle times, and multi-batch switching scenarios. Through this mechanism, a self-evolving closed loop of mechanical motion-physical feedback-incremental strategy update-re-execution verification is formed, ensuring that the mechanized auxiliary device always has multi-dimensional flexible matching capabilities, stability, and safety boundary maintenance capabilities in complex and ever-changing assembly environments, meeting the efficient adaptive control requirements of workshop-level intelligent manufacturing production lines.

[0188] On the operator side, the system provides a visual interface for the self-evolution process of mechanical motion, dynamically displaying key physical indicators such as the evolution of clamping force segments, changes in the flexible bandwidth of the support arm posture, and flexible adjustment of logistics buffer cycle time. Operators can make manual fine adjustments, and combined with the system's self-evolution updates, ensure the safety, adaptability, and process stability of the mechanized motion configuration when new tasks are launched on the production line.

[0189] This step achieves a complete path in the mechanized motion execution dimension, from "historical experience matching - multi-strategy fusion - safety self-check buffer - real-time feedback compensation - self-evolution update - manual fine-tuning" to "adaptive flexible closed loop". Especially in variable working conditions and highly flexible parallel environments with multiple processes, the mechanical motion configuration has stronger safety margins, flexible adaptability, and efficient execution robustness, supporting the production line to achieve intelligent adaptation capabilities for launching new tasks with "safety, flexibility, and efficiency".

[0190] S6. After the production line task is completed, the system will systematically collect and evaluate the action history and assembly results of the auxiliary devices in a multi-dimensional feedback manner to form a continuous optimization mechanism for flexible auxiliary actions in multiple processes.

[0191] First, the system collects full-process time-series data, including clamping force, multi-dimensional posture angle vector of the support arm, conveyor speed of the material rollers, and temperature and humidity at the workstation. Combining the workstation alignment results with the workpiece clearance distribution, the system constructs a production line feedback quality assessment dataset. :

[0192] ,

[0193] In the formula, This represents the residual attitude error after assembly. The failure rate of assembly operations (such as insertion failure, support instability, etc.). This is time-series data on the temperature and humidity of the workstation environment.

[0194] Based on this data, the system will perform the following multi-dimensional analysis in stages:

[0195] Dynamic matching accuracy analysis: based on residual attitude error after assembly By observing the time distribution trend, we can determine the evolution of docking errors during the assembly process and whether problems such as assembly eccentricity, fixture slippage, or support arm vibration occur.

[0196] Force-pose coupling characteristic analysis: Analyze the cross-correlation between clamping force and support arm pose fine-tuning curve to determine the collaborative adaptability of physical actions under complex assembly conditions.

[0197] Logistics cycle time buffering performance analysis: Analyze the cycle time matching degree of logistics roller conveyor speed during peak and slow periods to confirm the cycle time adjustment response of the flexible buffer device.

[0198] Environmental adaptability analysis: Analyze the potential correlation between fluctuations in workstation temperature and humidity and assembly quality, and further optimize the thermal deformation resistance of the support arm material or the micro-compensation mechanism of the fixture.

[0199] Based on this, the system defines the overall performance improvement rate. Its expression is:

[0200] ,

[0201] In the formula, To comprehensively consider multiple process indicators such as fixture loading efficiency, workpiece assembly accuracy, logistics cycle time matching, and energy consumption utilization, a global evaluation function is used as the auxiliary control action. If If the rate of increase continues throughout a continuous task cycle, it proves that the physical intervention of the auxiliary control strategy is effective; otherwise, it is necessary to trigger the parameter rollback of the local auxiliary action or the strategy re-optimization process. This refers to the comprehensive performance index value of the auxiliary control obtained from the system evaluation after the current task cycle is completed. These are the baseline auxiliary control performance index values ​​before the initial launch of this task.

[0202] To quantitatively characterize the robustness and safety of the auxiliary control device, the control anomaly rate is introduced. The calculation method is as follows:

[0203] ,

[0204] in, The number of abnormal physical intervention actions detected (including clamping force deviation, support arm vibration, logistics buffer instability, etc.). The total number of auxiliary actions performed in this round of tasks.

[0205] When controlling the abnormality rate Continuously exceeding the set threshold The system automatically marks the assistive action strategy as "low stability," triggering parameter fine-tuning or replacement mode to ensure the safety boundaries and long-term stability of physical assistive actions.

[0206] At the same time, the system will utilize a self-evolutionary learning module to incorporate auxiliary action history and control anomaly rate. The results are updated using a graph neural network, and the node embedding representation of the working condition map is used. Adaptive optimization will be performed based on the following formula:

[0207] ,

[0208] In this way, after each assembly process is completed, the node embedding representation not only contains the historical experience of assembly actions, but also achieves self-evolution and task adaptation based on the new round of physical history, forming a flexible and transferable auxiliary control capability.

[0209] To ensure operational safety, the residual error of each auxiliary action is displayed in real time and visually. Fixture loading force The system displays the conveyor speed curve of the logistics rollers, alerting station-level operators to deviations and equipment safety boundaries. It also supports a "manual safety window" function, which is activated when auxiliary devices detect excessively high safety risks (such as control anomaly rates). (Sudden rise), automatically requests operator confirmation or intervention for correction, to avoid production line damage caused by uncontrolled mechanized auxiliary actions.

[0210] Finally, the system writes the updated auxiliary action configuration into the work condition map G, forming a complete closed loop of "task execution - physical feedback - action correction - state update," realizing the adaptive evolution and long-term evolution of mechanized auxiliary actions. It is especially suitable for dynamic production lines with multiple product varieties and multiple cycle times, significantly improving the intelligence and safety level of flexible assembly assistance in the workshop.

[0211] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An auxiliary control system for intervening in manufacturing and assembly line anomalies, characterized in that, include: Multi-dimensional physical state acquisition unit: used to collect the operating status of equipment in the production line, the clamping force of fixtures, the workpiece arrival status of logistics path and the workpiece assembly angle error, and generate a multi-dimensional physical state dataset covering process-level actions. Dynamic mapping and anomaly detection unit: used to construct a multidimensional device-based mapping model containing multi-source physical state mapping based on the multidimensional physical state dataset, and to detect residual alignment error of workpiece in real time, wherein the residual alignment error of workpiece includes clamping force overload, support arm posture angle drift or logistics cycle deviation. Multi-process linkage intervention unit: used to automatically generate flexible intervention control commands when an abnormal state is detected, triggering dynamic correction of clamping force, multi-degree-of-freedom flexible fine adjustment of support arm, and adaptive adjustment of logistics buffer roller transmission rate; Auxiliary control unit: used for instruction decomposition and modular action issuance, refining system control instructions into mechanical low-level execution parameters, and executing intervention actions; Closed-loop optimization unit: Based on the real-time acquisition of the clamping force flexible loading curve, support arm posture compensation amplitude and logistics buffer state changes by the sensor array, it constructs feedback comparison of intervention action effect and residual posture error judgment, forming an adaptive correction closed loop of the auxiliary device. After the production line task is completed, the system will systematically collect and evaluate the action history and assembly results of the auxiliary devices in a multi-dimensional manner, forming a continuous optimization mechanism for flexible auxiliary actions in multiple processes. The system first collects full-process time-series data including clamping force, multi-dimensional posture angle vector of support arm, conveying speed of logistics rollers, and temperature and humidity at the workstation. Combining the workstation alignment results with the workpiece fit clearance distribution, a production line feedback quality assessment dataset is constructed. ; Based on production line feedback quality assessment dataset We will conduct dynamic matching accuracy analysis, force-pose coupling characteristic analysis, logistics cycle buffer performance analysis, and environmental condition adaptability analysis. Finally, calculate the overall performance improvement rate. Specifically: ， In the formula, This refers to the comprehensive performance index value of the auxiliary control obtained from the system evaluation after the current task cycle is completed. These are the baseline auxiliary control performance index values ​​before the initial launch of this task.

2. The auxiliary control system for abnormal intervention in manufacturing and assembly lines according to claim 1, characterized in that, The multidimensional physical state dataset includes: Clamping force signal of fixture, multi-dimensional attitude angle vector of support arm, conveying speed of logistics roller and temperature and humidity of workstation.

3. The auxiliary control system for abnormal intervention in manufacturing and assembly lines according to claim 2, characterized in that, The multidimensional physical state acquisition unit generates the multidimensional physical state dataset covering process-level actions through an adjustable weighting coefficient matrix and a nonlinear mapping activation function, specifically: ， In the formula, It is a nonlinear mapping activation function. To cover the multidimensional physical state dataset of process-level actions, Both are weight matrices. For clamping force, For the multi-dimensional attitude angle vector of the support arm, For logistics roller conveyor speed, This refers to the temperature and humidity at the workstation.

4. The auxiliary control system for abnormal intervention in manufacturing and assembly lines according to claim 1, characterized in that, The multi-source physical state mapping includes a clamp module, a flexible support arm buffer structure, and a logistics buffer device. The clamping module has a built-in servo electric cylinder and force sensor, which are used to adjust the clamping force of the clamp according to the real-time physical state of the workpiece. The flexible support arm buffer structure includes a multi-degree-of-freedom rotary joint, a guide rail slider, and an electric actuator, which is used to adaptively adjust the spatial posture of the support arm in real time according to the workpiece assembly posture to compensate for the posture deviation of the workpiece during the insertion process. The logistics buffer device consists of an electric roller drive module, a buffer support frame, a built-in encoder, and a buffer area sensor array. It is used to adjust the transmission speed of the buffer roller in real time according to the changes in the assembly task cycle and the dynamic accumulation of logistics.

5. The auxiliary control system for abnormal intervention in manufacturing and assembly lines according to claim 4, characterized in that, The clamping force of the clamp in the clamp module is adjusted in real time through a clamping control model, wherein the clamping control model is: ； In the formula, For clamping control model, This is the stiffness coefficient. The damping coefficient is... This refers to the residual alignment error of the workpiece. This represents the instantaneous rate of change of the workpiece docking error.

6. The auxiliary control system for abnormal intervention in manufacturing and assembly lines according to claim 4, characterized in that, The transmission speed of the buffer roller is calculated based on a dynamic adjustment model, specifically as follows: ； In the formula, To buffer the roller transfer speed, The reference speed of the drum. This is the cache flexibility adjustment coefficient.

7. The auxiliary control system for abnormal intervention in manufacturing and assembly lines according to claim 1, characterized in that, The dynamic mapping and anomaly detection unit integrates a pose sensor to detect residual alignment error of the workpiece in real time, wherein the residual alignment error of the workpiece is: ； In the formula, This refers to the residual alignment error of the workpiece. Assemble the pose of the desired target. This is the inspection pose after the actual assembly is completed. It is an L2 norm.

8. The auxiliary control system for abnormal intervention in manufacturing and assembly lines according to claim 7, characterized in that, When the multi-process linkage intervention unit detects that the residual alignment error exceeds the set safety tolerance, it triggers multi-dimensional flexible adaptive compensation through the auxiliary control unit, automatically reduces the upper limit of the clamping force of the fixture, extends the logistics buffer cycle, widens the production line buffer zone, and adjusts the support arm fine-tuning angle buffer zone in real time.

9. The auxiliary control system for abnormal intervention in manufacturing and assembly lines according to claim 1, characterized in that, The system also includes: The modular quick-release structure allows for the rapid replacement of clamp modules, flexible support arm buffer structures, and logistics buffer devices via standardized interfaces; Human-machine interface and visual operation interface: used to display the clamping force curve, the support arm posture buffer zone and logistics cycle fluctuation in real time, and supports manual fine-tuning and safety warning.

10. An auxiliary control method for intervening in manufacturing and assembly line anomalies, characterized in that, include: The multi-dimensional physical state acquisition unit collects the operating status of equipment in the production line, the clamping force of fixtures, the workpiece arrival status of logistics path, and the workpiece assembly angle error, generating a multi-dimensional physical state dataset covering process-level actions. Based on the multidimensional physical state dataset, a multidimensional device mapping model containing multi-source physical state mapping is constructed, and real-time detection of fixture clamping force overload, support arm attitude angle drift, or logistics cycle deviation is performed. Anomaly detection is performed through a multi-process linkage intervention unit. When an abnormal state is detected, a flexible intervention control command is automatically generated, triggering dynamic correction of clamping force, multi-degree-of-freedom flexible fine adjustment of support arm, and adaptive adjustment of logistics buffer roller transmission rate. By decomposing instructions and issuing modular actions through the auxiliary control unit, the system control instructions are refined into mechanical low-level execution parameters to execute intervention actions; Based on the real-time acquisition of the clamping force flexible loading curve, support arm posture compensation amplitude and logistics buffer status changes by the sensor array, a feedback comparison of the intervention action effect and residual posture error determination are constructed to form an adaptive correction closed loop for the auxiliary device. Among them, after the production line task is completed, the action history and assembly results of the auxiliary equipment will be systematically collected and the performance will be evaluated in a multi-dimensional feedback manner to form a continuous optimization mechanism for flexible auxiliary actions in multiple processes. First, time-series data of the entire process, including clamping force, multi-dimensional attitude angle vector of support arm, conveying speed of logistics roller, and temperature and humidity at the workstation, were collected. Combined with the workstation alignment results and workpiece clearance distribution, a production line feedback quality assessment dataset was constructed. ; Based on production line feedback quality assessment dataset We will conduct dynamic matching accuracy analysis, force-pose coupling characteristic analysis, logistics cycle buffer performance analysis, and environmental condition adaptability analysis. Finally, calculate the overall performance improvement rate. Specifically: ， In the formula, This refers to the comprehensive performance index value of the auxiliary control obtained from the system evaluation after the current task cycle is completed. These are the baseline auxiliary control performance index values ​​before the initial launch of this task.

Images