A production scheduling method based on new energy wharf tractor

Abstract

The present application relates to the field of new energy vehicle intelligent manufacturing technology, specifically to a production scheduling method based on new energy terminal tractor, comprising: constructing a dynamic production atlas containing multi-stage process parameters and real vehicle operation feedback information; stereoscopic mapping matching the parameter chain in the atlas with the real-time collected manufacturing process parameters to generate a production process state model; performing multi-dimensional stability evaluation based on the model to identify the deviation vector between the process parameters and the performance target; starting a closed-loop compensation mechanism based on the deviation vector to generate a production control instruction set for vehicle body assembly precision and power battery system adjustment; after the control instruction set is corrected by environmental factors, the final execution scheme deployed on the production line is formed. The present application realizes dynamic optimization and cross-domain precise control of the manufacturing process based on real vehicle operation data, and improves the production adaptability and product performance consistency.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology for new energy vehicles, and in particular to a production scheduling method based on new energy terminal tractors. Background Technology

[0002] The production of new energy terminal tractors involves complex processes across multiple stages, including vehicle body assembly and power battery system integration. Currently, the manufacturing process primarily relies on pre-set static process parameter files, with parameter settings for each stage depending on design theory and prior trial production experience. Quality control typically involves setting up inspection points at key workstations to assess the pass / fail status of indicators such as dimensions and electrical performance in stages. Under this model, the setting of process parameters lacks a direct, quantifiable, and dynamic correlation with the long-term operational performance of the final vehicle under real terminal conditions.

[0003] The existing technical solutions suffer from a disconnect between production data and actual vehicle operation data. Deviations in the manufacturing process can only be identified and locally corrected within the current process or a limited number of subsequent processes. They cannot effectively trace back and feed back to specific upstream processes to adjust performance degradation, efficiency fluctuations, or specific failure modes that emerge after the vehicle enters service. Adjustments to process parameters are often triggered by exceeding limits in single-point inspections or rely on trial and error based on expert experience, lacking systematic analysis of the coupled effects of multiple parameters. This makes it difficult for the production system to dynamically adapt to material fluctuations, changes in equipment status, and differentiated usage needs of end users. The optimal match between vehicle assembly quality and battery system performance is difficult to achieve and maintain accurately during the production stage. The production process faces bottlenecks in terms of real-time performance, accuracy, and systematic coordination. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a production scheduling method based on new energy terminal tractors.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a production scheduling method based on new energy terminal tractors, comprising:

[0006] Construct a dynamic production map that includes multi-stage process parameters and real-vehicle operation feedback information;

[0007] The parameter chain in the dynamic production map is matched with the real-time collected manufacturing process parameters in a three-dimensional mapping to generate a production process state model.

[0008] Based on the production process state model, a multi-dimensional stability evaluation is performed to identify the deviation vector between process parameters and performance targets;

[0009] Based on the deviation vector, a closed-loop compensation mechanism is initiated, which is used to generate a set of production control instructions for vehicle assembly accuracy and power battery system calibration.

[0010] The production control instruction set, after being modified by environmental factors, forms the final execution plan deployed on the production line.

[0011] Preferably, the step of constructing a dynamic production map that includes multi-stage process parameters and real-vehicle operation feedback information specifically includes:

[0012] The process parameter set of the tractor is extracted from the historical production database and real-time manufacturing flow at different assembly stages. The process parameter set includes the structural component press-fit force curve, welding heat input distribution and wire harness conduction resistance sequence.

[0013] Simultaneously acquire the operational feedback information of the off-line tractor under simulated dock conditions. The operational feedback information includes the torque response spectrum of the drive motor, the voltage fluctuation trajectory of the battery pack, and the displacement time history data of the suspension system.

[0014] Establish a causal relationship network between the set of process parameters and the operational feedback information. The causal relationship network reveals how specific press-fitting force characteristics affect torque response and how specific heat input patterns are related to voltage stability.

[0015] Using time axis and process flow as the framework, the causal relationship network is encoded into a dynamically updatable graph structure to form the dynamic production graph.

[0016] Preferably, the step of performing a stereo mapping and matching between the parameter chain in the dynamic production map and the real-time acquired manufacturing process parameters specifically includes:

[0017] A multi-sensor array is deployed on the final assembly line to continuously collect manufacturing process parameters of the current tractor frame. These manufacturing process parameters include deformation images of key riveting points, levelness data of the battery compartment tray, and fit waveform of the drive axle mounting surface.

[0018] From the dynamic production map, the standard parameter chain corresponding to the current production batch is retrieved. The standard parameter chain defines the ideal range and mutual constraint relationship of each process parameter.

[0019] Align the manufacturing process parameters with the standard parameter chain in the spatial dimension and synchronize them in the temporal dimension, and perform a stereo mapping matching operation;

[0020] The stereo mapping matching operation outputs a series of matching degree coefficients, and based on the distribution of the matching degree coefficients and the threshold comparison, a production process state model reflecting the difference between the current actual production state of the vehicle and the theoretical model is constructed.

[0021] Preferably, the step of performing multi-dimensional stability evaluation specifically includes:

[0022] The production process state model is decomposed into structural stability dimension, electrical stability dimension, and dynamic matching stability dimension;

[0023] In terms of structural stability, the synthesis error between the deformation image and the fit waveform is analyzed to calculate the rigidity index of the vehicle body basic structure.

[0024] In terms of electrical stability, the correlation between the on-resistance sequence of the wiring harness and the voltage fluctuation trajectory of the battery pack is analyzed, and the attenuation gradient of the electrical connection reliability is calculated.

[0025] In terms of power matching stability, the fluctuation bandwidth of the vehicle's power transmission efficiency is evaluated by combining the torque response spectrum of the drive motor and the displacement time history data of the suspension system.

[0026] The rigidity index, the decay gradient, and the fluctuation bandwidth are aggregated and compared with a preset stability reference surface to identify the deviation vector that characterizes the current production state from the ideal target.

[0027] Preferably, the step of initiating the closed-loop compensation mechanism based on the deviation vector specifically includes:

[0028] The deviation vector is input into a pre-trained compensation decision matrix, which stores the mapping rules between different deviation modes and compensation actions.

[0029] For deviations in vehicle body assembly accuracy, the compensation decision matrix outputs a set of compensation parameters including laser calibration coordinate correction values ​​and hydraulic fixture pressure adjustment values;

[0030] For deviations in the calibration of the power battery system, the compensation decision matrix outputs a set of compensation parameters, including updated values ​​of the battery management system calibration parameters and adaptation parameters of the charging pile handshake protocol.

[0031] The compensation parameters for the vehicle body and the battery system are time-series arranged and logically coupled to form a set of synergistic production control instructions, which aim to bring the production state back to within the stability reference plane.

[0032] Preferably, the step of modifying the production control instruction set via environmental factors specifically includes:

[0033] Monitor real-time environmental factors within the production workshop, including ambient temperature, air humidity, and foundation micro-vibration spectrum;

[0034] A thermal drift influence function of ambient temperature on the laser calibration coordinate correction value is established, and the coordinate correction value is compensated based on the current temperature.

[0035] A friction coefficient influence model is established for the air humidity value on the pressure adjustment value of the hydraulic clamp, and the pressure adjustment value is adaptively adjusted according to the current humidity.

[0036] The interference modes caused by the micro-vibration spectrum of the foundation to the high-precision assembly process are analyzed, and the timing of writing the calibration parameter update value of the battery management system is adjusted to avoid the resonance interference range.

[0037] Integrate all compensation parameters that have been adaptively modified for environmental factors, repackage the instruction format, and generate the final execution plan to be deployed on the production line.

[0038] Preferably, the application steps of the execution plan ultimately deployed on the production line specifically include:

[0039] The execution scheme is broken down into micro-instruction sequences that match specific workstation controllers;

[0040] The micro-instruction sequence is sent to the intelligent tightening shaft on the final assembly line to control it to complete the bolt tightening according to the corrected torque curve.

[0041] The micro-instruction sequence is synchronously sent to the battery pack assembly robot arm, guiding it to complete the battery pack insertion into the compartment at the adjusted path and speed, and triggering the updated handshake protocol to complete the electrical connection self-test;

[0042] At the critical inspection station, the micro-instruction sequence activates the 3D scanner, uses the compensated calibration coordinates as a reference, verifies the key dimensions of the vehicle body, and feeds the verification data back to the update node of the dynamic production map.

[0043] Preferably, it also includes an iterative evolutionary step for producing a knowledge base:

[0044] Collect final inspection data and initial operating data for each tractor unit after applying the aforementioned execution plan;

[0045] The final inspection data after the production line is compared with the initial operation data and the production process state model on which the execution plan of the vehicle is based for reverse correlation analysis.

[0046] Extract new causal patterns or optimized parameter matching relationships discovered in association analysis to form knowledge fragments;

[0047] After the knowledge fragments are evaluated for confidence, they are integrated into the dynamic production graph and the compensation decision matrix to complete the iterative update of the production control logic.

[0048] Preferably, a dynamic priority scheduling mechanism is also included during execution:

[0049] Real-time monitoring of queue status and equipment health at each workstation on the production line, generating a production resource load view;

[0050] When the production process state model identifies a high-risk deviation vector, the dynamic priority scheduling mechanism immediately intervenes.

[0051] The dynamic priority scheduling mechanism allocates higher computing resource priority and a better equipment scheduling sequence to the control tasks calculated by the closed-loop compensation mechanism for handling high-risk deviation vectors based on the production resource load view.

[0052] Ensure that compensation instructions for critical quality deviations can be generated quickly and executed with priority, thereby minimizing the impact of production fluctuations.

[0053] Preferably, the dynamic priority scheduling mechanism, based on the production resource load view, allocates higher computing resource priorities and a better equipment scheduling sequence to the control tasks calculated by the closed-loop compensation mechanism for handling high-risk deviation vectors, using the following method:

[0054] Real-time analysis of the current task queue length and processor utilization of each workstation controller in the production resource load view;

[0055] When the production process state model identifies a high-risk deviation vector, the estimated computational complexity and execution time window of the output result of the compensation decision matrix corresponding to the high-risk deviation vector are extracted.

[0056] The estimated computational complexity is matched with the remaining computing power of each workstation controller to filter out the set of available computing resources;

[0057] Based on the urgency of the execution time window and the distribution of the available computing resources, the highest computing resource priority is allocated to the control task;

[0058] Based on the real-time device status data in the production resource load view, a device instruction path with the lightest load and shortest transmission delay is planned for the control task, and the better device scheduling sequence is generated.

[0059] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0060] By constructing a dynamic production map and mapping it in a three-dimensional manner with real-time manufacturing parameters, historically accumulated feedback data from actual vehicle operation can be integrated into current production decisions. This mechanism establishes a real-time data flow from the product user end to the manufacturing end, transforming vehicle performance in real-world operating environments into inputs for process optimization. The manufacturing process no longer merely follows fixed parameter tables but can dynamically evaluate and reverse-calibrate the front-end process chain based on the actual operating results of the vehicle. Once the correlation pattern between operating performance and specific process parameter sets is identified, the system can proactively adjust the settings of relevant processes, thereby achieving predictive optimization and continuous improvement of the manufacturing process, enhancing the directionality and adaptability of the production process to the final product quality objectives.

[0061] By generating deviation vectors through multi-dimensional stability evaluation and using these vectors to drive a closed-loop compensation mechanism, precise and coordinated control of production deviations is achieved. The deviation vectors comprehensively reflect the overall gap between the output and the target under the coupling effect of multiple parameters. The control instruction set generated based on these vectors can simultaneously and collaboratively adjust key parameters in different domains, such as mechanical assembly and electrical calibration. This solves the matching problems that may arise from independent adjustments of subsystems in traditional methods, directly impacting the optimization of the core coupling relationship between the vehicle body and the battery system. The production system thus possesses the ability to analyze the root causes of complex deviations and execute cross-domain collaborative compensation, enabling the key performance indicators of each product to be proactively controlled to the optimal range, improving the consistency and reliability of product performance. Attached Figure Description

[0062] Figure 1 This is a flowchart of the production scheduling method based on new energy terminal tractors described in this invention;

[0063] Figure 2 A flowchart for constructing a dynamic production graph;

[0064] Figure 3 A flowchart for generating a production process state model using 3D mapping matching;

[0065] Figure 4 A grouped bar chart showing the execution effect of micro-instructions in the production scheduling of tractor vehicles at the new energy terminal;

[0066] Figure 5 This is a grouped bar chart showing the performance indicators of each workstation on the new energy terminal tractor production line. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0068] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0069] See Figure 1 The implementation scheme includes constructing a dynamic production map containing multi-stage process parameters and real vehicle operation feedback information, performing a three-dimensional mapping and matching of the parameter chain in the dynamic production map with the real-time collected manufacturing process parameters to generate a production process state model, performing a multi-dimensional stability evaluation based on the production process state model and identifying the deviation vector between the process parameters and the performance target, and initiating a closed-loop compensation mechanism based on the deviation vector to generate a production control instruction set for vehicle assembly accuracy and power battery system calibration, and finally forming an execution scheme deployed on the production line after the production control instruction set is corrected by environmental factors.

[0070] In one embodiment of the present invention, see [reference] Figure 2 In practical implementation, a dynamic production map is constructed, incorporating multi-stage process parameters and real-vehicle operation feedback information. Taking a new energy terminal tractor production line as an example, a complete set of process parameters for the same model of tractor produced in the past quarter is extracted from the historical production database of the central manufacturing execution system. The set of process parameters is indexed by production batch and vehicle serial number. The structural component pressing force curve is extracted from the historical data of pressure sensors at the chassis assembly station. Each curve records the continuous waveform of pressure changing over time from the start to the end of pressing. The welding heat input distribution is obtained from the process monitoring system of the welding robot. The system records the welding current, voltage, and travel speed of each weld and calculates the heat input distribution along the weld length using formulas.

[0071]

[0072] in: Linear energy, or heat input, is measured in joules per millimeter. The thermal efficiency coefficient represents the welding process. Represents welding voltage. Represents welding current. The welding torch travel speed is represented by the wiring harness continuity resistance sequence, which comes from the electrical safety inspection station before the vehicle leaves the production line. This sequence includes the resistance measurements between various connection points in the high-voltage wiring harness. The process of synchronously acquiring operational feedback information is completed on the factory simulation test platform. The delivered tractor performs standard dock cycle conditions on the test platform. The drive motor torque response spectrum is recorded by a data acquisition card, showing the frequency spectrum of the difference between the motor controller's commanded torque and the actual feedback torque over time. The battery pack voltage fluctuation trajectory is captured by the high-speed data port of the battery management system, recording the dynamic fluctuation curve of the total battery pack voltage during the high-power discharge of the vehicle during simulated loading and unloading. The suspension system displacement time history data is measured by linear displacement sensors installed on the suspension struts, recording the continuous time series data of wheel displacement relative to the vehicle body when the vehicle passes through obstacles set on the test platform.

[0073] In some embodiments, establishing a causal relationship network between the set of process parameters and operational feedback information requires time alignment and feature extraction of the data. The alignment process uses the production number of each vehicle as a key field, correlates the structural component press-fit force curve in the production database with the drive motor torque response spectrum of the same vehicle collected by the test platform, extracts the pressure integral value of the press-fit force curve within the key assembly displacement range as a feature, and extracts the energy value of the torque response spectrum within a specific frequency band as a feature. Statistical methods are used to analyze hundreds of sample pairs, revealing the pattern that vehicle groups with higher pressure integral values ​​have lower energy values ​​in the torque response spectrum within a specific frequency band. This indicates that specific press-fit force characteristics affect the stiffness of the transmission system, thereby changing the torque response characteristics. The correlation analysis between the distribution of welding heat input and the battery pack voltage fluctuation trajectory involves calculating the correlation between the average heat input of a specific weld area on the battery pack housing and the rate of voltage drop during rapid vehicle acceleration. The calculation found that when the average heat input in this area exceeds the upper limit of the process specification, the voltage drop rate of the corresponding vehicle increases significantly, revealing the correlation between specific welding thermal modes and the dynamic performance of the power system.

[0074] It can be understood that a causal relationship network is implemented in a computer system as a directed weighted graph data structure. The vertices of the graph represent feature variables extracted from multi-source data, such as "pressure integral value of the pressing force curve", "average heat input in the weld area", "energy in the characteristic frequency band of the torque response spectrum", and "voltage drop rate". The edges of the graph represent the causal relationships between the feature variables, with the edge direction pointing from process parameter features to operational feedback performance features. The edge weights are assigned by the correlation strength coefficient derived from historical data statistical analysis. Optionally, the causal relationship network supports the mining of multi-level correlations. For example, it can analyze not only the first-order correlation between "stability of the wiring harness on-resistance sequence" and "smoothness of the battery pack voltage fluctuation trajectory", but also the second-order correlation between "voltage fluctuation smoothness" and "drive motor torque response delay", thereby constructing a transmission path from electrical connection to power response.

[0075] In some embodiments, the causal relationship network is encoded into a dynamically updatable graph structure using the time axis and process flow as the coordinates. The time axis is divided into discrete time intervals synchronized with the production cycle, with each interval corresponding to the time consumption of a major assembly process. The process flow is defined as an ordered sequence of workstations from the chassis assembly line to the vehicle roll-off line. The construction of the dynamic production graph involves mapping each causal relationship in the causal relationship network to a specific interval on the time axis based on the specific production stage of the process parameters involved, and simultaneously mapping the physical workstation generated by the parameter to a specific node on the process flow. For example, the feature "pressure integral value of the pressing force curve" is generated at the "chassis assembly workstation," corresponding to the early interval of the time axis and the third node of the process flow. Its causal relationship edge with "torque response spectrum characteristic frequency band energy" is encoded as a connection from the intersection of the early interval of the time axis and the third node of the process flow to the operational feedback data space. This encoding makes the graph contain the spatiotemporal background of causal relationships. It is understandable that the updability of the dynamic production graph is achieved through an iterative algorithm of its internal weight coefficients. When a new vehicle completes production and testing, and its data undergoes feature extraction to generate a new association strength coefficient ρ, the historical weight coefficients of the corresponding edges in the graph will be updated according to the following formula:

[0076]

[0077] in: This represents the causal weight coefficient after this update. This represents the causal weight coefficients before the update. This represents the correlation coefficient calculated based on the latest batch of data. The forgetting factor, representing historical data, is set to 0.9. This formula enables the smooth inheritance of historical experience and the incorporation of new knowledge. Optionally, the dynamic production map update mechanism is triggered not only by data from newly produced vehicles but also by receiving real-time deviation feedback from downstream quality inspection stations. This feedback is incorporated into the map as correction signals for new nodes or edges, ensuring that the map continuously reflects the latest state of the production system.

[0078] In one embodiment of the present invention, see [reference] Figure 3 In practical implementation, the process involves three-dimensional mapping and matching of the parameter chain in the dynamic production map with the real-time acquired manufacturing process parameters to generate a production process state model. Taking a specific production batch of new energy terminal tractor vehicles as an example, a multi-sensor array is deployed at key workstations on the final assembly line. The multi-sensor array includes three industrial cameras installed at the frame assembly workstation to continuously acquire high-resolution deformation images of key riveting points at different angles. The deformation images are processed in real time by an image processing unit to calculate the local strain distribution of the sheet metal after riveting. The multi-sensor array also includes four sets of laser levels installed at the battery compartment installation workstation to continuously acquire the levelness data of the battery compartment pallet relative to the frame reference surface. The levelness data is expressed as tilt in millimeters per meter. The multi-sensor array also includes six sets of high-precision inductive displacement sensors installed at the drive axle assembly workstation to continuously acquire the gap waveform between the drive axle mounting surface and the axle support. The gap waveform records the change in fit along the circumferential direction of the mounting surface with micron-level precision. These data together constitute the real-time manufacturing process parameters of the current tractor vehicle frame. The standard parameter chain corresponding to the current production batch is retrieved from the dynamic production map. The standard parameter chain is generated based on the process specifications and historical high-quality data of this batch of vehicles. It defines the ideal strain threshold range of the deformation image of key rivet points, the allowable tolerance zone of the battery compartment tray levelness data, and the ideal envelope shape and amplitude limit of the drive axle mounting surface fit waveform. At the same time, it defines the mutual constraint relationship between these parameters. For example, the levelness deviation of the battery compartment tray must meet certain spatial coupling conditions with the strain distribution pattern of the key rivet points of the frame.

[0079] In some embodiments, manufacturing process parameters are aligned spatially with a standard parameter chain, and synchronized temporally with a stereo mapping matching operation. Spatial alignment means registering the feature point positions in the deformation images captured by industrial cameras with the ideal riveting point coordinate system defined in the standard parameter chain, mapping the levelness data collected by laser level to the global three-dimensional coordinate grid of the vehicle frame design, and mapping the fit waveform data points collected by displacement sensors to the theoretical design positions of the drive axle mounting surfaces. Temporal synchronization means stamping all manufacturing process parameters for the same vehicle collected by all sensors with a unified timestamp according to the production line cycle, ensuring that data at different assembly stages can be compared within the same production time frame. The core of the stereo mapping matching operation is to calculate the similarity or deviation between each real-time acquired manufacturing process parameter data sequence and the corresponding ideal parameter sequence in the standard parameter chain. For deformation images, the matching operation calculates the structural similarity index between the real-time strain distribution and the ideal strain template. For levelness data, the matching operation calculates the Euclidean distance between the real-time tilt vector and the ideal zero vector. For fit waveforms, the matching operation calculates the dynamic time warping distance between the real-time waveform and the ideal envelope.

[0080] It is understandable that the stereo mapping matching operation ultimately outputs a series of matching coefficients. Each matching coefficient corresponds to the matching result of a process parameter, such as the deformation image matching coefficient, the levelness data matching coefficient, and the fit waveform matching coefficient. The distribution of the matching coefficients is compared with a preset threshold. When the matching coefficient is lower than the threshold, it indicates a significant deviation in that parameter. By integrating the matching status, deviation direction, and deviation magnitude of all parameters, a production process state model is constructed that comprehensively reflects the difference between the current actual production state of the vehicle and the theoretical model. Internally, the production process state model can be represented as a multi-dimensional state vector, where each dimension corresponds to the evaluation result of a process parameter. Optionally, the stereo mapping matching operation not only performs one-to-one parameter comparisons but also cross-parameter correlation matching. For example, it checks whether the riveting point area with a low deformation image matching coefficient is spatially adjacent to the area with an abnormal fit waveform matching coefficient. This correlation matching is a direct application of the mutual constraint relationships between parameters defined in the standard parameter chain, and its matching result, as an additional constraint satisfaction coefficient, is also incorporated into the production process state model.

[0081] The steps for performing multi-dimensional stability evaluation involve decomposing the production process state model into structural stability, electrical stability, and dynamic matching stability dimensions. In the structural stability dimension, the synthesis error between the deformation image and the fit waveform is analyzed. The calculation of this synthesis error is not a simple addition; instead, a fusion model is used to overlay the spatial coordinates of local strain anomaly regions extracted from the deformation image with the position coordinates of excessively large gap points identified in the fit waveform. This overlay analysis is used to identify whether there are systematic assembly errors caused by the accumulation of local deformation. The formula for calculating the stiffness index is:

[0082]

[0083] in: This represents the calculated rigidity index of the vehicle body structure, with a value range of 0 to 1. This represents the number of key anomalies identified. Representing the The composite error value of each outlier point The weight representing the importance of this outlier within the overall structure is defined in the design documents. The theoretical maximum value representing the error at all possible anomalies is used; the closer the stiffness exponent is to 1, the higher the structural stability. In terms of electrical stability, the correlation between the wiring harness on-resistance sequence and the battery pack voltage fluctuation trajectory is analyzed. Correlation analysis calculates the statistical correlation between the dispersion of the wiring harness on-resistance sequence detected before the vehicle leaves the production line and the smoothness index of the typical battery pack voltage fluctuation trajectory of the same model vehicle under simulated dock conditions, retrieved from the dynamic production map, and the attenuation gradient. The attenuation gradient is obtained by calculating the decreasing slope of the correlation coefficient as a function of electrical load (expressed as current). The larger the absolute value, the more pronounced the decrease in electrical connection reliability with increasing load. In terms of power matching stability, the evaluation considers both the drive motor torque response spectrum and the suspension system displacement time history data. The drive motor torque response spectrum is the ideal response spectrum of this vehicle model under standard operating conditions, obtained from dynamic production data. The suspension system displacement time history data is the actual data collected on a vibration test bench before the vehicle leaves the factory. The evaluation process calculates the phase difference and amplitude ratio between the Fourier transform spectrum of the actual displacement data and the ideal torque response spectrum at the main characteristic frequencies, as well as the fluctuation bandwidth. The fluctuation bandwidth is defined as the sum of the standard deviations of these phase differences and amplitude ratios. The larger the value, the worse the consistency of the vehicle's power transmission efficiency.

[0084] It can be understood that the rigidity index, decay gradient, and fluctuation bandwidth are aggregated and compared with a preset stability benchmark, which is defined in three-dimensional space as a surface defined by the lower limit of the rigidity index. Upper limit of decay gradient and fluctuation bandwidth limit The enclosed area will be calculated The three-dimensional vector is compared with this reference plane to identify whether the vector falls within the stable region defined by the reference plane. If it falls outside, the direction and distance from the current vector to the center point of the stable region are the identified deviation vectors. The deviation vectors indicate the direction and magnitude of the current production state's deviation from the ideal target in the three dimensions of structure, electrical, and power matching. In some embodiments, the stability reference plane is not fixed but adaptively adjusted according to the changes in production batches and the updates of the dynamic production map. For example, when the introduction of new materials or new processes causes changes in the performance baseline, the boundary parameters of the stability reference plane will be updated accordingly. Optionally, the execution of multi-dimensional stability evaluation is periodic. Each time the manufacturing process parameters of a vehicle are collected and a production process state model is generated, an evaluation is automatically triggered, generating the corresponding deviation vectors to provide accurate input for subsequent closed-loop compensation.

[0085] In one embodiment of the present invention, in a specific implementation, the process of generating a production control instruction set by activating a closed-loop compensation mechanism based on the deviation vector involves the production process state model outputting a specific deviation vector after multi-dimensional stability evaluation. For example, this deviation vector indicates that the rigidity index is lower than the benchmark value by 0.1 in the structural stability dimension, the attenuation gradient exceeds the benchmark value by 5% in the electrical stability dimension, and the fluctuation bandwidth exceeds the benchmark value by 3% in the dynamic matching stability dimension. This deviation vector is then input into a pre-trained compensation decision matrix. The compensation decision matrix is ​​a multi-dimensional lookup table or rule reasoning engine stored in an industrial control computer. The compensation decision matrix stores the mapping rules between different deviation patterns and effective compensation actions summarized from historical production data. The mapping rules are obtained by training a large number of historical amendment examples using a machine learning algorithm. For deviations in the assembly precision of the vehicle body, the compensation decision matrix matches and searches based on the specific negative deviation of the rigidity index and the location of abnormal areas in the deformation image, and outputs a set of compensation parameters. The compensation parameters include laser calibration coordinate correction values, which are used to guide the adjustment of the coordinate system reference in subsequent measurements or assembly stations. The specific value may be to correct and increase the theoretical X coordinate value of the front reference point of the frame by 0.2 mm. The compensation parameters also include hydraulic clamp pressure adjustment values, which are used to apply different clamping forces to specific clamping points in the next assembly station. For example, the output pressure of the third hydraulic clamp is adjusted from the standard 120 bar to 135 bar.

[0086] For deviations in the calibration of the power battery system, the compensation decision matrix matches and searches based on the positive deviation of the attenuation gradient and the abnormal frequency band of the voltage fluctuation trajectory, and outputs another set of compensation parameters. The compensation parameters include the updated values ​​of the battery management system calibration parameters. The updated values ​​of the battery management system calibration parameters may involve a certain compensation coefficient of the battery state of charge estimation model, which needs to be updated from the current 1.0 to 0.97. The compensation parameters also include the charging pile handshake protocol adaptation parameters. The charging pile handshake protocol adaptation parameters are used to adjust the delay parameters in the specific message sent by the vehicle when communicating with the charging pile for the first time, in order to adapt to the slight changes in electrical characteristics.

[0087] In some embodiments, compensation parameters for the vehicle body and the battery system are time-seriesd and logically coupled to form a set of synergistic production control instructions. The time-series arrangement assigns a precise trigger timestamp to each compensation parameter instruction based on the production line cycle time and the readiness time of each actuator. For example, the hydraulic clamp pressure adjustment instruction is timed to trigger 3 seconds after the vehicle enters the next workstation, while the battery management system calibration parameter update instruction is timed to trigger when the vehicle completes all mechanical assembly and enters the software writing workstation. Logical coupling checks for dependencies or conflicts between different compensation instructions and coordinates them. For example, it ensures that measurement verification instructions based on the new coordinates can only be issued after the laser calibration coordinate correction takes effect. The production control instruction set is structurally represented as a list of instructions with temporal and logical relationships, aiming to pull the production state back to within the stability baseline through orderly and coordinated compensation actions. It is understood that the output of the compensation decision matrix is ​​not fixed; its internal mapping rules are continuously optimized with the iterative evolution of the production knowledge base, giving the closed-loop compensation mechanism the ability to self-improve.

[0088] The production control command set, after undergoing environmental factor correction, is executed before being sent to the production line actuators. Real-time environmental factors within the production workshop are monitored through a sensor network deployed at key locations. These environmental factors include: ambient temperature values ​​continuously measured by temperature sensors installed near the final assembly line (uploaded in real-time in degrees Celsius); air humidity values ​​measured by humidity sensors installed in the battery pack assembly area (expressed as a percentage of relative humidity); and foundation micro-vibration spectra collected by vibration sensors embedded in the foundation, which describe the distribution of vibration energy with frequency in the foundation environment during production line operation. A thermal drift influence function is established for the effect of ambient temperature on the laser calibration coordinate correction value. This function describes the thermal expansion effect of the laser probe itself and the chassis metal materials due to changes in ambient temperature. The function may be a linear compensation model. Compensation calculations are performed on the coordinate correction value based on the current temperature. For example, if the current ambient temperature is 5 degrees Celsius higher than the standard operating temperature, the thermal drift influence function calculates that an additional -0.05 mm compensation is needed on top of the original laser calibration coordinate correction value. A friction coefficient influence model was established to reflect the effect of air humidity on the hydraulic clamp pressure adjustment value. The friction coefficient influence model reflects the change in the friction coefficient between the clamp jaws and the workpiece surface caused by changes in air humidity, which in turn affects the transmission of effective clamping force. The pressure adjustment value was adaptively adjusted according to the current humidity. For example, if the current air humidity is 20% higher than the standard value, the friction coefficient influence model calculates that the hydraulic clamp pressure adjustment value needs to be adjusted from 135 bar to 138 bar to ensure the same clamping effect.

[0089] It is understandable that analyzing the interference modes caused by the foundation micro-vibration spectrum to the high-precision assembly process reveals several characteristic frequency peaks in the foundation micro-vibration spectrum. These characteristic frequency peaks may be coupled with the operating frequency of the production line equipment or external excitations. Interference modes refer to the resonance or precision drift phenomena that may be caused when the inherent frequency of the assembly process (such as the servo control bandwidth of the robotic arm) is close to the main frequency components of the foundation micro-vibration. The phase adjustment of the writing timing of the battery management system calibration parameter update value is performed by fine-tuning the trigger time of the write command to avoid the high-energy interference range in the identified foundation micro-vibration spectrum. For example, the data writing pulse is arranged to be issued within the time window with the lowest vibration spectrum energy to avoid the resonance interference range.

[0090] In some embodiments, the modification of environmental factors is dynamic and predictive. The system not only considers the current instantaneous environmental factors but also combines historical data to predict short-term future environmental change trends and adjusts the compensation parameters accordingly. Optionally, all compensation parameters modified for environmental factors need to be repackaged into instruction formats to match the communication protocols of different actuators. The process of repackaging the instruction format includes adding checksums, converting data byte order, and conforming to the syntax of a specific controller instruction set, ultimately generating a structurally complete, parameter-accurate, and environmentally adaptive final execution plan for deployment on the production line. The logic for integrating the modified parameters is completed by a dedicated environmental compensation middleware. This middleware receives the original production control instruction set and real-time environmental factor data stream, performs calculations and integration based on preset influence functions and models, and outputs the final execution plan.

[0091] In one embodiment of the present invention, the application steps of the execution plan ultimately deployed on the production line are illustrated by an example of a specific execution plan that has been modified by environmental factors. This execution plan is a collection of multiple specific operation instructions and their parameters in terms of data structure. The process of decomposing the execution plan into a sequence of micro-instructions that matches the specific workstation controller is completed by the central scheduling system of the production line. The central scheduling system parses each macro-instruction in the execution plan and converts it into a low-level micro-instruction sequence that can be directly recognized and executed by the corresponding workstation controller according to the target workstation and equipment type of the instruction. The format of the micro-instruction sequence follows the proprietary protocol of each controller manufacturer or the industry standard fieldbus protocol. For example, for a smart tightening shaft controller, the micro-instruction sequence may include a series of atomic operation instructions and their parameters such as "axis start", "position to coordinates (X,Y,Z)", "run in torque mode", "apply torque according to the given curve", and "complete signal feedback". For a battery pack assembly robotic arm controller, the micro-instruction sequence may include instructions such as "path point planning", "speed and acceleration setting", "end effector action", and "communication protocol trigger".

[0092] In some embodiments, a micro-instruction sequence is sent to the intelligent tightening shaft on the assembly line to control it to complete bolt tightening according to the modified torque curve. The modified torque curve is derived from the tightening strategy update associated with the hydraulic clamp pressure adjustment value in the execution scheme. After receiving the micro-instruction sequence, the intelligent tightening shaft controller parses the bolt position coordinate sequence that needs to be tightened and loads the torque-angle-time curve parameters attached to the micro-instruction. This curve is different from the standard curve, and its pre-tightening stage slope, target torque value, or holding time may have been adjusted according to compensation logic. The intelligent tightening shaft monitors the torque and angle in real time during the tightening process to ensure that the actual output strictly follows the modified torque curve defined by the micro-instruction sequence, and uploads the actual tightening process data (such as final torque, angle, and yield point) back to the system after completion. It is understandable that the micro-instruction sequence is synchronously sent to the battery pack assembly robot arm to guide it to complete the battery pack insertion into the battery compartment at an adjusted path and speed, and to trigger the updated handshake protocol to complete the electrical connection self-test. The adjustment of the path and speed may be due to the response of the laser calibration coordinate correction value in the execution plan. The robot arm control system re-plans the motion trajectory based on the new path point coordinates and speed parameters in the micro-instruction. After the battery pack is smoothly transported and accurately placed into the battery compartment tray, the communication interface at the end of the robot arm will execute a specific command in the micro-instruction sequence. This command writes the updated charging pile handshake protocol adaptation parameters into the connector communication module between the battery pack and the vehicle, and then triggers an electrical connection self-test process. The self-test process will use the new protocol parameters to perform message exchange and insulation detection, and feed the self-test results back to the control system.

[0093] At key inspection stations, micro-instruction sequences activate 3D scanners to verify key dimensions of the vehicle body using compensated calibration coordinates. Upon receiving the trigger micro-instruction, the 3D scanner adjusts the origin or reference axis of its scanning coordinate system based on the compensated laser calibration coordinate correction value carried in the instruction. It then performs high-speed non-contact scanning of key features on the vehicle body, such as door frame diagonals and chassis mounting holes, acquiring point cloud data. The verified data is then fed back to the update node of the dynamic production map. The returned data packet contains the vehicle identification code, scan timestamp, measured dimensions, and deviations from the compensated theoretical values. This data is transmitted to the data access layer of the dynamic production map system via the workshop network. Optionally, the issuance and execution of micro-instruction sequences follow a strict time synchronization and condition triggering mechanism. When a vehicle arrives at a specific station, the vehicle identification system on the production line (such as RFID) triggers the station controller to request the corresponding micro-instruction sequence from the central scheduling system, ensuring precise matching between the instruction and the physical vehicle.

[0094] After implementing the execution plan for each tractor unit, we collect final inspection data and initial operating data. The final inspection data comes from the vehicle inspection line and includes hundreds of indicators such as braking performance test values, headlight angles, emissions tests (if applicable), and airtightness test results. The initial operating data refers to the operating status data periodically transmitted back through the onboard telematics system within the first three months after vehicle delivery, including average energy consumption, fault code frequency, and operating temperature and voltage-current curve segments of key components (such as motors and batteries). We then perform a reverse correlation analysis between the final inspection data and the initial operating data and the production process state model on which the vehicle execution plan was based. This reverse correlation analysis traces the vehicle's final performance (final inspection and operating data) back to its specific state during the production process (production process state model), analyzing the statistical relationship between various parameters in the process state (such as rigidity index and attenuation gradient) and the final performance indicators. This process aims to verify the accuracy of the causal relationships established by the dynamic production map and the effectiveness of the compensation measures.

[0095] It is understandable that extracting new causal patterns or optimized parameter matching relationships discovered in correlation analysis forms knowledge fragments. New causal patterns might manifest as a waveform feature of fit that was not significantly marked in the production process state model, but is found to have a strong correlation with a specific anomaly in the initial operating data. Optimized parameter matching relationships might manifest as the updated values ​​of the battery management system calibration parameters recommended for a certain attenuation gradient deviation in the original compensation decision matrix, which have been proven by a large amount of data to be further optimized by 0.5% to obtain better voltage stability. These findings are structured into new knowledge fragments, which contain a triplet of preconditions (production process state characteristics), actions (compensation instructions or parameters), and results (changes in performance indicators). In some embodiments, the confidence level of the knowledge fragment is assessed using statistical methods, such as calculating the frequency of the pattern or relationship in historical data and its consistency with the resulting outcome. A simple confidence assessment formula can be expressed as:

[0096]

[0097] in: The confidence score representing a knowledge segment. This represents the number of cases where the preconditions for the knowledge fragment are met and the result is as expected. This represents the total number of cases where the preconditions are met. This represents the standard deviation of the outcome indicator across all valid cases. The average value of the change in the representative result indicator is considered. A higher confidence score indicates a more reliable knowledge fragment. After confidence evaluation, the knowledge fragment is integrated into the dynamic production graph and compensation decision matrix to iteratively update the production control logic. For the dynamic production graph, the integration operation may involve adding an edge to the causal relationship network or updating the weight coefficients of existing edges. For the compensation decision matrix, the integration operation may involve adding a new "IF-THEN" rule to the mapping rule base or modifying the output parameters in existing rules. This allows the entire production scheduling system to learn from the actual performance of each produced vehicle and continuously optimize the production accuracy and quality of subsequent vehicles. Optionally, the integration of knowledge fragments is not immediately effective but requires an offline testing and verification phase. During the verification phase, the new knowledge fragment is applied to a simulated production environment for deduction. Only after confirming its logical rationality and lack of conflict is it officially released to the online dynamic production graph and compensation decision matrix (see Table 1).

[0098] Table 1: Microinstruction Sequence Decomposition and Application Table

[0099]

[0100] See Figure 4 This is a grouped bar chart illustrating the performance of micro-instructions in the production scheduling of new energy terminal tractor units. It primarily showcases the performance of four types of micro-instructions across different dimensions. All micro-instructions have an accuracy rate close to 100%, reflecting the reliability of the production scheduling system's instruction execution. The path guidance instruction takes the longest (approximately 25 seconds), while the protocol writing instruction takes the shortest (approximately 5 seconds), consistent with the difference in business complexity between robotic arm path planning and communication protocol writing. The protocol writing instruction has a relatively high latency (approximately 15 × 0.01 seconds), while the calibration activation instruction has the lowest latency (approximately 5 × 0.01 seconds), reflecting the instruction processing efficiency of different devices. The retry rate for all instructions is close to 0, indicating high stability in instruction execution. This chart is used to evaluate the execution performance of micro-instructions on the new energy tractor unit production line, helping technicians identify optimization opportunities for different instructions and ensuring efficient and accurate production scheduling.

[0101] In one embodiment of the present invention, the production scheduling method based on new energy terminal tractors includes a dynamic priority scheduling mechanism during execution. Taking a final assembly line with twenty workstations as an example, the real-time monitoring of the queue status and equipment health of each workstation is achieved through the agent program deployed on each workstation controller and the workshop IoT gateway. The queue status information includes the number of vehicle identification codes waiting to be processed in the buffer at the entrance of each workstation, the percentage of completion of the current process step for each vehicle, and the estimated waiting time. The equipment health information is obtained from the equipment management system and includes the servo motor current fluctuation value of the intelligent tightening shaft, the most recent calibration deviation of the positioning repeatability accuracy of the robotic arm, the light source intensity attenuation coefficient of the 3D scanner, and the port error packet rate of the network switch. After these data are aggregated, they are integrated and visualized by the scheduling server to generate a production resource load view. The production resource load view is presented on the monitoring screen in the form of a topology map, where the size and color of each workstation node represent its load, and the thickness of the connection between nodes represents the real-time latency of data transmission. When the production process state model identifies a high-risk deviation vector, the dynamic priority scheduling mechanism immediately intervenes. The determination of a high-risk deviation vector is based on preset multi-dimensional thresholds. For example, when the identified deviation vector deviates more than 0.15 in the rigidity index of the structural stability dimension or more than 8% in the attenuation gradient of the electrical stability dimension, the system marks it as a high-risk event. The intervention of the dynamic priority scheduling mechanism is manifested by interrupting the current regular task scheduling cycle and starting a high-priority compensation task scheduling session.

[0102] In some embodiments, the dynamic priority scheduling mechanism allocates higher computing resource priority and a better equipment scheduling sequence to the control tasks calculated by the closed-loop compensation mechanism for handling high-risk deviation vectors based on the production resource load view. The allocation of higher computing resource priority occurs in the resource management module of the scheduling server. This module moves the computational processes corresponding to the control tasks (such as real-time querying of the compensation decision matrix and rapid calculation of compensation parameters) from the ordinary queue to the real-time queue, allocating them more CPU time slices and higher memory access permissions. Allocating a better equipment scheduling sequence involves planning an optimal execution path in terms of time and resource conflict for the specific production instructions driven by the control tasks (such as hydraulic fixture pressure adjustment instructions and laser calibration coordinate correction instructions). This ensures that compensation instructions for critical quality deviations can be quickly generated and prioritized for execution, thereby minimizing the impact of production fluctuations. It can be understood that the core of the dynamic priority scheduling mechanism lies in its responsiveness and optimization capabilities. It ensures that when deviations occur during production that may significantly affect the final vehicle performance, the decision-making and execution processes used to correct these deviations can obtain system resource support exceeding the normal production pace, avoiding delays in correction due to queuing.

[0103] The dynamic priority scheduling mechanism allocates higher computing resource priority and better equipment scheduling sequence to the control tasks calculated by the closed-loop compensation mechanism for handling high-risk deviation vectors based on the production resource load view. The method is to analyze the current task queue length and processor utilization of each workstation controller in the production resource load view in real time. The analysis process continuously reads the operating system performance counters and message queue depth of each workstation controller through the monitoring agent of the scheduling server. For example, it reads that the current task queue of PLC controller A located in the chassis assembly workstation has 5 pending instruction blocks and its CPU utilization is consistently at 78%, while the task queue length of the robotic arm controller B in the battery pack assembly workstation is 2 and its CPU utilization is 45%. When the production process state model identifies a high-risk deviation vector, it extracts the estimated computational complexity and execution time window of the compensation decision matrix output corresponding to the high-risk deviation vector. The estimated computational complexity is estimated based on the query dimension of the compensation decision matrix, the historical average computation time, and the number of features of the current input vector. The execution time window is calculated based on the current position of the vehicle on the production line, its speed, and the remaining time to reach the target compensation station. For example, it is estimated that the matrix query and parameter calculation required to process the current high-risk deviation vector will occupy one CPU core for about 150 milliseconds, and the compensation command must be issued to the adjustment station located in the middle of the production line within the next 90 seconds.

[0104] The estimated computational complexity is matched with the remaining computing power of each workstation controller to filter out the available computing resource set. Remaining computing power is estimated by subtracting the actual computing power corresponding to the current processor utilization from the workstation controller's nominal computing power, and then subtracting the sum of the estimated time consumption of all tasks in its current task queue. The matching process involves finding controller nodes whose remaining computing power is greater than or equal to the estimated computational complexity of this control task; these controller nodes constitute the available computing resource set. Based on the urgency of the execution time window and the distribution of the available computing resource set, the control task is assigned the highest computing resource priority. The priority can be numerically represented by a priority scoring function.

[0105]

[0106] in: This represents a dynamic priority score for the control task; the higher the score, the higher the priority of the resources received. This represents the number of seconds remaining in the execution time window. This represents a predefined baseline time constant whose value is related to the standard cycle time of the production line, and is used to normalize the time dimension. This represents the ratio of the remaining computing power of the optimal candidate controller node to the estimated computational complexity. It is a weighting coefficient between 0 and 1, used to balance time urgency and computing resource sufficiency, for example, setting... This indicates a preference for time urgency. Based on real-time equipment status data in the production resource load view, a device instruction path with the lightest load and shortest transmission delay is planned for the control task, generating a better device scheduling sequence. The real-time equipment status data includes network link latency, communication bandwidth utilization between controllers at each workstation, and material flow time between adjacent workstations. The planning process can be modeled as a shortest path problem in a weighted directed graph. The nodes of the graph are the various device controllers participating in the compensation execution, and the edge weights combine transmission delay, queue waiting time, and device switching costs. An algorithm is used to solve for an instruction transmission and execution sequence with the minimum total cost, which is the better device scheduling sequence. Optionally, when planning the device instruction path, the logical dependencies between compensation instructions must also be considered. For example, the writing of battery management system calibration parameters can only be performed after the battery pack is physically installed. The dynamic priority scheduling mechanism will treat these constraints as hard constraints for path planning, ensuring that the generated device scheduling sequence is not only optimized in terms of time and resources but also logically feasible. In some embodiments, when multiple high-risk deviation vectors occur simultaneously, the dynamic priority scheduling mechanism independently calculates the priority score and equipment scheduling sequence for each corresponding control task, and coordinates the competition of these high-priority tasks for shared resources (such as a high-speed bus or a key robotic arm) through a global arbitrator. The arbitration strategy may be based on the priority score of the task or the potential severity of its impact on the final vehicle quality.

[0107] See Figure 5 This is a grouped bar chart showing the performance indicators of each workstation on the new energy terminal tractor production line. It primarily displays the performance of different workstations in three dimensions: load, equipment health, and response time. The battery pack assembly and electrical system workstations have the highest load (over 85%), reflecting the heavy workload of assembling core components. The interior installation workstation has the highest health (over 85%), while the battery pack assembly workstation has a relatively lower health (approximately 68%), requiring attention to equipment maintenance at high-load workstations. This chart is used to assess the operational status of each workstation on the production line, helping dispatchers identify high-load, low-health workstations (such as battery pack assembly) to optimize task allocation and increase equipment maintenance resources, ensuring the overall efficiency and stability of the production line.

[0108] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A production scheduling method based on new energy terminal tractors, characterized in that, include: Construct a dynamic production map that includes multi-stage process parameters and real-vehicle operation feedback information; The parameter chain in the dynamic production map is matched with the real-time collected manufacturing process parameters in a three-dimensional mapping to generate a production process state model. Based on the production process state model, a multi-dimensional stability evaluation is performed to identify the deviation vector between process parameters and performance targets; Based on the deviation vector, a closed-loop compensation mechanism is initiated, which is used to generate a set of production control instructions for vehicle assembly accuracy and power battery system calibration. The production control instruction set, after being modified by environmental factors, forms the final execution plan deployed on the production line; The steps for constructing a dynamic production map that includes multi-stage process parameters and real-vehicle operation feedback information specifically include: The process parameter set of the tractor at different assembly stages is extracted from the historical production database and real-time manufacturing process. The process parameter set includes the structural component press-fit force curve, welding heat input distribution and wire harness conduction resistance sequence. Simultaneously acquire the operational feedback information of the off-line tractor under simulated dock conditions. The operational feedback information includes the torque response spectrum of the drive motor, the voltage fluctuation trajectory of the battery pack, and the displacement time history data of the suspension system. Establish a causal relationship network between the set of process parameters and the operational feedback information. The causal relationship network reveals how specific press-fitting force characteristics affect torque response and how specific heat input patterns are related to voltage stability. Using time axis and process flow as the framework, the causal relationship network is encoded into a dynamically updatable graph structure to form the dynamic production graph.

2. The production scheduling method based on new energy terminal tractors according to claim 1, characterized in that, The step of performing a three-dimensional mapping and matching between the parameter chain in the dynamic production map and the real-time acquired manufacturing process parameters specifically includes: A multi-sensor array is deployed on the final assembly line to continuously collect manufacturing process parameters of the current tractor frame. These manufacturing process parameters include deformation images of key riveting points, levelness data of the battery compartment tray, and fit waveform of the drive axle mounting surface. From the dynamic production map, the standard parameter chain corresponding to the current production batch is retrieved. The standard parameter chain defines the ideal range and mutual constraint relationship of each process parameter. Align the manufacturing process parameters with the standard parameter chain in the spatial dimension and synchronize them in the temporal dimension, and perform a stereo mapping matching operation; The stereo mapping matching operation outputs a series of matching degree coefficients, and based on the distribution of the matching degree coefficients and the threshold comparison, a production process state model reflecting the difference between the current actual production state of the vehicle and the theoretical model is constructed.

3. The production scheduling method based on new energy terminal tractors according to claim 2, characterized in that, The steps for performing the multi-dimensional stability evaluation specifically include: The production process state model is decomposed into structural stability dimension, electrical stability dimension, and dynamic matching stability dimension; In terms of structural stability, the synthesis error between the deformation image and the fit waveform is analyzed to calculate the rigidity index of the vehicle body basic structure. In terms of electrical stability, the correlation between the on-resistance sequence of the wiring harness and the voltage fluctuation trajectory of the battery pack is analyzed, and the attenuation gradient of the electrical connection reliability is calculated. In terms of power matching stability, the fluctuation bandwidth of the vehicle's power transmission efficiency is evaluated by combining the torque response spectrum of the drive motor and the displacement time history data of the suspension system. The rigidity index, the decay gradient, and the fluctuation bandwidth are aggregated and compared with a preset stability reference surface to identify the deviation vector that characterizes the current production state from the ideal target.

4. The production scheduling method based on a new energy terminal tractor as described in claim 3, characterized in that, The step of initiating the closed-loop compensation mechanism based on the deviation vector specifically includes: The deviation vector is input into a pre-trained compensation decision matrix, which stores the mapping rules between different deviation modes and compensation actions. For deviations in vehicle body assembly accuracy, the compensation decision matrix outputs a set of compensation parameters including laser calibration coordinate correction values ​​and hydraulic fixture pressure adjustment values; For deviations in the calibration of the power battery system, the compensation decision matrix outputs a set of compensation parameters, including updated values ​​of the battery management system calibration parameters and adaptation parameters of the charging pile handshake protocol. The compensation parameters for the vehicle body and the battery system are time-series arranged and logically coupled to form a set of synergistic production control instructions, which aim to bring the production state back to within the stability reference plane.

5. A production scheduling method based on a new energy terminal tractor as described in claim 4, characterized in that, The production control instruction set, through environmental factor correction steps, specifically includes: Monitor real-time environmental factors within the production workshop, including ambient temperature, air humidity, and foundation micro-vibration spectrum; A thermal drift influence function of ambient temperature on the laser calibration coordinate correction value is established, and the coordinate correction value is compensated based on the current temperature. A friction coefficient influence model is established for the air humidity value on the pressure adjustment value of the hydraulic clamp, and the pressure adjustment value is adaptively adjusted according to the current humidity. The interference modes caused by the micro-vibration spectrum of the foundation to the high-precision assembly process are analyzed, and the timing of writing the calibration parameter update value of the battery management system is adjusted to avoid the resonance interference range. Integrate all compensation parameters that have been adaptively modified for environmental factors, repackage the instruction format, and generate the final execution plan to be deployed on the production line.

6. A production scheduling method based on a new energy terminal tractor as described in claim 5, characterized in that, The application steps of the final deployment plan on the production line specifically include: The execution scheme is broken down into micro-instruction sequences that match specific workstation controllers; The micro-instruction sequence is sent to the intelligent tightening shaft on the final assembly line to control it to complete the bolt tightening according to the corrected torque curve. The micro-instruction sequence is synchronously sent to the battery pack assembly robot arm, guiding it to complete the battery pack insertion into the compartment at the adjusted path and speed, and triggering the updated handshake protocol to complete the electrical connection self-test; At the critical inspection station, the micro-instruction sequence activates the 3D scanner, uses the compensated calibration coordinates as a reference, verifies the key dimensions of the vehicle body, and feeds the verification data back to the update node of the dynamic production map.

7. A production scheduling method based on a new energy terminal tractor as described in claim 6, characterized in that, It also includes an iterative evolutionary step for producing a knowledge base: Collect final inspection data and initial operating data for each tractor unit after applying the aforementioned execution plan; The final inspection data after the production line is compared with the initial operation data and the production process state model on which the execution plan of the vehicle is based for reverse correlation analysis. Extract new causal patterns or optimized parameter matching relationships discovered in association analysis to form knowledge fragments; After the knowledge fragments are evaluated for confidence, they are integrated into the dynamic production graph and the compensation decision matrix to complete the iterative update of the production control logic.

8. A production scheduling method based on a new energy terminal tractor as described in claim 7, characterized in that, It also includes a dynamic priority scheduling mechanism during execution: Real-time monitoring of queue status and equipment health at each workstation on the production line, generating a production resource load view; When the production process state model identifies a high-risk deviation vector, the dynamic priority scheduling mechanism immediately intervenes. The dynamic priority scheduling mechanism allocates higher computing resource priority and a better equipment scheduling sequence to the control tasks calculated by the closed-loop compensation mechanism for handling high-risk deviation vectors based on the production resource load view. Ensure that compensation instructions for critical quality deviations can be generated quickly and executed with priority, thereby minimizing the impact of production fluctuations.

9. A production scheduling method based on a new energy terminal tractor as described in claim 8, characterized in that, The dynamic priority scheduling mechanism, based on the production resource load view, allocates higher computing resource priorities and a better equipment scheduling sequence to the control tasks calculated by the closed-loop compensation mechanism for handling high-risk deviation vectors, as follows: Real-time analysis of the current task queue length and processor utilization of each workstation controller in the production resource load view; When the production process state model identifies a high-risk deviation vector, the estimated computational complexity and execution time window of the output result of the compensation decision matrix corresponding to the high-risk deviation vector are extracted. The estimated computational complexity is matched with the remaining computing power of each workstation controller to filter out the set of available computing resources; Based on the urgency of the execution time window and the distribution of the available computing resources, the highest computing resource priority is allocated to the control task; Based on the real-time device status data in the production resource load view, a device instruction path with the lightest load and shortest transmission delay is planned for the control task, and the better device scheduling sequence is generated.

Images