Automobile wire harness terminal pull-out force detection method

Abstract

This invention relates to the field of mechanical engineering and discloses a method for detecting the pull-out force of automotive wiring harness terminals. The method includes: real-time acquisition of pressure, stroke, speed, temperature, and wire core compression ratio using a sensor array during the crimping process; obtaining the actual pull-out force value at the pull-out force testing station; constructing a process-performance sample set and training a pull-out force prediction regression model; optimizing crimping parameters using a gradient descent algorithm based on the model's prediction results, and sending the results to a servo control system for adaptive adjustment. By integrating a multi-source sensor array and pull-out force testing equipment on the crimping production line, a complete data link from process execution to performance verification is established, solving the problem of information separation between testing and production in traditional models; and by using a machine learning model to accurately model the nonlinear mapping relationship between crimping parameters and pull-out force, the method eliminates reliance on operator experience.

Description

Technical Field

[0001] This invention belongs to the field of mechanical engineering, specifically relating to a method for detecting the pull-out force of automotive wiring harness terminals. Background Technology

[0002] With the continuous improvement of automotive electronics and intelligence, wiring harnesses, as the "neural network" of the vehicle's electrical system, directly affect the vehicle's safety and functional stability through their connection reliability. Terminal crimping is a key process in wiring harness manufacturing, and its quality is usually verified through pull-out force testing, which measures the minimum force required for the terminal and wire to separate under axial tensile force. This indicator not only reflects the mechanical bonding strength of the crimped interface but also indirectly characterizes the electrical contact performance and long-term service reliability. Therefore, pull-out force testing has become an indispensable quality control method in the wiring harness production process.

[0003] Terminal pull-out force testing methods primarily rely on offline sampling tests or online fixed-parameter testing. Traditional solutions typically separate pull-out force testing from the crimping production process, making it difficult to provide real-time feedback of test data to the crimping equipment control system. Process engineers must manually adjust key parameters such as crimping machine pressure, stroke, and die clearance based on experience, leading to delayed parameter optimization, high subjectivity, and an inability to adapt to dynamic changes in wire specifications, terminal materials, or environmental temperature and humidity. This break in the "test-analysis-adjustment" chain makes it difficult to guarantee consistent crimping quality, easily resulting in defects such as false crimping and over-crimping, which in turn can lead to increased contact resistance, signal interruption, and even short-circuit risks.

[0004] While sensors are introduced to monitor the force-displacement curve during the crimping process, this method remains limited to single-process judgment and lacks closed-loop learning and adaptive capabilities based on historical and real-time pull-out force data. Especially in flexible production scenarios with multiple product types and small batches, traditional methods cannot achieve rapid migration and accurate adaptation of crimping parameters when facing frequently changing wire harness models. Furthermore, the testing strategy itself is statically set, failing to dynamically adjust the sampling frequency or test thresholds based on current process stability, resulting in both wasted testing resources and the risk of missed detections. Summary of the Invention

[0005] This invention provides a method for testing the pull-out force of automotive wiring harness terminals. By constructing a closed-loop feedback mechanism between the production process and testing data, it achieves automatic iterative optimization of crimping process parameters, thereby improving the consistency of terminal crimping quality and production efficiency. The method establishes a multi-dimensional feature fusion model based on real-time collected physical parameters of the crimping process, pull-out force test results, and historical process data. It dynamically generates optimal crimping parameter instructions and drives the crimping equipment to perform adaptive adjustments, ensuring that the pull-out force index of each batch of products remains stably within a preset acceptable range.

[0006] This invention provides a method for detecting the pull-out force of automotive wiring harness terminals, comprising:

[0007] A sensor array is deployed at the crimping station to collect data in real time on the applied pressure, crimping stroke, crimping speed, die temperature, and wire core compression ratio during the crimping process.

[0008] Obtain the actual pull-out force value of the crimped terminal sample at the pull-out force testing station;

[0009] The physical parameters of the pressing process are time-aligned and data-bound with the corresponding actual pull-out force values ​​to form a tagged process-performance sample set;

[0010] Based on the process-performance sample set, a pull-out force prediction regression model is constructed. This model takes the physical parameters of the pressing process as input and the predicted pull-out force value as output.

[0011] The pull-out force prediction regression model is used to predict the pull-out force of the current batch of crimping parameter combinations. If the predicted value deviates from the preset qualified range, the parameter optimization module is activated.

[0012] The parameter optimization module uses the gradient descent algorithm to iteratively calculate the new pressing pressure value and pressing stroke value by taking the objective function as minimizing the square error between the predicted pull-out force value and the median target pull-out force value.

[0013] The new crimping pressure and crimping stroke values ​​are sent to the servo control system of the crimping equipment to update the parameters for the next round of crimping operations.

[0014] Furthermore, the crimping process sensing array includes a high-precision strain gauge pressure sensor, a laser displacement sensor, an encoder, a thermocouple temperature sensor, and an image recognition unit. The high-precision strain gauge pressure sensor is installed at the end of the piston rod of the crimping cylinder and is used to measure the instantaneous pressure applied to the terminal during the crimping process. Its range is 0 to 10 kN, and its accuracy is 0.5%. The laser displacement sensor is aligned with the movement trajectory of the crimping mold and is used to measure the crimping stroke in real time. Its resolution is 1 micrometer. The encoder is coaxially connected to the crimping spindle and is used to record the crimping speed. Its sampling frequency is 1000 Hz. The thermocouple temperature sensor is embedded inside the crimping mold, no more than 5 mm away from the crimping contact surface, and is used to monitor the working temperature of the mold. The image recognition unit consists of an industrial camera and a ring light source. The industrial camera has a frame rate of 120 frames per second and a resolution of no less than 2 million pixels. It is used to capture the terminal cross-section image after crimping and calculate the core compression ratio through an edge detection algorithm.

[0015] Furthermore, the pull-out force testing station adopts a servo motor-driven tensile testing machine with a maximum tensile force of 5000 N, a force sensor accuracy of 0.2%, and a clamping mechanism with a self-aligning function to ensure that the pull-out direction coincides with the terminal axis. The pull-out force test is completed within 30 minutes after crimping, and the test environment temperature is controlled at 25 degrees Celsius ± 2 degrees Celsius, and the relative humidity is controlled at 50% ± 5%. For each test, 5 samples from the same batch of continuous production are selected, and the average pull-out force is taken as the actual pull-out force value of that batch.

[0016] Furthermore, the process of constructing the process-performance sample set includes: assigning a unique serial number to each crimped workpiece; during the crimping process, synchronously recording the data of each sensor with a sampling period of 10 milliseconds and associating it with the serial number; during the pull-out force test, reading the sample with the corresponding serial number and writing the measured pull-out force value into the data record of that serial number; and storing all data in a central database and indexing it by timestamp, equipment number, wire harness model, and terminal specification.

[0017] Furthermore, the pull-out force prediction regression model adopts a multilayer perceptron neural network structure, including an input layer, two hidden layers, and an output layer; the input layer has 5 nodes, corresponding to pressure value, pressing stroke, pressing speed, die temperature, and wire core compression ratio, respectively; each hidden layer contains 30 neurons, and the activation function is a modified linear unit; the output layer is a single node, outputting the predicted pull-out force value; the model training uses the mean squared error loss function, the optimizer is the adaptive moment estimation algorithm, the initial learning rate is 0.001, and the batch size is 64; incremental training is triggered every 500 newly added valid samples.

[0018] Furthermore, the preset qualified range is pre-set according to the terminal specifications, with its lower limit not lower than the minimum pull-out force value specified by the industry standard, and its upper limit not exceeding 90% of the theoretical maximum pull-out force value corresponding to the yield strength of the terminal material; the median value of the target pull-out force is the arithmetic mean of the qualified range.

[0019] Furthermore, when the parameter optimization module performs gradient descent iteration, the constraints include: the crimping pressure value must not be lower than the minimum safe pressure threshold of the equipment, and must not be higher than the pressure limit corresponding to the fatigue limit of the mold material; the crimping stroke must not be less than the minimum stroke required for the wire core to be fully embedded in the terminal, and must not be greater than the mold closing limit position; each parameter adjustment range shall not exceed 5% of the current value to prevent system oscillation.

[0020] Furthermore, after the parameters are updated, a full range of tests are performed on the first product, including pull-out force test, cross-sectional microscopic observation and conduction resistance measurement. If the results of the full range of tests meet the quality specifications, the parameter update is confirmed to be valid and the new parameters are used as the benchmark for subsequent production. If they do not meet the requirements, the process is rolled back to the previous valid parameter set and a manual review process is triggered.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0022] 1. By integrating a multi-source sensor array and pull-out force testing equipment into the crimping production line, a complete data link from process execution to performance verification was established, solving the problem of information separation between testing and production processes in the traditional model;

[0023] 2. By using machine learning models to accurately model the nonlinear mapping relationship between pressing parameters and pull-out force, the reliance on operator experience is eliminated;

[0024] 3. An automatic parameter optimization mechanism based on gradient descent was introduced, which realized the closed-loop adaptive adjustment of crimping process parameters, significantly improved the consistency of terminal crimping quality, and increased the pull-out force qualification rate from 85% of the traditional method to more than 99%.

[0025] 4. By limiting the range of parameter adjustments and setting physical constraint boundaries, the safety and stability of process adjustments are ensured, avoiding equipment damage or batch scrapping caused by radical adjustments. This method can be seamlessly integrated into existing smart manufacturing production lines without replacing core equipment, only requiring the addition of sensing modules and edge computing units, and has good engineering feasibility and promotion value. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall technical solution architecture of the automotive wiring harness terminal pull-out force detection method proposed in this invention;

[0027] Figure 2 This is a schematic diagram of the core principle framework of pull-off force prediction and parameter adaptive adjustment based on multi-dimensional feature fusion model and gradient descent optimization in this invention;

[0028] Figure 3 This is a logical flowchart of the process-performance sample set formed by binding the sensor array data acquisition and pull-out force test results during the crimping process in this invention.

[0029] Figure 4 This is a flowchart illustrating the logical flow of the pull-off force prediction regression model construction and incremental training mechanism in this invention.

[0030] Figure 5 This is a flowchart illustrating the logical flow of the parameter optimization module in this invention, which performs constrained gradient descent iterations to generate new pressing parameters.

[0031] Figure 6 This is a schematic diagram of the multi-level interaction relationship and data flow between the crimping device, the sensing system and the central database in this invention. Detailed Implementation

[0032] Please refer to the attached document. Figures 1 to 6 This invention provides a method for detecting the pull-out force of automotive wiring harness terminals. Its core lies in constructing a closed-loop feedback mechanism between the crimping production process and the pull-out force performance test. Through multi-source sensor data fusion, machine learning modeling, and constraint parameter optimization, adaptive iterative adjustment of the crimping process parameters is achieved. The following will describe this embodiment in detail, combining each of the S-steps in the technical solution.

[0033] The method first deploys a crimping process sensor array at the crimping station to collect real-time data on the applied pressure, crimping stroke, crimping speed, die temperature, and wire core compression ratio during the crimping process. This sensor array consists of a high-precision strain gauge pressure sensor, a laser displacement sensor, an encoder, a thermocouple temperature sensor, and an image recognition unit. The high-precision strain gauge pressure sensor is installed at the end of the piston rod of the crimping cylinder and is used to measure the instantaneous pressure acting on the terminal during the crimping process. Its measurement range covers 0 to 10 kN, with an accuracy class of 0.5% and a sampling period of 10 milliseconds.

[0034] A laser displacement sensor is aligned with the movement trajectory of the pressing die, recording the pressing stroke in real time with a resolution of 1 micrometer to ensure the accuracy of the stroke data. An encoder is coaxially connected to the pressing spindle, recording the pressing speed at a sampling frequency of 1000 Hz to reflect the dynamic characteristics of the pressing action. A thermocouple temperature sensor is embedded inside the pressing die, no more than 5 mm from the pressing contact surface, to monitor the temperature rise of the die during continuous operation.

[0035] The image recognition unit consists of an industrial camera and a ring light source. The industrial camera has a frame rate of 120 frames per second and a resolution of no less than 2 million pixels. It immediately captures a cross-sectional image of the terminal after each crimping. After the image is processed by the edge detection algorithm, the actual cross-sectional area of ​​the wire core after crimping is calculated, and the ratio of this ratio to the original cross-sectional area of ​​the wire core is used as the wire core compression ratio. The value range is usually between 60% and 85%.

[0036] After the crimping process is completed, the terminal samples are transferred to the pull-out force testing station. This station is equipped with a servo motor-driven tensile testing machine with a maximum output tensile force of 5000 N and a built-in force sensor with an accuracy class of 0.2%. The clamping mechanism has a self-centering function to ensure that the pull-out direction is strictly applied along the terminal axis, avoiding test errors caused by off-center loading. The pull-out force test must be performed within 30 minutes after crimping to prevent material stress relaxation from interfering with the test results. The test environment temperature is controlled at 25°C ± 2°C, and the relative humidity is maintained at 50% ± 5% to eliminate the influence of environmental fluctuations on the mechanical properties of the metal material. For each test, five samples are randomly selected from the same batch of continuously produced terminals, and the pull-out force is tested sequentially. The arithmetic mean of the results is taken as the actual pull-out force value for that batch, which is used for subsequent data binding and model training.

[0037] After acquiring the physical parameters of the crimping process and the corresponding actual pull-out force values, time alignment and data binding operations are performed to form a tagged process-performance sample set. This process includes assigning a unique serial number to each crimped workpiece. This serial number is generated by the central control system when crimping begins and synchronously written into the data streams of all sensors. During crimping, each sensor synchronously collects data at 10-millisecond intervals, and parameters such as pressure value, crimping stroke, crimping speed, die temperature, and wire core compression ratio are associated with this serial number according to a timestamp. When the sample enters the pull-out force testing station, the operating terminal reads its physical identifier or obtains the serial number through visual recognition, and writes the measured pull-out force value into the record field of the corresponding serial number in the central database. All data is indexed in multiple dimensions by timestamp, equipment number, wire harness model, and terminal specification, facilitating subsequent conditional retrieval and batch calling. The structure of this sample set is a mapping relationship between a 5-dimensional input feature vector and a 1-dimensional output label. The input features are, in order, pressure value, crimping stroke, crimping speed, die temperature, and wire core compression ratio, and the output label is the actual pull-out force value in Newtons.

[0038] Based on the aforementioned process-performance sample set, a pull-out force prediction regression model was constructed. This model employs a multilayer perceptron neural network structure, comprising an input layer, two hidden layers, and an output layer. The input layer has five nodes, each receiving one of the five physical parameters of the aforementioned pressing process. Each hidden layer contains 30 neurons, with the activation function uniformly using a modified linear unit to enhance the model's ability to fit nonlinear relationships. The output layer is a single node, directly outputting the predicted pull-out force value. The model training uses mean squared error as the loss function, and the optimizer employs an adaptive moment estimation algorithm. The initial learning rate is set to 0.001, and the batch size is 64. Training is performed on edge computing units, with incremental training triggered every 500 newly added valid samples. This involves fine-tuning the model using new samples while retaining the original model weights to adapt to distribution shifts caused by material batch variations, mold wear, or environmental fluctuations. The model convergence criterion is a decrease in mean squared error on the validation set of less than 0.1% for three consecutive training cycles.

[0039] Before the current batch of crimping operations begins, a trained pull-out force prediction regression model is used to predict the pull-out force of the currently set combination of crimping parameters. If the predicted value falls within the preset acceptable range, the existing parameters remain unchanged; if the predicted value is lower than the lower limit of the acceptable range or higher than the upper limit, the parameter optimization module is activated. The preset acceptable range is pre-set according to the terminal specifications, with its lower limit not lower than the minimum pull-out force value specified by industry standards. For example, for the combination of AWG 16 gauge wire and standard yellow terminal, the minimum pull-out force is 350 Newtons; the upper limit does not exceed 90% of the theoretical maximum pull-out force value corresponding to the yield strength of the terminal material, to avoid excessive crimping leading to terminal cracking or conductor damage. The target pull-out force median is defined as the arithmetic mean of this acceptable range, serving as the target reference point for parameter optimization.

[0040] The parameter optimization module employs a gradient descent algorithm, with the objective function being to minimize the squared error between the predicted pull-out force and the median target pull-out force. Let the current pressing parameter vector be... The predicted pull-off force is The target pull-off force median value is Then the objective function is:

[0041]

[0042] The parameter update formula is:

[0043]

[0044] in The learning rate is 0.01. The gradient of the loss function with respect to the parameter vector is calculated using the backpropagation algorithm. Gradient descent iterations are executed until any of the following termination conditions are met: the predicted pull-off force value enters the acceptable range; the number of iterations reaches 10; or the parameter change in two consecutive iterations is less than a preset threshold.

[0045] During gradient descent iterations, multiple physical constraints are applied to ensure the safety and feasibility of parameter adjustments. These constraints include: the crimping pressure must not be lower than the equipment's minimum safe pressure threshold, determined by the minimum stable operating pressure of the pneumatic system, typically 500 Newtons; and it must not exceed the pressure limit corresponding to the fatigue limit of the mold material, determined through mold life testing, generally not exceeding 8000 Newtons. The crimping stroke must not be less than the minimum stroke required for the wire core to be fully embedded in the terminal, a value determined by the terminal structure depth and wire core diameter, typically 1.2 mm; nor must it exceed the mold's closing limit position to prevent mechanical overload, typically set at 2.5 mm. Furthermore, each parameter adjustment is limited to no more than 5% of the current value, i.e.:

[0046]

[0047] This limitation prevents system oscillations or the production of defective products due to sudden parameter changes.

[0048] After calculating the new crimping pressure and stroke values, the parameter optimization module sends this parameter combination to the servo control system of the crimping equipment. Upon receiving the command, the servo control system updates the opening setting of the pressure regulating valve and the target stroke value of the servo motor, and executes the new parameters in the next round of crimping operations. To verify the effectiveness of the parameter update, a full range of tests are performed on the first product after the parameter switch, including pull-out force testing, cross-sectional microscopic observation, and continuity resistance measurement.

[0049] Pull-out force testing is performed according to the aforementioned standard procedure; cross-sectional microscopic observation is conducted using a metallographic microscope at 200x magnification to check for cracks, voids, or wire breakage in the terminal crimping area; continuity resistance is measured using a four-wire micro-ohmmeter to ensure that the contact resistance after crimping is less than 10 milliohms. If all three test results meet the quality specifications, the parameter update is confirmed to be valid, and the new parameters are stored in the process parameter library as the benchmark parameters for subsequent production; if any test item fails to meet the requirements, it is immediately rolled back to the previous valid parameter set, and a manual review process is triggered, with a process engineer intervening to analyze the cause of the anomaly.

[0050] Throughout the process, data flows through multiple levels of interaction between the crimping equipment, sensing system, pull-out force testing machine, and central database. The crimping equipment uploads execution parameters and status signals in real time; the sensing system continuously pushes high-frequency physical quantity data; the pull-out force testing machine sends back performance labels after completing the test; and the central database, acting as a data hub, is responsible for storing, indexing, and distributing all information, and providing data services for model training and parameter optimization. Edge computing units are deployed on the production line side, handling model inference and lightweight optimization tasks, reducing reliance on the central server and improving response speed.

[0051] This method achieves a complete closed loop from data acquisition, model building, prediction and evaluation to adaptive parameter adjustment through the aforementioned steps. Each pressing operation becomes a new sample for model learning, and each parameter adjustment is based on quantified error and physical constraints, thereby ensuring that the pull-out force index remains stable within the qualified range over the long term. Compared to the traditional method that relies on manual experience to adjust parameters, this method increases the pull-out force qualification rate from 85% to over 99%, significantly reducing rework and scrap, while also reducing reliance on highly skilled operators and improving the automation and intelligence level of the production line.

[0052] In summary, the automotive wiring harness terminal pull-out force detection method described in this embodiment solves the problems of disconnect between testing and production and lag in parameter adjustment in traditional crimping processes by integrating multi-source sensing, constructing a data closed loop, and applying machine learning models and constraint optimization algorithms. This achieves consistent control of crimping quality and synergistic improvement of production efficiency.

Claims

1. A method of detecting a terminal pull-out force of an automobile wire harness, characterized by, include: A sensor array is deployed at the crimping station to collect data in real time on the applied pressure, crimping stroke, crimping speed, die temperature, and wire core compression ratio during the crimping process. Obtain the actual pull-out force value of the crimped terminal sample at the pull-out force testing station; The physical parameters of the pressing process are time-aligned and data-bound with the corresponding actual pull-out force values ​​to form a tagged process-performance sample set; Based on the process-performance sample set, a pull-out force prediction regression model is constructed. This model takes the physical parameters of the pressing process as input and the predicted pull-out force value as output. The pull-out force prediction regression model is used to predict the pull-out force of the current batch of crimping parameter combinations. If the predicted value deviates from the preset qualified range, the parameter optimization module is activated. The parameter optimization module uses the gradient descent algorithm to iteratively calculate the new pressing pressure value and pressing stroke value by taking the objective function as minimizing the square error between the predicted pull-out force value and the median target pull-out force value. The new crimping pressure and crimping stroke values ​​are sent to the servo control system of the crimping equipment to update the parameters for the next round of crimping operations.

2. The automobile wire harness terminal pull-out force detection method according to claim 1, characterized by A sensor array is deployed at the crimping station to collect data in real time on the applied pressure, crimping stroke, crimping speed, die temperature, and wire core compression ratio during the crimping process, including: The instantaneous pressure applied to the terminals during the crimping process is measured using a high-precision strain gauge pressure sensor. The crimping stroke is measured in real time using a laser displacement sensor; The crimping speed is recorded using an encoder; The working temperature of the mold is monitored using a thermocouple temperature sensor; The terminal cross-section image is captured by an image recognition unit, and the core compression ratio is calculated based on an edge detection algorithm.

3. The automobile wire harness terminal pull-out force detection method according to claim 2, characterized by Obtain the actual pull-out force value of the crimped terminal sample at the pull-out force testing station, including: A tensile testing machine driven by a servo motor is used to perform pull-out force tests; The clamping mechanism has a self-centering function to ensure that the pull-out direction coincides with the terminal axis; The pull-out force test should be completed within 30 minutes after the crimping is completed; Five samples from the same batch of consecutive production were selected, and the average pull-out force was taken as the actual pull-out force value for that batch.

4. The automobile wire harness terminal pull-out force detection method according to claim 3, characterized by The physical parameters of the pressing process are time-aligned and data-bound with the corresponding actual pull-out force values ​​to form a tagged process-performance sample set, including: Assign a unique serial number to each crimped workpiece; Data from each sensor is recorded synchronously with a sampling period of 10 milliseconds and associated with the serial number. During the pull-out force test, the sample with the corresponding serial number is read, and the measured pull-out force value is written into the data record of that serial number; All data is indexed and stored in a central database by timestamp, device number, harness model, and terminal specification.

5. The automobile wire harness terminal pull-out force detection method according to claim 4, characterized by Based on the aforementioned process-performance sample set, a pull-out force prediction regression model is constructed, including: It adopts a multilayer perceptron neural network structure, which includes an input layer, two hidden layers and an output layer; The number of input layer nodes is 5, corresponding to pressure value, crimping stroke, crimping speed, die temperature and wire core compression ratio respectively; Each hidden layer contains 30 neurons, and the activation function is a modified linear unit; The output layer is a single node, outputting the predicted pull-out force value; The mean squared error loss function and adaptive moment estimation algorithm are used for model training. Incremental training is triggered after every 500 new valid samples are accumulated.

6. The automobile wire harness terminal pull-out force detection method according to claim 5, characterized by The preset qualified range is set in advance according to the terminal specifications. Its lower limit is not lower than the minimum pull-out force value specified by the industry standard, and its upper limit is not more than 90% of the theoretical maximum pull-out force value corresponding to the yield strength of the terminal material. The median value of the target pull-out force is the arithmetic mean of the qualified range.

7. The method for detecting the pull-out force of automotive wiring harness terminals according to claim 6, characterized in that, When the parameter optimization module performs gradient descent iterations, the constraints include: The pressing pressure value shall not be lower than the minimum safe pressure threshold of the equipment, and shall not be higher than the pressure limit corresponding to the fatigue limit of the mold material; The crimping stroke must not be less than the minimum stroke required for the wire core to be fully embedded in the terminal, and must not be greater than the limit position of the die closure. Each parameter adjustment should not exceed 5% of the current value.

8. The method for detecting the pull-out force of automotive wiring harness terminals according to claim 7, characterized in that, After sending the new crimping pressure and crimping stroke values ​​to the servo control system of the crimping equipment and updating the parameters for the next round of crimping operations, the system also includes: The first product undergoes full testing, including pull-out force testing, cross-sectional microscopic observation, and continuity resistance measurement. If all test results meet the quality specifications, the parameter update is confirmed to be valid, and the new parameters will be used as the benchmark for subsequent production. If it does not meet the requirements, it will roll back to the previous valid parameter set and trigger a manual review process.

9. The method for detecting the pull-out force of automotive wiring harness terminals according to claim 8, characterized in that, The cross-sectional microscopic observation was performed using a metallographic microscope at 200x magnification to check for cracks, voids, or broken wires in the terminal crimping area; the conduction resistance was measured using a four-wire micro-ohmmeter.

10. The method for detecting the pull-out force of automotive wiring harness terminals according to claim 1, characterized in that, The pull-out force prediction regression model is deployed on an edge computing unit to perform model inference and lightweight parameter optimization on the production line side, reducing reliance on a central server.

Images