Intelligent hierarchical control method for wire harness terminal crimping process

Abstract

The application provides an intelligent grading control method for a wire harness terminal crimping process, comprising the following steps: acquiring multi-source real-time detection data in a terminal processing process; based on a preset process standard database, performing feature extraction and health degree evaluation on the multi-source real-time detection data to generate corresponding processing quality grades; wherein the processing quality grades at least include a first grade, a second grade and a third grade. The application significantly reduces the waste rate, improves the quality consistency, reduces manual intervention, realizes intelligent, self-adaptive and self-optimizing control of the terminal processing process, and has outstanding substantial features and significant progress.

Description

Technical Field

[0001] This invention relates to an intelligent hierarchical control method for the crimping process of wire harness terminals. Background Technology

[0002] Electronic wire harnesses, as key components for signal transmission and power delivery in electronic devices, are widely used in automotive electronics, aerospace, and home appliances. Terminal crimping is a core process in electronic wire harness production, and its quality directly affects the electrical connection reliability and mechanical strength of the entire wire harness. Traditional terminal crimping mainly relies on manual experience to set fixed process parameters, such as crimping force and crimping height, and then mass production is carried out through open-loop control. With the development of sensor technology, terminal crimping equipment has begun to be equipped with pressure sensors or displacement sensors, and quality monitoring is carried out by setting upper and lower threshold values. When the detected value exceeds the threshold range, the equipment automatically alarms or stops. However, single-parameter threshold control has inherent limitations: normal pressure does not mean that the insulation is not damaged, displacement meets the standard, but core wire is not fully deformed, and a single perspective cannot cover all defect types.

[0003] To address these issues, some high-end equipment incorporates machine vision systems to photograph and inspect the terminals after crimping, identifying visual defects. However, this method is an offline or near-line inspection, unable to intervene in the crimping process in real time, and powerless to assess internal crimping quality (such as core wire deformation).

[0004] By analyzing existing technical literature and commercially available products, the applicant found that: (1) Both threshold alarms and visual inspections are binary judgment modes of "qualified / unqualified", lacking early warning and fine-tuning mechanisms for intermediate states, resulting in a large number of critical defective products flowing into the next process or eventually becoming potential failure points; (2) Sensor data such as pressure, displacement, vision, and vibration are processed independently, failing to establish a correlation model between data, and making it impossible to assess the health of the crimping process from a system level; (3) Existing controls are all feedback controls, that is, adjustments are made after a problem occurs, failing to optimize process parameters in advance according to the characteristics of incoming materials, and having a weak ability to cope with fluctuations in incoming materials. Therefore, how to achieve intelligent, adaptive, and self-optimizing control of the electronic wire harness terminal processing process has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] This invention provides an intelligent hierarchical control method for the wire harness terminal crimping process, which can effectively solve the above-mentioned problems.

[0006] This invention is implemented as follows: A smart hierarchical control method for wire harness terminal crimping process includes the following steps: S1. Acquire multi-source real-time detection data during the terminal processing process; S2. Based on a preset process standard database, feature extraction and health assessment are performed on the multi-source real-time detection data to generate corresponding processing quality levels; wherein, the processing quality levels include at least a first level, a second level, and a third level; S3. Based on the processing quality level, match the corresponding control strategy in the pre-stored hierarchical control strategy library and generate control instructions; S4. Send the control command to the terminal processing actuator to dynamically adjust the subsequent terminal processing process.

[0007] The beneficial effects of this invention are: (1) This invention provides an intelligent hierarchical control method for wire harness terminal crimping process. Compared with the prior art, it has the following advantages: By setting a first threshold range and a second threshold range, the processing quality is divided into four levels. For the first time, the critical state of "deviation from the optimal but not exceeding the standard" is identified as the second level, creating a window for early warning and intervention before the production of defective products, realizing a fundamental shift from "post-processing" to "pre-processing". At the same time, it integrates multi-source data of pressure, displacement, vision and vibration, extracts three types of feature values ​​of waveform, geometry and spectrum for comprehensive evaluation, and overcomes the one-sidedness of single sensor judgment. For the critical state of the second level, a fuzzy neural network controller is introduced to calculate the optimal compensation value online, realizing adaptive and precise fine-tuning of nonlinear and time-varying crimping process. By cross-station feedforward control, it adapts to the characteristics of incoming materials in advance, reducing quality problems caused by wire diameter fluctuations. It constructs a self-learning optimization closed loop, automatically updates the threshold range and parameter mapping table, so that the system can adapt to equipment aging and environmental changes, and achieve continuous evolution. In addition, the data cliff change is separately divided into the fourth level, distinguishing between process problems and system faults, and avoiding incorrect adjustments. Through the synergistic effect of the above-mentioned technical means, the present invention significantly reduces the scrap rate, improves quality consistency, reduces manual intervention, and realizes intelligent, adaptive, and self-optimizing control of the terminal processing process, which has outstanding substantive features and significant progress. Attached Figure Description

[0008] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0009] Figure 1 This is a flowchart of the quality grade determination logic of the present invention.

[0010] Figure 2 This is a diagram of the architecture of the intelligent hierarchical control system for electronic wire harness terminals of the present invention. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention.

[0012] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0013] Reference Figure 1-2 As shown, an intelligent hierarchical control method for the wire harness terminal crimping process includes the following steps: S1. Acquire multi-source real-time detection data during the terminal processing process; S2. Based on a pre-set process standard database, feature extraction and health assessment are performed on multi-source real-time detection data to generate corresponding processing quality levels; wherein, the processing quality levels include at least the first level, the second level, and the third level; S3. Based on the processing quality level, match the corresponding control strategy in the pre-stored hierarchical control strategy library and generate control instructions; S4. Send control commands to the terminal processing actuator to dynamically adjust the subsequent terminal processing process.

[0014] It should be noted that in step S2, this application divides the quality status into four levels. This is because although the crimping process is continuously changing, control decisions must be discrete. The system cannot do different things at every moment. Therefore, it is necessary to map the continuous quality changes to a finite number of levels, with each level corresponding to a type of state that requires the same handling method. Through the analysis of a large amount of crimping process data, the applicant found three key dividing points in the changes in crimping quality, thus dividing it into four levels, as follows: The first dividing line: the boundary between ideal and acceptable states: when all parameters are within a narrow ideal range, the crimping quality is optimal and stable, requiring no intervention. This range is defined as the first threshold range. States within this range are defined as the first level (excellent state). The second dividing line: the boundary between acceptable and unacceptable states: when parameters exceed the ideal range but remain within certain limits, the product, though imperfect, is still usable. This limit is the bottom line of quality requirements, defined as the second threshold range. Exceeding this bottom line renders the product unacceptable. It should be noted that products within the range between the first and second thresholds are neither optimal (exceeding the first threshold) nor unacceptable (not exceeding the second threshold). According to binary judgment, they would be categorized as "acceptable" and allowed to pass. However, research has found that this area is precisely the most dangerous because quality is deteriorating; without intervention, it will enter the unacceptable zone. Therefore, this case separates this area as the second level (critical state). The third dividing line: the boundary between routine changes and abrupt changes: In addition to the gradual changes caused by process drift mentioned above, there is a special type of change, such as abrupt, precipitous changes. Examples include pressure dropping to zero instantaneously, displacement data suddenly interrupting, and image signals being completely lost. These changes are not caused by the process itself, but by system-level anomalies such as sensor malfunctions, communication interruptions, and power failures. Their mathematical characteristics are completely different: it's not "exceeding a threshold," but rather a "fundamental change in data characteristics." Therefore, this type of state cannot be classified into the above three levels and must be handled separately, defined as the fourth level (abnormal state).

[0015] Thus, this invention has constructed a health assessment model comprising four levels: First level (excellent condition): Within the first threshold range, the quality is optimal; Level 2 (Critical State): The quality is deteriorating, exceeding the first threshold but within the second threshold. Level 3 (Unqualified): Exceeds the second threshold, quality is no longer qualified; Level 4 (Abnormal State): Sudden changes in data characteristics, system-level anomalies.

[0016] Based on the above model, this invention provides a specific determination method that is calculable and executable. First, three types of feature values ​​are extracted from multi-source real-time detection data: waveform feature values, including the peak value, valley value, rising slope, and integral area of ​​the crimping force curve; geometric feature values, including crimping height, crimping width, core wire visibility, and insulation position; and spectral feature values, including the dominant frequency amplitude and harmonic components. Second, a first threshold range and a second threshold range are preset in the process standard database. The first threshold range represents the ideal range with optimal quality, and the second threshold range represents the acceptable baseline range with acceptable quality. The first threshold range is narrower than the second threshold range and is completely contained within it. Finally, the grade determination is performed according to the following logic: The first-level determination: The first level requires all feature values ​​to be in their ideal state. Let the set of collected feature values ​​be F, and the first threshold range be [T1min, T1max]. Then the determination condition for the first level is: for all feature values ​​f belonging to F, f is between T1min and T1max. That is, all feature values ​​are within the first threshold interval.

[0017] The determination of the second level: The characteristic of the second level is that at least one feature value has deviated from the ideal state, but has not yet exceeded the quality baseline. Let the range of the second threshold be [T2min, T2max], and T2min is less than T1min, and T1max is less than T2max. Then the determination condition for the second level is: there exists at least one feature value f that satisfies f less than T1min or f greater than T1max, and all feature values ​​f are between T2min and T2max. That is, at least one feature value exceeds the first threshold, but all feature values ​​are still within the second threshold.

[0018] The determination of the third level: The third level is marked by the breach of the quality baseline. The determination condition is that there exists at least one feature value f that satisfies either f less than T2min or f greater than T2max. That is, at least one feature value exceeds the range of the second threshold.

[0019] Level 4 Judgment: Level 4 corresponds to anomalies in the data itself, rather than deviations in values, requiring separate detection logic. Define a data quality detection function, and classify data as abnormal when it meets any of the following conditions: the rate of change between adjacent sampling points exceeds the physical limit; the data value exceeds the sensor's range; multiple consecutive sampling points are missing or have invalid values. The judgment condition for Level 4 is: at least one feature value is judged as abnormal.

[0020] To further clarify the innovative aspects of this invention, the technical reasons why the fourth level must be handled separately are explained in detail below: First, the nature of the problems differs. The third level is a process problem: parameter deviations are caused by the pressing process itself, such as mold wear, wire diameter fluctuations, or improper parameter settings. These problems can be resolved by adjusting process parameters. The fourth level is a system problem: data anomalies are caused by faults in basic systems such as sensors, communication, and power supplies. These problems cannot be resolved through process adjustments and require equipment maintenance or system reset. If the two are confused, when a sensor fails, the system may incorrectly adjust the pressing parameters, not only failing to solve the problem but also potentially masking the true fault and causing greater losses. Second, the characteristics of the data differ. The characteristic of the third level is that the values ​​are still within the physically possible range, but deviate from the target value. For example, the pressing force gradually increases from 1000N to 1050N, which is a reasonable value. The characteristic of the fourth level is that the values ​​exhibit physically impossible changes, such as the pressing force instantly dropping from 1000N to 0. This is physically impossible (the pressing force does not disappear instantly) and only indicates sensor failure. Third, the handling methods differ. The third level requires process adjustments: compensating for deviations by fine-tuning parameters such as pressing height and pressure. Level 4 requires system recovery: checking sensor connections, restarting the data acquisition module, replacing faulty components, etc.; fourth, the urgency of the response differs. Level 3 degradation is gradual, allowing time for system fine-tuning. Level 4 abrupt changes are instantaneous, potentially generating a large number of defective products every second, requiring immediate production halt and rapid isolation of abnormal products.

[0021] The above-described grading system achieves the following technical effects, as shown in the chart below:

[0022] In step S1, the multi-source real-time detection data includes at least two of the following: crimping force data collected by a pressure sensor, crimping displacement data collected by a displacement sensor, terminal appearance image data collected by a vision sensor, and feeding vibration data collected by a vibration sensor.

[0023] Feature extraction and health assessment are performed on multi-source real-time detection data to generate corresponding processing quality levels. Specifically, this includes: extracting waveform feature values, geometric feature values, and spectral feature values ​​from the multi-source real-time detection data; comparing the extracted waveform feature values, geometric feature values, and spectral feature values ​​with the standard feature threshold ranges in the process standard database; if all feature values ​​are within the first threshold range, the system is classified as Level 1; if at least one feature value exceeds the first threshold range but is within the second threshold range, the system is classified as Level 2; if at least one feature value exceeds the second threshold range, the system is classified as Level 3. Specifically, this invention transforms raw data collected by multiple sensors, including those for pressure, displacement, vision, and vibration, into waveform feature values ​​(such as peak value and slope), geometric feature values ​​(such as pressing height and width), and spectral feature values ​​(such as main frequency and harmonics) through feature extraction and health assessment steps. This achieves data dimensionality reduction and information condensation, enabling the processing of originally heterogeneous mechanical, geometric, and dynamic data within a unified framework. Subsequently, these feature values ​​are compared with a preset first threshold range (representing the optimal state) and a second threshold range (representing the quality baseline) in the process standard database. Based on the comparison results, the processing quality is divided into two levels: a first level (all feature values ​​are within the optimal range) and a second level (at least one feature value is within the optimal range). The invention categorizes products into two levels: a first level (characteristic value deviating from the optimal but not exceeding the minimum standard) and a second level (at least one characteristic value exceeding the minimum quality standard). The identification of the second level is a key innovation of this invention. It is the first time that the critical state of "deviation but not exceeding the standard" has been separated from the traditional qualified products, creating a window for early warning and intervention before the production of unqualified products. This step plays a crucial role in connecting the preceding and following steps: it not only receives the raw data from multiple sensor sources, but also provides accurate decision-making basis for subsequent hierarchical control strategies, enabling the system to take differentiated control actions for different levels. At the same time, it accumulates valuable sample data for self-learning optimization, solving the defects of existing technologies that only have binary judgment, lose intermediate state information, and cannot achieve preventive control.

[0024] The method also includes: if the key feature values ​​in the multi-source real-time detection data show a cliff-like change or are missing, it will be directly judged as the fourth level and a waste rejection instruction will be triggered.

[0025] Specifically, a "cliff-like change" refers to a significant jump in a feature value between adjacent sampling points that exceeds the physically possible range. It can be quantified and detected in the following ways: Let fi(t) be the value of the i-th feature at sampling time t, and let fi(t) be the value of the i-th feature at the adjacent previous sampling time t. The value of 1 is fi(t) 1): The absolute value of the change in this feature value within a single sampling period is defined as:

[0026] Let Δ maxi This is the maximum possible rate of change of this characteristic value in physical terms, which can be obtained through statistical analysis of the device's physical characteristics and historical data.

[0027] If the following conditions are met:

[0028] This indicates that a "cliff-like change" has occurred at the current moment. Here, K is a preset threshold coefficient, which is usually set to 3 to 5.

[0029] To further illustrate, refer to the following examples: The crimping force drops from 1000N to 0N within 10ms: This is physically impossible (the crimping force does not disappear instantaneously), and Δmax usually does not exceed 200N / 10ms. Here, the change of 1000N is much greater than the threshold, and it is judged as a cliff-like change; The displacement sensor signal suddenly changes from 5.0mm to -100mm: This exceeds the sensor's range and is obviously abnormal; The visual image signal is completely lost: Empty data is received for several consecutive cycles.

[0030] Furthermore, the essential differences between the fourth level and the first three levels are shown in the chart below:

[0031] Based on the processing quality level, the corresponding control strategy is matched from the pre-stored hierarchical control strategy library, specifically including: If the current processing quality level is the first level, then the first control strategy is matched. The first control strategy is to maintain the current process parameters and record the current processing data to the excellent process sample library. Specifically, the first control strategy is applicable to the processing process with a health assessment of the first level (excellent state). At this time, all characteristic values ​​are within the first threshold range, indicating that the current pressing process is in the optimal stable state and the product quality is reliable. "Maintaining the current process parameters" includes the following specific operations: (1) No adjustment command is sent: the system does not send any parameter adjustment command to the processing actuator; (2) Keep the original set value: all process parameters (such as pressing force set value, pressing height set value, feeding speed, etc.) keep the current value unchanged; (3) Control cycle is skipped: no parameter update calculation is performed within the current control cycle.

[0032] In summary, when the system determines the first level, all eigenvalue are within the optimal range, and the production process is in an ideal state. At this time, the strategy of "maintaining the current process parameters unchanged" is adopted, which avoids unnecessary parameter fluctuations and enables the excellent state to be continuously maintained. This "non-interference" itself is the optimal intervention, ensuring the continuity and stability of the production process, and recording the data of the first level in the excellent process sample library, which is equivalent to establishing a quantitative standard for the system. These positive samples have multiple uses: one is as a benchmark for subsequent threshold optimization, making the setting of the first threshold range more suitable for actual production; the second is as training data for the fuzzy neural network controller, enabling it to learn "the characteristics of the excellent state"; the third is as the basis for quality traceability, which can be traced and compared when quality problems occur. Moreover, the more data of the first level are accumulated, the richer the excellent process sample library will be, and the more accurate the subsequent threshold optimization and model training will be. The optimization of the threshold and the model will in turn enable more products to enter the first level, forming a virtuous cycle of "more stable → more data → more accurate model → more stable".

[0033] If the current processing quality level is the second level, the second control strategy is matched. The second control strategy is to calculate the deviation between the current eigenvalue and the standard threshold, and generate fine-tuning control parameters according to the deviation to perform online compensation for the process parameters of the processing actuator. When deviations occur simultaneously in multiple eigenvalues, by analyzing the combination pattern of the deviation amounts of each eigenvalue, the root cause of the deviation can be identified, such as distinguishing wire diameter fluctuations, die wear or equipment looseness. Specifically, the second control strategy is applicable to the processing process with a health assessment of the second level (critical state). At this time, at least one eigenvalue has exceeded the first threshold range but is still within the second threshold range, indicating that the quality is deteriorating but not yet unqualified, and it is in a critical window period that can be repaired. The calculation of the deviation amount adopts different methods according to the eigenvalue type; For eigenvalues with a clear target value, let the first threshold range of eigenvalue f be [L1, U1]. Usually, the midpoint of the interval is taken as the ideal target value, and the calculation formula is:

[0034] Then the calculation formula for the absolute deviation amount e is:

[0035] The relative deviation amount e rel The calculation formula is: .

[0036] For eigenvalues with only upper and lower limit requirements, if the eigenvalue only requires to fall within the interval and there is no clear ideal target value, the degree of deviation can be calculated: When f < L1 (too small): e = (L1 - f) / L1 × 100%; When f > U1 (too large): e = (f-U1) / U1×100%.

[0037] In the above formula, f represents the currently acquired feature value; L1 represents the lower limit of the first threshold range; U1 represents the upper limit of the first threshold range; t represents the ideal target value, taken as the midpoint of the first threshold range; e represents the absolute deviation, reflecting the degree of absolute deviation between the feature value and the ideal target value; e rel It represents the relative deviation, reflecting the degree of deviation between the characteristic value and the ideal target value, and is expressed as a percentage.

[0038] In summary, the second control strategy intervenes when quality just begins to deteriorate but before defective products are produced, using online fine-tuning to bring the condition back to the excellent range. Compared to the reactive processing of existing technologies, the second control strategy achieves proactive prevention, eliminating potential problems before defective products are generated and reducing the scrap rate at the source. For example, consider a specific scenario: the crimping force increases continuously for five cycles, gradually rising from 1000N to 1040N (the first threshold is 1020N). Existing technologies only trigger an alarm after the limit is exceeded, by which time defective products have already been generated; however, the second control strategy of this invention determines the second level in the third cycle (1030N) and initiates fine-tuning, appropriately increasing the crimping height to bring subsequent products back down to 1010N, thus preventing the generation of defective products.

[0039] Specifically, the generation of fine-tuning control parameters is achieved through a fuzzy neural network controller. This controller takes the deviation and historical adjustment effects as input and outputs the optimal compensation value to the machining actuator. The second control strategy, also implemented through the fuzzy neural network controller, achieves a precise mapping between the deviation and the compensation value. The fuzzy neural network can handle complex nonlinear relationships during the crimping process, outputting targeted compensation values ​​for deviations caused by different reasons: excessive crimping force due to larger wire diameter → compensate for crimping height; insufficient crimping force due to mold wear → compensate for the crimping force setpoint; abnormal spectrum caused by feeding vibration → compensate for feeding speed. This precise compensation avoids the over-adjustment or under-adjustment problems that may result from a "one-size-fits-all" approach.

[0040] If the current processing quality level is level three, a third control strategy is applied. This strategy generates an alarm signal and controls the sorting mechanism to divert the currently processed products to the inspection area. Specifically, the third control strategy automatically pushes non-conforming products to the inspection area via the diversion mechanism, physically eliminating the possibility of non-conforming products mixing into the packaging of conforming products. This is the bottom-line guarantee of quality control, ensuring that 100% of the products flowing to downstream processes or customers are conforming. Simultaneously, the alarm signal not only provides audible and visual alerts on-site but also records detailed information about non-conforming products (time, characteristic values, deviation, product image, etc.) in the system database. This information forms a complete quality traceability chain, allowing for precise backtracking to the production time and specific parameters of each non-conforming product when analyzing the causes of quality problems.

[0041] Furthermore, the fuzzy neural network controller employs a five-layer structure: the first layer is the input layer, receiving the deviation and historical adjustment effect features; the second layer is the fuzzification layer, using a Gaussian membership function to fuzzify the input; the third layer is the rule layer, where each node represents a fuzzy rule, and a multiplication operator is used to calculate the rule trigger strength; the fourth layer is the normalization layer; and the fifth layer is the output layer, which outputs the optimal compensation value using a weighted summation method. The initial parameters of the network can be set through expert experience, and subsequent training and updates are performed using data accumulated through self-learning optimization steps.

[0042] The method also includes a feedforward control step: acquiring the incoming wire harness characteristic data transmitted from the previous station, which includes at least the measured wire diameter and conductor flexibility coefficient; using the incoming wire harness characteristic data as feedforward input to pre-adjust the reference process parameters of the current station; the specific calculation model for feedforward control is as follows: Pre-adjustment formula for crimping height:

[0043] Pre-adjustment formula for clamping force:

[0044] Among them, H adj Indicates the pre-adjusted crimping height; H base Indicates the reference crimping height; k d D represents the wire diameter compensation coefficient, used to quantify the impact of wire diameter deviation on the crimping height. 实测 Indicates the measured wire diameter; D 标准 Indicates the standard wire diameter value; F adj Indicates the pre-adjusted pressing force; F base Indicates the reference pressing force; k f S represents the softness compensation coefficient, used to quantify the degree of influence of softness deviation on the pressing force; 实测 This indicates the measured value of softness; S 标准This indicates the standard softness coefficient.

[0045] The above wire diameter compensation coefficient k d And softness compensation coefficient k f It can be determined through experimental calibration or self-learning optimization.

[0046] The method also includes a self-learning optimization step: S5. Record the adjustment amount and corresponding processing effect data for each graded control process; S6. After the batch processing is completed, call the optimization algorithm to perform correlation analysis on the adjustment amount and processing effect data; S7. Update the standard feature threshold range in the process standard database and / or the control parameter mapping table in the hierarchical control strategy library based on the analysis results.

[0047] In step S7, the first and second threshold ranges in the process standard database are updated using an exponential moving average. The threshold update uses an exponential moving average method, and the calculation formula is as follows:

[0048] Among them, T new The updated threshold is represented by α; the learning rate is represented by α, which ranges from 0.1 to 0.3; T opt T represents the optimal threshold obtained from this batch of analysis; old This represents the original threshold before the update.

[0049] Similarly, the control parameter mapping table (including the weights of the fuzzy neural network and the feedforward compensation coefficients) in the hierarchical control strategy library is updated using the same exponential moving average method.

[0050] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements an intelligent hierarchical control method for a wire harness terminal crimping process.

[0051] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the invention by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the invention should be included within the scope of protection of the invention.

Claims

1. A smart hierarchical control method for the wire harness terminal crimping process, characterized in that, Includes the following steps: S1. Acquire multi-source real-time detection data during the terminal processing process; S2. Based on a preset process standard database, feature extraction and health assessment are performed on the multi-source real-time detection data to generate corresponding processing quality levels; wherein, the processing quality levels include at least a first level, a second level, and a third level; S3. Based on the processing quality level, match the corresponding control strategy in the pre-stored hierarchical control strategy library and generate control instructions; S4. Send the control command to the terminal processing actuator to dynamically adjust the subsequent terminal processing process.

2. The intelligent hierarchical control method for wire harness terminal crimping process according to claim 1, characterized in that, The multi-source real-time detection data includes at least two of the following: crimping force data collected by a pressure sensor, crimping displacement data collected by a displacement sensor, terminal appearance image data collected by a vision sensor, and feeding vibration data collected by a vibration sensor.

3. The intelligent hierarchical control method for wire harness terminal crimping process according to claim 1, characterized in that, Feature extraction and health assessment are performed on the multi-source real-time detection data to generate corresponding processing quality levels. Specifically, this includes: extracting waveform feature values, geometric feature values, and spectral feature values ​​from the multi-source real-time detection data; comparing the extracted waveform feature values, geometric feature values, and spectral feature values ​​with the standard feature threshold ranges in the process standard database; if all feature values ​​are within the first threshold range, the system is classified as a first level; if at least one feature value exceeds the first threshold range but is within the second threshold range, the system is classified as a second level; if at least one feature value exceeds the second threshold range, the system is classified as a third level.

4. The intelligent hierarchical control method for wire harness terminal crimping process according to claim 3, characterized in that, The method further includes: If the key feature values ​​in the multi-source real-time detection data show a sharp change or are missing, it will be directly determined as Level 4 and a scrap rejection instruction will be triggered.

5. The intelligent hierarchical control method for wire harness terminal crimping process according to claim 1, characterized in that, Based on the processing quality level, a corresponding control strategy is matched from a pre-stored hierarchical control strategy library, specifically including: If the current processing quality level is the first level, then the first control strategy is matched, which is to maintain the current process parameters and record the current processing data to the excellent process sample library. If the current processing quality level is the second level, then the second control strategy is matched. The second control strategy is to calculate the deviation between the current feature value and the standard threshold, and generate fine-tuning control parameters based on the deviation to compensate the process parameters of the processing actuator online. If the current processing quality level is level three, then a third control strategy is matched. The third control strategy is to generate an alarm signal and control the sorting mechanism to divert the currently processed products to the inspection area.

6. The intelligent hierarchical control method for wire harness terminal crimping process according to claim 5, characterized in that, The fine-tuning control parameters are generated by a fuzzy neural network controller; the fuzzy neural network controller takes the deviation and historical adjustment effects as input and outputs the optimal compensation value to the processing execution mechanism.

7. The intelligent hierarchical control method for wire harness terminal crimping process according to claim 1, characterized in that, The method further includes a feedforward control step: Acquire the wire harness incoming material characteristic data transmitted from the previous workstation. The wire harness incoming material characteristic data includes at least the measured wire diameter and the conductor flexibility coefficient. The incoming wire harness feature data is used as a feedforward input to pre-adjust the baseline process parameters of the current station, so that the processing actuator can adapt to the characteristics of the incoming material in advance.

8. The intelligent hierarchical control method for wire harness terminal crimping process according to claim 1, characterized in that, The method also includes a self-learning optimization step: S5. Record the adjustment amount and corresponding processing effect data for each graded control process; S6. After the batch processing is completed, call the optimization algorithm to perform correlation analysis on the adjustment amount and processing effect data; S7. Update the standard feature threshold range in the process standard database and / or the control parameter mapping table in the hierarchical control strategy library based on the analysis results.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the intelligent hierarchical control method as described in any one of claims 1 to 8.

Images