Part selection control method and system based on reinforcement learning

Abstract

The present disclosure provides a kind of based on reinforcement learning's spare part selection control method, belong to intelligent manufacturing technical field.The method is by constructing the structured feature set of spare part, based on KNN algorithm and decision tree algorithm, automatically realize the screening and accurate matching of terminal model and seal, greatly reduce the artificial intervention under traditional experience rule, improve the efficiency and accuracy of spare part selection.At the same time, introduce abnormal plating detection and preset matching rule set self-update mechanism, the abnormal plating data found in selection is automatically fed back to matching rule base, realizes the dynamic closed-loop optimization of matching strategy.Compared with the traditional one-time matching rule set or the way that needs to be manually adjusted by R&D personnel, it can respond to supplier data changes and actual application scenario dynamic changes in real time, reduce the operation complexity and error risk, continuously improve the selection accuracy and system robustness, reduce the R&D and production cost, meet the complex and changeable automobile wiring harness manufacturing scene demand.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of intelligent manufacturing, in particular to a part selection control method and system based on reinforcement learning. BACKGROUND

[0002] At present, the selection of automobile wiring harness parts is a crucial link in automobile manufacturing, directly affecting the reliability, cost and production efficiency of the wiring harness system. As the core component of the automobile electrical system, the selection of terminals and seals of the wiring harness needs to ensure accurate matching with various complex parameters to meet the stringent performance requirements. However, the existing selection methods have significant limitations in dealing with complex matching relationships, for example, the same hole position number may correspond to multiple cross-sectional areas and wire types, making it difficult to accurately determine the matching terminals and seals. At the same time, many methods rely on manual experience or static rule bases, which are difficult to adapt to dynamically changing part data and diverse matching requirements, especially when faced with multiple parameter combinations and abnormal data, which can easily lead to low selection efficiency and accuracy, increasing research and production costs.

[0003] The above information disclosed in the background section is only for the purpose of enhancing the understanding of the background of the present disclosure, and therefore it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0004] Therefore, the present disclosure provides a part selection control method based on reinforcement learning, which improves the efficiency and accuracy of part selection.

[0005] In a first aspect, the embodiments of the present application provide a part selection control method based on reinforcement learning, which comprises: obtaining part data from a part database, the part data comprising a stand number, a supplier number, a hole position number, a wire cross-sectional area, a wire type, a terminal model, a seal, and plating data; based on a set of pre-set matching rules, matching the part data to generate a matching data set, the matching data set comprising the corresponding relationship between each part data; determining a structured feature set according to the matching data set, the structured feature set comprising a plurality of data combinations of hole position numbers, wire cross-sectional areas, wire types and terminal models; based on a KNN algorithm, training the structured feature set to determine a plurality of candidate terminal models matched with the wire type; calculating the total score of the matching degree of each candidate terminal model, and taking the candidate terminal model with the highest total score of the matching degree as the terminal model with the highest priority; based on a decision tree algorithm, generating a seal matched with the terminal model with the highest priority; determining abnormal plating data on the seal according to the seal; updating the set of pre-set matching rules according to the abnormal plating data; determining the final matching terminal model and seal according to the updated set of pre-set matching rules.

[0006] Secondly, embodiments of this application provide a component selection control system based on reinforcement learning. This system may include an acquisition module, a matching module, a first determination module, a second determination module, a calculation module, a third determination module, a fourth determination module, an update module, and a fifth determination module. Specifically: The acquisition module is used to acquire component data from a component database. The component data includes Luxshare part numbers, supplier numbers, hole positions, wire cross-sectional areas, wire types, terminal models, seals, and plating data. The matching module is used to match the component data based on a preset matching rule set to generate a matching dataset. The matching dataset includes the correspondence between the various component data. The first determination module is used to determine a structured feature set based on the matching dataset. The structured feature set includes multiple combinations of hole positions, wire cross-sectional areas, wire types, and terminal models. The second determination module is used to perform a KNN algorithm on the structured feature set... The system is trained to identify multiple candidate terminal models that match the wire type; a calculation module calculates the total matching score for each candidate terminal model and selects the candidate terminal model with the highest total matching score as the highest priority terminal model; a third determination module determines the seal that matches the highest priority terminal model based on a decision tree algorithm; a fourth determination module determines abnormal plating data on the seal based on the seal; an update module updates the preset matching rule set based on the abnormal plating data; and a fifth determination module determines the final mutually matching terminal models and seals based on the updated preset matching rule set.

[0007] This application provides a component selection control method based on reinforcement learning. By constructing a structured feature set of components, the method automatically and accurately matches terminal models with seals based on KNN and decision tree algorithms, significantly reducing manual intervention under traditional empirical rules and improving the efficiency and accuracy of component selection. Simultaneously, an abnormal plating detection and preset matching rule set self-updating mechanism are introduced, automatically feeding back abnormal plating data discovered during selection to the matching rule base, achieving dynamic closed-loop optimization of the matching strategy. Compared to traditional one-time matching rule sets or methods requiring manual adjustments by R&D personnel, this method can respond in real-time to changes in supplier data and dynamic changes in actual application scenarios, reducing operational complexity and error risks, continuously improving component selection accuracy and system robustness, reducing R&D and production costs, and meeting the needs of complex and ever-changing automotive wiring harness manufacturing scenarios. Attached Figure Description

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only represent some embodiments of the present disclosure, and for those skilled in the art, other drawings can also be obtained from these drawings without creative effort.

[0009] Figure 1 is a flowchart of a parts selection control method based on reinforcement learning provided by an exemplary embodiment of the present application.

[0010] Figure 2 is a flowchart of a parts selection control method based on reinforcement learning provided by another exemplary embodiment of the present application.

[0011] Figure 3 is a flowchart of a parts selection control method based on reinforcement learning provided by yet another exemplary embodiment of the present application.

[0012] Figure 4 is a flowchart of a parts selection control method based on reinforcement learning provided by still another exemplary embodiment of the present application.

[0013] Figure 5 is a flowchart of a parts selection control method based on reinforcement learning provided by still another exemplary embodiment of the present application.

[0014] Figure 6 is a flowchart of a parts selection control method based on reinforcement learning provided by still another exemplary embodiment of the present application.

[0015] Figure 7 is a flowchart of a parts selection control method based on reinforcement learning provided by still another exemplary embodiment of the present application.

[0016] Figure 8 is a flowchart of a parts selection control method based on reinforcement learning provided by still another exemplary embodiment of the present application. DETAILED DESCRIPTION

[0017] Example embodiments now will be described more fully hereinafter with reference to the accompanying drawings. Example embodiments, however, can be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of example embodiments to those skilled in the art. The features, structures, or characteristics described in connection with the embodiments can be combined in any suitable manner in one or more embodiments.

[0018] The terms "one", "an", "said" are used to denote the presence of one or more elements / components / etc.; the terms "include" and "has" are used to indicate an open-ended inclusion of the elements / components / etc. listed in the following description and that additional elements / components / etc. can be present. The terms "first" and "second" are used only as labels and not as a limitation on the number of objects.

[0019] Currently, the selection of automobile wiring harness components is a crucial link in automobile manufacturing, directly affecting the reliability, cost and production efficiency of the wiring harness system. As the core component of the automobile electrical system, the selection of terminals and seals needs to ensure accurate matching with a variety of complex parameters to meet the stringent performance requirements. However, the existing selection methods have significant limitations in dealing with complex matching relationships, for example, the same hole number may correspond to multiple cross-sectional areas and wire types, making it difficult to accurately determine the matching terminals and seals. At the same time, many methods rely on human experience or static rule bases, which are difficult to adapt to dynamically changing component data and diverse matching requirements, especially when faced with multiple parameter combinations and abnormal data, which can easily lead to low selection efficiency and accuracy, increasing research and production costs.

[0020] In the embodiments of the present disclosure, a component selection control method based on reinforcement learning is provided, as shown in the component selection control method based on reinforcement learning. Figure 1 The method can include the following steps:

[0021] Step S110: Obtain component data from a component database, the component data including a stand number, a supplier number, a hole number, a wire cross-sectional area, a wire type, a terminal model, a seal, and plating data;

[0022] Step S120: Based on a set of preset matching rules, match the component data to generate a matching data set, the matching data set including the corresponding relationship between each component data;

[0023] Step S130: Determine a structured feature set according to the matching data set, the structured feature set including a plurality of data combinations of hole numbers, wire cross-sectional areas, wire types, and terminal models;

[0024] Step S140: Based on a KNN algorithm, train the structured feature set to determine a plurality of candidate terminal models matched with the wire type;

[0025] Step S150: Calculate the total score of the matching degree of each candidate terminal model, and select the candidate terminal model with the highest total score of the matching degree as the terminal model with the highest priority;

[0026] Step S160: Based on a decision tree algorithm, generate a seal matched with the terminal model with the highest priority;

[0027] Step S170: determining abnormal plating layer data on the seal according to the seal;

[0028] Step S180: updating the preset matching rule set according to the abnormal plating layer data;

[0029] Step S190: determining the final mutually matched terminal type and seal according to the updated preset matching rule set.

[0030] According to the component selection control method based on reinforcement learning provided by the present disclosure, the method can extract multi-dimensional component data from the component database, generate corresponding matching relationships between each component data according to the preset matching rule set, train the structured feature set based on the KNN algorithm, and filter out the candidate terminal type matched with the wire type; then, the total score of the matching degree is calculated to determine the terminal type with the highest priority, and the seal is matched with the decision tree algorithm; subsequently, the matching rule set is dynamically updated by detecting the abnormal plating layer data of the seal; and finally, the final mutually matched terminal type and seal are determined according to the updated preset matching rule set.

[0031] In the above method, by constructing the structured feature set of the component, the KNN algorithm and the decision tree algorithm are used to automatically realize the screening and accurate matching of the terminal type and the seal, greatly reducing the manual intervention under the traditional experience rule, improving the component selection efficiency and accuracy. At the same time, the abnormal plating layer detection and the preset matching rule set self-updating mechanism are introduced, the abnormal plating layer data found in the selection is automatically fed back to the matching rule library, and the dynamic closed-loop optimization of the matching strategy is realized. Compared with the traditional one-time matching rule set or the way that needs to be manually adjusted by the R&D personnel, it can respond to the changes of supplier data and the dynamic changes of actual application scenarios in real time, reduce the operation complexity and error risk, continuously improve the selection accuracy and system robustness, reduce the R&D and production costs, and meet the needs of complex and variable automobile wire harness manufacturing scenes.

[0032] The steps of the component selection control method based on reinforcement learning provided by the embodiment of the present disclosure are described in detail as follows:

[0033] In an embodiment of the present disclosure, in step S110, the component data is obtained from the component database, and the component data includes Luxon part number, supplier number, hole number, wire cross-sectional area, wire type, terminal type, seal, and plating layer data. Specifically, an SQL query statement is generated to extract the relevant fields from the database. It is assumed that the component database is named WireHarnessDB, the table name is PartsInfo, and the fields include LuxonPartNo, SupplierNo, HoleNo, CrossSectionArea, and WireType.

[0034] The SQL statement is SELECT LuxonPartNo, SupplierNo, HoleNo, CrossSectionArea, WireType FROM PartsInfo WHERE CrossSectionArea > 0. Assume that the query returns a dataset containing 100 records, such as {LuxonPartNo: 'LX12345', SupplierNo: 'SP6789', HoleNo: 'H01', CrossSectionArea: 2.5, WireType: 'PVC'}. Here, LuxonPartNo is the Luxon part number, SupplierNo is the supplier number, HoleNo is the hole number, CrossSectionArea is the cross-sectional area of the wire, and WireType is the type of wire, with PVC (Polyvinyl Chloride) being the polyvinyl chloride.

[0035] For example, the following component data can be obtained from the component database:

[0036] The Luxon part numbers are LX1, LX2, and LX3, respectively; the supplier numbers are SUP01, SUP02, and SUP03, respectively; the hole numbers are H01, H02, and H03; the cross-sectional areas of the wires are 2.5 mm 2 , 4 mm 2 , and 6 mm 2 , respectively; the wire types are BV (copper core polyvinyl chloride insulated hard wire) and BVR (copper core polyvinyl chloride insulated soft wire); the terminal models are T1, T2, and T3, respectively; the seals can be S001, S002, and S003; and the plating data can include plating thickness, hardness, and material, such as a plating thickness of 0.02 mm, a hardness of HV700, and a material of Ni-P (nickel-phosphorus alloy). It should be noted that the component data is not limited to the above-mentioned list, and can also include other types of different quantities of data, which are not enumerated here.

[0037] Through the above method, all the key parameters required for selecting an automobile wire harness can be extracted from the enterprise-stored component information library, providing a structured basis for subsequent matching, screening, and optimization. These component data cover dimensions such as "identity", "physical parameters", and "performance attributes" of the components, which are the premise for realizing automatic selection.

[0038] In an embodiment of the present disclosure, in step S120, the part data is matched based on the preset matching rule set to generate a matching data set, which includes the corresponding relationship between each part data. Specifically, for example, the MatchingRules in the preset matching rule set stipulates that WireType = 'PVC' and CrossSectionArea = 2.5 correspond to StandardPartNo = 'STD9876'. The data set and the rule table are loaded using the Python pandas library, and a merge operation is performed: df_parts.merge(df_rules, on=[ 'WireType', 'CrossSectionArea' ]) to generate a new column StandardPartNo. For example, terminal model T1, hole position number H01, and sealing member S001 need to be matched to generate a piece of matching data; terminal model T2, hole position number H02, and sealing member S002 need to be matched to generate another piece of matching data, and so on. In this way, multiple pieces of matching data can be combined to generate a matching data set. In the above method, the part data in the database is associated and matched through the predefined matching rules to generate a structured corresponding relationship

[0039] In an embodiment of the present disclosure, in step S130, the structured feature set is determined according to the matching data set, and the structured feature set includes a plurality of data combinations of hole position numbers, wire cross-sectional areas, wire types, and corresponding terminal models. Specifically, the following data combinations can be formed:

[0040] Data combination 1: hole position number H01, wire cross-sectional area 2.5 mm 2 , wire type BV, and corresponding terminal model T1; data combination 2: hole position number H02, wire cross-sectional area 4 mm 2 , wire type BVR, and corresponding terminal model T2; data combination 3: hole position number H03, wire cross-sectional area 4 mm 2 , wire type BV, and corresponding terminal model T3.

[0041] Through the above method, irrelevant data in the matching data set, such as supplier numbers and Liantronics part numbers, is removed, and only key variables that affect terminal selection are retained. This avoids misleading the model with irrelevant data, and enables the subsequent KNN algorithm to efficiently learn the association rules of hole position numbers, wire cross-sectional areas, wire types, and terminal models.

[0042] It should be noted that the KNN (K-Nearest Neighbors) algorithm is a supervised learning algorithm based on "similarity matching." It calculates the feature similarity (distance) between a sample and known samples (training data), selects the K most similar samples (nearest neighbors), and uses the majority result of these K samples as the prediction result for the new sample. In the component selection method of this disclosure, the KNN algorithm is mainly used to screen candidate terminal models that match the target wire type from the structured feature set, and is the core step in the initial terminal selection.

[0043] In one embodiment of this disclosure, step S140, based on the KNN algorithm, trains a structured feature set to determine multiple candidate terminal models that match the wire type. The step also includes the following steps: Figure 2 As shown, the specific content is as follows:

[0044] Step S210: Obtain the dataset of candidate components, which includes hole number, wire cross-sectional area and wire type;

[0045] Step S220: Based on the mapping rules, map the candidate component dataset to a first feature vector and map the structured feature set to multiple second feature vectors;

[0046] Step S230: Based on the KNN algorithm, calculate the Euclidean distance between the first feature vector and multiple second feature vectors respectively;

[0047] Step S240: Sort the multiple Euclidean distance results in order from minimum to maximum value, and remove the wire type results that are different from those in the candidate component dataset to generate the sorting results;

[0048] Step S250: Select the first three from the sorting results as candidate terminal models.

[0049] Specifically, for example, the dataset of components to be selected is {hole number H01, wire cross-sectional area 2.5mm²}. 2 The wire type is BV, meaning it needs to be hole number H01 with a wire cross-sectional area of ​​2.5mm². 2 The component with wire type BV is matched with a suitable terminal model. The structured dataset is {hole number, wire cross-sectional area, wire type, terminal model}, which can include A1{H01, 2.5mm}. 2 BV, terminal model T1}、B1{H02, 4.0mm 2 BV, terminal type T2}, C1{H03, 2.5mm 2 BVR, terminal model T3}, D1{H04, 3mm 2BV, terminal model T4}, wherein the mapping rule is that the hole number is coded as a numerical characteristic according to the number body, the cross-sectional area is directly coded as a numerical characteristic according to the numerical size, and the wire type is coded as a vector represented by 0 and 1. For example, H01=1, H02=2, H03=3, H04=4; BV=[1,0], BVR=[0,1]; 2.5mm2=2.5, 4.0mm2=4.0. According to the above mapping rule, the first feature vector of the selected part data set is [1, 2.5, 1, 0], and the second feature vectors are respectively: A1=[1, 2.5, 1, 0]; B1=[2, 4, 1, 0]; C1=[3, 2.5, 0, 1]; D1=[4, 3, 1, 0]. Based on the KNN algorithm, the Euclidean distances between the first feature vector and the plurality of second feature vectors are calculated respectively, and the calculation formula is as follows:

[0050]

[0051] wherein D is the Euclidean distance, x1, y1, z1, w1 are respectively the four coordinate values of the first feature vector, and x2, y2, z2, w2 are respectively the four coordinate values of the second feature vector.

[0052] According to the above formula, the Euclidean distance D1 between the first feature vector and the second feature vector A1 is calculated as follows:

[0053]

[0054] The Euclidean distance D2 between the first feature vector and the second feature vector A2 is:

[0055]

[0056] The Euclidean distance D3 between the first feature vector and the second feature vector A3 is:

[0057]

[0058] The Euclidean distance D4 between the first feature vector and the second feature vector A4 is:

[0059]

[0060] The above Euclidean distance results are sorted in order from the smallest value to the largest value, and the sorting result is D1<D2<D3<D4. Since the input selected part wire type is BV, the terminal models T1, T2 and T4 with the same wire type are preferentially selected. And the first three in the sorting result are T1, T2 and T4 as candidate terminal models.

[0061] In the above method, by uniformly mapping the hole position number, cross-sectional area and wire type into a multi-dimensional feature vector, the to-be-selected wire and the existing structured data set are calculated in the same dimensional space, which can accurately quantify the similarity between different parts. Compared with artificial experience or static rule base, this method does not need to manually set complex matching threshold, but relies on pure Euclidean distance sorting to quickly lock the most similar terminal model, which significantly improves the candidate screening efficiency and matching accuracy.

[0062] In an embodiment of the present disclosure, in step S150, the total score of the matching degree of each candidate terminal model is calculated, and the candidate terminal model with the highest total score of the matching degree is taken as the terminal model with the highest priority. It further includes the following steps, as shown in Figure 3 The specific content is as follows:

[0063] Step S310: Obtain the wire cross-sectional area range, maximum current value and connection stability value corresponding to the candidate terminal model;

[0064] Step S320: Determine the cross-sectional area matching degree according to the wire cross-sectional area range;

[0065] Step S330: Obtain the weight value corresponding to the cross-sectional area matching degree, maximum current value and connection stability value;

[0066] Step S340: Perform weighted operation on the weight value, and take the weighted operation result as the total score of the matching degree of each candidate terminal model.

[0067] Specifically, assuming that a terminal model needs to be selected for a circuit with a copper core wire type and a cross-sectional area of 2.5mm 2 , the candidate terminal models include candidate terminal model T1 (applicable to copper core wire, cross-sectional area range 1.5-4.0mm 2 , maximum current 20A, connection stability 0.9), candidate terminal model T2 (applicable to copper core wire, cross-sectional area range 2.0-6.0mm 2 , maximum current 25A, connection stability 0.85), and candidate terminal model T3 (applicable to aluminum core wire, cross-sectional area range 1.0-3.0mm 2 , maximum current 18A, connection stability 0.8). First, the terminal models matching the wire type are screened out, and the candidate terminal model T3 is excluded for the copper core wire, and the candidate terminal models T1 and T2 are retained. Then, the cross-sectional area matching degree is calculated, and the calculation formula is:

[0068]

[0069] According to the above formula, the median of the candidate terminal model T1 is (1.5+4.0) / 2=2.75, and the cross-sectional area matching degree is 1-|2.5-2.75| / 2.75=0.909; the median of the candidate terminal model T2 is (2.0+6.0) / 2=4.0, and the cross-sectional area matching degree is 1-|2.5-4.0| / 4.0=0.625. For example, the weight value corresponding to the wire type is 0.4, the weight value corresponding to the cross-sectional area is 0.3, the weight value corresponding to the maximum current is 0.2, and the weight value corresponding to the connection stability value is 0.1. The above weight values are weighted and operated, and the weighted operation result is taken as the total score of the matching degree of each candidate terminal model, and the calculation result is as follows:

[0070] The total score of the matching degree of the candidate terminal model T1 is 0.4x1+0.3x0.909+0.2x(20 / 25)+0.1x0.9=0.8627; and the total score of the matching degree of the candidate terminal model T2 is 0.4x1+0.3x0.625+0.2x(25 / 25)+0.1x0.85=0.8725.

[0071] As can be seen from the above, the candidate terminal model T2 has the highest score, and therefore the candidate terminal model T2 is taken as the terminal model with the highest priority.

[0072] Alternatively, if the business scenario requires consideration of cost, the cost data of the candidate terminal model T2 (such as 0.5 yuan per piece) can be further queried and compared with the cost data of the candidate terminal model T1 (0.6 yuan per piece), and the score and the cost are combined to confirm that the terminal model T2 is the optimal choice.

[0073] In the above method, the cross-sectional area matching degree, the maximum current carrying capacity and the connection stability and other multi-dimensional indexes are quantified, and different weights are given according to the business requirements, so as to realize the comprehensive evaluation of the performance of the candidate terminal. Compared with single-dimensional judgment, this way can more comprehensively consider the electrical and mechanical properties of the terminal in the actual working environment, ensure that the selected terminal can not only meet the current load requirement, but also has excellent contact reliability, thereby greatly reducing the risk of poor contact or overload failure caused by terminal mismatch. This flexible weighting mechanism not only makes the terminal optimization result have higher explainability and controllability, but also provides a convenient way for subsequent system online weight adjustment, rapid response to new business scenarios or supply chain changes, and significantly improves the intelligentization and adaptability of the entire selection process.

[0074] In an embodiment of the present disclosure, in step S160, the sealing element matched with the terminal model with the highest priority is generated based on the decision tree algorithm, and further includes the following steps, as shown in Figure 4 The specific content is as follows:

[0075] Step S410: Obtain a terminal data set corresponding to a terminal model with the highest priority, the terminal data set including a terminal model, a terminal rated current, a terminal aperture, and a terminal material;

[0076] Step S420: Extract a plurality of hole position numbers from a hole position number database, and determine apertures corresponding to the hole position numbers according to the hole position numbers;

[0077] Step S430: Determine whether a difference between the apertures corresponding to the hole position numbers and the terminal aperture is not greater than a first preset difference value;

[0078] Step S440: If the difference is not greater than the first preset difference value, calculate hole position number matching degrees of the plurality of hole position numbers;

[0079] Step S450: Sort the hole position number matching degrees, and take a maximum value in a sorting result as a hole position number with the highest priority;

[0080] Step S460: Construct a decision tree model, an input of the decision tree model being the hole position number with the highest priority, the terminal aperture, the terminal rated current, and the terminal material;

[0081] Step S470: Output a sealing element matching the terminal model with the highest priority according to the decision tree model.

[0082] Specifically, for example, the terminal model with the highest priority is T2, and a terminal data set of T2 includes a rated current of 25 A, an aperture of 3.0 mm, and a material of copper. Hole positions H02 (an aperture of 3.1 mm), H05 (an aperture of 3.3 mm), and H07 (an aperture of 2.9 mm) are obtained from a hole position number database. It is assumed that the first preset difference value is 0.2 mm. Differences between the apertures corresponding to the hole position numbers and the terminal aperture are respectively calculated as follows:

[0083] A difference between the aperture of H02 and the aperture of T2 is |3.1-3.0|=0.1 mm≤0.2 mm; a difference between the aperture of H05 and the aperture of T2 is |3.3-3.0|=0.3 mm>0.2 mm (eliminated); and a difference between the aperture of H07 and the aperture of T2 is |2.9-3.0|=0.1 mm≤0.2 mm.

[0084] A hole position number matching degree calculation formula is as follows:

[0085]

[0086] According to the above formula, the matching degree of H02 is calculated as 1-0.1 / 3.1≈96.8%; the matching degree of H07 is calculated as 1-0.1 / 2.9≈96.6%. It can be seen that the highest priority hole number is H02 (with a higher matching degree). The decision tree model is constructed, and the highest priority hole number H02, the terminal hole diameter 3.0 mm, the current 25 A, and the material copper are input into the model, and the sealing member S002 (thickness 1.3 mm, silicone, temperature resistance-40℃ to 150℃) matching the highest priority terminal model is output.

[0087] In the above method, by refining the traditional experience rule into an interpretable quantitative decision node, the transparency of the selection process is ensured, and the model structure or feature weight can be flexibly updated according to business needs; at the same time, by using the natural multi-condition branching capability of the decision tree, the changes of different terminal parameters and hole numbers can be quickly adapted, and a sealing protection scheme with high robustness and high reliability is provided for the automobile wiring harness system under varying working conditions.

[0088] In an embodiment of the present disclosure, in step S470, the sealing member matching the highest priority terminal model is output according to the decision tree model, and further includes the following steps, as shown in Figure 5 The specific content is as follows:

[0089] Step S510: determining whether the difference between the sealing member hole diameter and the terminal hole diameter is not greater than a second preset difference value;

[0090] Step S520: if the difference is not greater than the second preset difference value, determining whether the sealing member material meets the terminal material compatibility requirement;

[0091] Step S530: if the sealing member material meets the terminal material compatibility requirement, determining whether the sealing member temperature resistance range is not greater than the working temperature corresponding to the terminal rated current;

[0092] Step S540: if the sealing member temperature resistance range is not greater than the working temperature corresponding to the terminal rated current, outputting the sealing member matching the highest priority terminal model.

[0093] Specifically, the above method ensures the physical compatibility, chemical stability and thermal adaptability of the seal and the terminal through triple constraint judgment. It can be verified whether the difference between the seal aperture and the terminal aperture is less than or equal to a second preset difference value (such as 0.1 mm), to ensure that the seal can tightly wrap the terminal and prevent loosening or water ingress. It can also be determined whether the seal material is chemically compatible with the terminal material (such as copper terminals avoiding contact with acidic rubber), to prevent electrochemical corrosion from causing poor contact. It can also confirm whether the temperature resistance range of the seal covers the working temperature of the terminal (such as 25A current corresponding to a temperature rise of about 30°C, so the temperature resistance of the seal needs to be ≥120°C), to avoid seal failure at high temperatures. In this way, physical fit, chemical compatibility and thermal adaptability can be combined organically to ensure that the selected seal is fully matched with the terminal in terms of size, material and temperature, greatly improving the long-term reliability and safety of the wiring harness interface.

[0094] At the same time, the method has high configurability and extensibility: the second preset difference value, the compatibility criteria and the temperature threshold can be flexibly adjusted according to different terminal models and application scenarios, without frequent manual intervention, which is conducive to quickly adapting to new products or supply chain changes, realizing real-time online verification and intelligent recommendation, thereby significantly shortening the design verification period, reducing on-site maintenance costs, and minimizing the risk of failure due to mismatch.

[0095] In an embodiment of the present disclosure, in step S170, determining the abnormal plating layer data on the seal according to the seal further includes the following steps, as shown in the following table: Figure 6

[0096] Step S610: Obtain the plating layer parameters matched with the seal, the plating layer parameters including plating layer thickness and plating layer hardness;

[0097] Step S620: Determine whether the plating layer thickness exceeds a preset thickness threshold;

[0098] Step S630: If it is determined that the preset thickness threshold is exceeded, mark the current plating layer thickness as a potential abnormality, and calculate the standard score of the current plating layer thickness;

[0099] Step S640: Determine whether the standard score of the current plating layer thickness exceeds a first preset standard score threshold;

[0100] Step S650: If it is determined that the first preset standard score threshold is exceeded, determine that it is an abnormal plating layer thickness, and output the abnormal plating layer thickness data;

[0101] Step S660: If it is determined that the preset thickness threshold is not exceeded, determine whether the plating layer hardness exceeds a preset hardness threshold;

[0102] ​Step S670: If it is determined that the preset hardness threshold is exceeded, the current plating layer hardness is marked as a potential abnormality, and a standard score of the current plating layer hardness is calculated;

[0103] Step S680: It is determined whether the standard score of the current plating layer hardness exceeds a second preset standard score threshold;

[0104] Step S690: If it is determined that the second preset standard score threshold is exceeded, the abnormal plating layer hardness is determined, and abnormal plating layer hardness data is output;

[0105] Step S695: Abnormal plating layer data is generated according to the abnormal plating layer thickness data and the abnormal plating layer hardness data.

[0106] Specifically, it is assumed that the plating layer quality requirement of a certain sealing element is that the preset thickness threshold is 50 μm (exceeding which is marked as a potential abnormality), and the first preset standard score threshold is 2 (measuring the severity of the abnormality, exceeding which is determined as an abnormality); the preset hardness threshold is 300 HV (Vickers hardness, exceeding which is marked as a potential abnormality), and the second preset standard score threshold is 1.5 (exceeding which is determined as an abnormality). For example, the actual thickness of the plating layer of a certain sealing element is 62 μm, which exceeds the preset thickness threshold of 50 μm, and is marked as an abnormality. The standard score is calculated through historical data (assuming that the average plating layer thickness of this type of sealing element is 45 μm, and the standard deviation is 6 μm), and the standard score is (62-45) / 6≈2.83, which exceeds the first preset standard score threshold of 2, and thus is determined as an “abnormal plating layer thickness”, and the data is output (such as “thickness 62 μm, standard score 2.83”). For example, the thickness of the plating layer of another sealing element is 48 μm (which does not exceed the preset thickness threshold), but the actual hardness is 330 HV, which exceeds the preset hardness threshold of 300 HV, and is marked as a potential abnormality. The standard score is calculated through historical data (assuming that the average plating layer hardness of this type of sealing element is 280 HV, and the standard deviation is 30 HV), and the standard score is (330-280) / 30≈1.67, which exceeds the second preset standard score threshold of 1.5, and thus is determined as an “abnormal plating layer hardness”, and the data is output (such as “hardness 330 HV, standard score 1.67”). The above abnormal thickness and hardness data are integrated to generate “abnormal plating layer data” of the batch of sealing elements.

[0107] In the above method, through the double judgment mechanism of "threshold judgment + standard score verification", the abnormality of the plating layer thickness and hardness is accurately identified, that is, as long as any one of the thickness and hardness indicators does not meet the requirements, it is marked as a potential abnormality, and when the standard score threshold is not met, it is confirmed that there is an abnormality, and the existence of the abnormality is confirmed through two verifications, so as to avoid misjudgment caused by a single threshold. The core indicators of the plating layer (thickness affects sealing, hardness is related to wear resistance) are comprehensively covered, unqualified sealing elements can be intercepted in time, and the risks of sealing failure and shortened service life caused by plating layer defects (such as excessive thickness cracking and excessive hardness embrittlement) are reduced, which not only improves the reliability of the sealing element, but also promotes the continuous and stable production process.

[0108] In one embodiment of the present disclosure, in step S180, updating the preset matching rule set according to the abnormal plating layer data further includes the following steps, as shown in Figure 7 The specific content is as follows:

[0109] Step S710: Extracting key abnormal features from the abnormal plating layer data;

[0110] Step S720: Determining an abnormal feature set according to the key abnormal features;

[0111] Step S730: Determining an abnormal influence weight based on a decision tree algorithm according to the abnormal feature set;

[0112] Step S740: Judging whether the abnormal influence weight exceeds a preset weight threshold;

[0113] Step S750: If it is judged that the preset weight threshold is exceeded, the abnormal plating layer data in the preset matching rule set is removed to generate an updated preset matching rule set.

[0114] Specifically, for example, the abnormal plating layer data shows that the terminal corrosion rate of a batch of sealing elements may rise by 30% due to the plating layer thickness average of 62 μm (exceeding the preset thickness threshold of 50 μm). The key abnormal features are extracted: thickness > 50 μm, hardness 330 HV (exceeding the preset hardness threshold of 300 HV), and the abnormal feature set is formed: {thickness > 50 μm, hardness > 300 HV}. The abnormal influence weight is calculated by using the decision tree: the preset weight threshold is 0.5, wherein the thickness abnormal weight 0.75 > 0.5, the abnormal plating layer data in the component database needs to be removed, and the "allowable plating layer thickness ≤ 60 μm" in the preset matching rule set is adjusted to "plating layer thickness ≤ 50 μm", so as to avoid matching to similar abnormal samples in the future. The hardness abnormal weight 0.3 < 0.5, which is not processed temporarily.

[0115] In the above method, through the data-driven rule updating mechanism, the matching rule set can automatically adapt to production fluctuations (such as plating process deviation), and the preset matching rule set no longer depends on static preset, but continuously changes based on real-time abnormal plating data, thereby improving the fit degree of the preset matching rule set and the actual working condition, and significantly enhancing the fault tolerance and long-term stability of the component selection system.

[0116] In an embodiment of the present disclosure, in step S190, the final mutually matched terminal type and seal are determined according to the updated preset matching rule set, and the updated preset matching rule set is applied to the structured feature set, so that the final mutually matched terminal type and seal are output based on the KNN algorithm and the decision tree algorithm. In the above method, the updated preset matching rule set (such as plating thickness standard) is used as a hard screening condition to ensure that all output combinations meet the latest quality standards and prevent abnormal parameters from re-entering the supply chain. When the production process fluctuates (such as the plating thickness generally increases), the preset matching rule set is automatically adjusted, and the component selection accuracy can be maintained without manual intervention.

[0117] In an embodiment of the present disclosure, after determining the final mutually matched terminal type and seal according to the updated preset matching rule set in step S190, the following steps are further included, as shown in Figure 8 The specific content is as follows:

[0118] Step S810: Data verification is performed on the final matching result to eliminate matching results that do not meet the data verification. The types of data verification include hole number, wire cross-sectional area, and wire type.

[0119] Step S820: Component selection data is generated according to the eliminated matching results.

[0120] Step S830: Updated Liantronics part numbers and supplier numbers are obtained according to the component selection data.

[0121] Step S840: The component database is updated according to the updated Liantronics part numbers and supplier numbers.

[0122] Specifically, assuming that the final matching result is "terminal type T2 (rated current 25A, hole diameter 3.0mm) + seal S002 (silicone, temperature resistance -50~180℃)". At this time, data verification needs to be performed on the matching result to eliminate matching results that do not meet the data verification. For example, hole number verification: the design requires H02 hole (hole diameter 3.1mm), T2 hole diameter 3.0mm, difference 0.1mm ≤ preset threshold 0.2mm, i.e. passing the verification; wire cross-sectional area verification: terminal type T2: applicable range 2.0-6.0mm 2 , actual wire cross-sectional area 2.5mm 2, data verification passed; wire type verification: T2 is applicable to copper core wire, the actual wire type is copper core, data verification passed. Further output component selection data: {terminal model: T2, seal model: S002, matching parameters: [current 25A, hole diameter 3.0mm, material copper]}. Query the internal code from the component database: T2 corresponds to "LX-T-0025", S002 corresponds to "LX-S-0108"; query the supplier number: T2 is provided by supplier "ABC Electronics", S002 is provided by "SealTech Inc". Add a record in the component database: {internal code: LX-T-0025, supplier number: ABC001, matching rule: [terminal model T2 + seal S002]}, to update the corresponding relationship table in the database, complete the automatic selection process.

[0123] In the above method, all verification and update operations are automatically performed by the system program, avoiding manual entry and communication errors; the updated database record can be used in real time for the next selection and quality tracking, forming a dynamic feedback mechanism of selection rules and actual procurement data. Thus, not only the overall efficiency of the wiring harness product development is improved, but also a reliable data basis is provided for subsequent version iteration, supplier management and quality control.

[0124] In the embodiment of the present disclosure, a component selection control system based on reinforcement learning is also provided, which can include an acquisition module, a matching module, a first determination module, a second determination module, a calculation module, a third determination module, a fourth determination module, an update module, and a fifth determination module. The acquisition module is configured to acquire component data from a component database, the component data including internal codes, supplier numbers, hole numbers, wire cross-sectional areas, wire types, terminal models, seals, and plating data. The matching module is configured to match the component data based on a preset matching rule set to generate a matching data set, the matching data set including corresponding relationships between the component data. The first determination module is configured to determine a structured feature set from the matching data set, the structured feature set including a plurality of data combinations of hole numbers, wire cross-sectional areas, wire types, and corresponding terminal models. The second determination module is configured to train the structured feature set based on a KNN algorithm to determine a plurality of candidate terminal models matched with the wire type. The calculation module is configured to calculate a total score of the matching degree of each candidate terminal model and take the candidate terminal model with the highest total score of the matching degree as the terminal model with the highest priority. The third determination module is configured to determine a seal matched with the terminal model with the highest priority based on a decision tree algorithm. The fourth determination module is configured to determine abnormal plating data on the seal based on the seal. The update module is configured to update the preset matching rule set based on the abnormal plating data. The fifth determination module is configured to determine the final mutually matched terminal model and seal based on the updated preset matching rule set.

[0125] It should be noted that the embodiments of the part selection control system based on reinforcement learning provided by the present application can be specifically used to execute the processing flow of the embodiments of the part selection control method based on reinforcement learning in the above embodiments, and the functions thereof will not be repeated here. Please refer to the detailed description of the above method embodiments.

[0126] From the above description, the part selection control system based on reinforcement learning provided by the embodiments of the present disclosure can automatically realize the screening and accurate matching of terminal types and sealing elements based on KNN algorithm and decision tree algorithm by constructing the structured feature set of the parts, greatly reducing the manual intervention under the traditional experience rule, and improving the part selection efficiency and accuracy. At the same time, the abnormal plating layer detection and the preset matching rule set self-updating mechanism are introduced, the abnormal plating layer data found in the selection is automatically fed back to the matching rule library, and the dynamic closed-loop optimization of the matching strategy is realized. Compared with the traditional one-time matching rule set or the way that needs to be manually adjusted by the R&D personnel, it can respond to the changes of supplier data and the dynamic changes of actual application scene in real time, reduce the operation complexity and error risk, continuously improve the selection accuracy and system robustness, reduce the R&D and production cost, and meet the needs of complex and variable automobile wiring harness manufacturing scene.

[0127] In the embodiments of the present disclosure, an electronic device is also provided, which includes one or more processors, and a memory resource represented by a memory for storing instructions executable by the processor, such as an application program. The application program stored in the memory can include one or more modules each corresponding to a set of instructions. In addition, the processor is configured to execute the instructions to perform the part selection control method based on reinforcement learning described above.

[0128] The electronic device can also include a power supply component configured to perform power management of the electronic device, a wired or wireless network interface configured to connect the electronic device to a network, and an input / output (I / O) interface. The electronic device can be operated based on an operating system stored in the memory, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM or the like.

[0129] In an embodiment, a computer device is also provided, which can be a server. The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store data. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through a network connection. The computer program is configured to be executed by the processor to implement a component selection control method based on reinforcement learning.

[0130] In an embodiment, a computer device is also provided, which can be a server. The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store data. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through a network connection. The computer program is configured to be executed by the processor to implement a component selection control method based on reinforcement learning.

[0131] Also provided in the embodiments of the present disclosure is a non-transitory computer-readable storage medium, when instructions in the storage medium are executed by a processor of an electronic device, the instructions enable the electronic device to perform a component selection control method based on reinforcement learning, including: obtaining component data from a component database, the component data including a Liantronics part number, a supplier number, a hole site number, a wire cross-sectional area, a wire type, a terminal model number, a seal, and plating data; matching the component data based on a preset matching rule set to generate a matching data set, the matching data set including a correspondence between each component data; determining a structured feature set according to the matching data set, the structured feature set including a plurality of data combinations of the hole site number, the wire cross-sectional area, the wire type, and the corresponding terminal model number; training the structured feature set based on a KNN algorithm to determine a plurality of candidate terminal model numbers matched with the wire type; calculating a total score of a matching degree of each candidate terminal model number, and taking a candidate terminal model number with the highest total score of the matching degree as a terminal model number with the highest priority; generating a seal matched with the terminal model number with the highest priority based on a decision tree algorithm; determining abnormal plating data on the seal according to the seal; updating the preset matching rule set according to the abnormal plating data; and determining a final mutually matched terminal model number and seal according to the updated preset matching rule set.

[0132] The present disclosure can take the form of a computer program product implemented on one or more storage media (including but not limited to magnetic storage media, CD-ROM, optical storage media, etc.) including program code. The computer-readable storage media include permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device, or any other non-transmission medium that can be used to store information accessible by a computing device.

[0133] It is to be understood that even though various steps of the method of part selection control based on reinforcement learning in the present disclosure are described in a particular order in the drawings, this is not required or implied as to the order of execution of the steps, nor is it required that all of the steps be executed to achieve the desired result. Additional or alternative, some steps can be omitted, multiple steps can be combined into one step, and / or one step can be broken into multiple steps, all of which are considered part of the present disclosure.

[0134] It will be appreciated that the present disclosure does not limit its application to the details of the module of the part selection control system based on reinforcement learning set forth in the specification. The present disclosure is capable of other embodiments and of being practiced or being carried out in various ways. The foregoing description and variations and modifications of the embodiments disclosed herein are considered to be part of the disclosure. It will be appreciated that the disclosure disclosed and defined in the specification extends to all alternative combinations of two or more of the individual features mentioned or evident from the description and / or drawings. All of these different combinations constitute various alternative aspects of the disclosure. The embodiments of the present disclosure illustrate the best ways known to the inventors of carrying out the disclosure and enabling others to utilize the disclosure.

Claims

1. A component selection control method based on reinforcement learning, characterized in that, include: Component data is obtained from the component database, including Luxshare part number, supplier number, hole position number, wire cross-sectional area, wire type, terminal model, seal and plating data; Based on a preset set of matching rules, the component data is matched to generate a matching dataset, which includes the correspondence between the various component data. A structured feature set is determined based on the matching dataset. The structured feature set includes a combination of data such as multiple hole positions, wire cross-sectional area, wire type, and terminal model. The structured feature set is trained based on the KNN algorithm to determine multiple candidate terminal models that match the wire type; Calculate the total matching score for each candidate terminal model, and select the candidate terminal model with the highest total matching score as the terminal model with the highest priority; Based on the decision tree algorithm, a seal matching the terminal model with the highest priority is generated; Determine abnormal coating data on the seal based on the seal; Update the preset matching rule set based on the abnormal coating data; The final matching terminal types and seals are determined based on the updated preset matching rule set.

2. The component selection control method based on reinforcement learning according to claim 1, characterized in that, The KNN algorithm is used to train the structured feature set to determine multiple candidate terminal models that match the wire type, including: Obtain a dataset of candidate components, which includes hole number, conductor cross-sectional area, and conductor type; Based on the mapping rules, the dataset of candidate parts is mapped to a first feature vector, and the structured feature set is mapped to multiple second feature vectors; Based on the KNN algorithm, the Euclidean distances between the first feature vector and multiple second feature vectors are calculated respectively. The multiple Euclidean distance results are sorted in order from minimum to maximum, and results with wire types different from those in the candidate component dataset are removed to generate a sorting result; The first three from the sorting results are selected as the candidate terminal models.

3. The component selection control method based on reinforcement learning according to claim 2, characterized in that, The hole number is encoded as a numerical feature according to the number body, the cross-sectional area is directly encoded as a numerical feature according to the numerical value, and the wire type is encoded as a vector represented by 0 and 1 according to the unique thermal encoding.

4. The component selection control method based on reinforcement learning according to claim 2, characterized in that, The calculation of the total matching score for each candidate terminal model, and the selection of the candidate terminal model with the highest total matching score as the highest priority terminal model, includes: Obtain the range of conductor cross-sectional area, maximum current value, and connection stability value corresponding to the candidate terminal models; Determine the cross-sectional area matching degree based on the range of the conductor cross-sectional area; Obtain the weight values ​​corresponding to the cross-sectional area matching degree, the maximum current value, and the connection stability value; The weight values ​​are weighted and the result is used as the total matching score for each candidate terminal model.

5. The component selection control method based on reinforcement learning according to claim 1, characterized in that, The step of generating a seal that matches the highest priority terminal model based on a decision tree algorithm includes: Obtain the terminal dataset corresponding to the highest priority terminal model. The terminal dataset includes terminal model, terminal rated current, terminal hole diameter, and terminal material. Extract multiple hole positions from the hole position number database, and determine the hole diameter corresponding to the hole position number based on the hole position number; Determine whether the difference between the hole diameter corresponding to the hole position number and the terminal hole diameter is not greater than a first preset difference; If the difference is not greater than the first preset difference, then calculate the hole position number matching degree of the multiple hole position numbers; The hole position numbers are sorted according to their matching degree, and the maximum value in the sorting result is taken as the hole position number with the highest priority; Construct a decision tree model, the inputs of which are the highest priority hole number, the terminal hole diameter, the terminal rated current, and the terminal material; The decision tree model outputs a seal that matches the terminal model with the highest priority.

6. The component selection control method based on reinforcement learning according to claim 5, characterized in that, The step of matching the seal that matches the highest priority terminal model output by the decision tree model includes: Determine whether the difference between the diameter of the sealing element hole and the diameter of the terminal hole is not greater than a second preset difference; If the difference is not greater than the second preset value, then it is determined whether the material of the sealing element meets the terminal material compatibility requirements; If it is determined that the terminal material compatibility requirement is met, then it is determined whether the temperature resistance range of the seal is not greater than the operating temperature corresponding to the rated current of the terminal. If the operating temperature is determined to be no greater than the rated current of the terminal, then a seal matching the terminal number with the highest priority is output.

7. The component selection control method based on reinforcement learning according to claim 1, characterized in that, The step of determining abnormal coating data on the seal based on the seal includes: Obtain coating parameters that match the seal, including coating thickness and coating hardness; Determine whether the coating thickness exceeds a preset thickness threshold; If the thickness exceeds the preset thickness threshold, the current coating thickness is marked as a potential anomaly, and a standard score for the current coating thickness is calculated. Determine whether the standard score of the current coating thickness exceeds a first preset standard score threshold. If the thickness exceeds the first preset standard score threshold, it is determined to be an abnormal coating thickness, and the abnormal coating thickness data is output. If it is determined that the thickness does not exceed the preset thickness threshold, then it is determined whether the hardness of the coating exceeds the preset hardness threshold. If the hardness exceeds the preset hardness threshold, the current coating hardness is marked as a potential anomaly, and a standard score for the current coating hardness is calculated. Determine whether the standard score of the hardness of the current coating exceeds the second preset standard score threshold. If the hardness of the coating exceeds the second preset standard threshold, it is determined to be an abnormal coating hardness, and the abnormal coating hardness data is output. The abnormal coating data is generated based on the abnormal coating thickness data and the abnormal coating hardness data.

8. The component selection control method based on reinforcement learning according to claim 1, characterized in that, The step of updating the preset matching rule set based on the abnormal coating data includes: Extract key anomalous features from the anomalous coating data; Determine the set of abnormal features based on the key abnormal features; Based on the decision tree algorithm, the weight of the anomaly impact is determined according to the set of anomaly features; Determine whether the weight of the abnormal impact exceeds the preset weight threshold; If the determination exceeds the preset weight threshold, the abnormal coating data in the preset matching rule set is removed to generate an updated preset matching rule set.

9. The component selection control method based on reinforcement learning according to claim 1, characterized in that, After determining the final matching terminal type and seal according to the updated preset matching rule set, the method further includes: The final matching results are validated to remove those that do not meet the validation criteria. The validation criteria include hole number, conductor cross-sectional area, and conductor type. Component selection data is generated based on the eliminated matching results; Obtain updated Luxshare part numbers and supplier numbers based on the component selection data; The component database is updated based on the updated Luxshare part number and the updated supplier number.

10. A component selection control system based on reinforcement learning, characterized in that, include: The acquisition module is used to acquire component data from the component database. The component data includes Luxshare part number, supplier number, hole position number, wire cross-sectional area, wire type, terminal model, seal and plating data. The matching module is used to match the component data based on a preset matching rule set to generate a matching dataset, wherein the matching dataset includes the correspondence between the various component data. The first determining module is used to determine a structured feature set based on the matching dataset. The structured feature set includes a combination of multiple hole positions, wire cross-sectional area, wire type, and corresponding terminal model data. The second determining module is used to train the structured feature set based on the KNN algorithm to determine multiple candidate terminal models that match the wire type. The calculation module calculates the total matching score for each candidate terminal model and selects the candidate terminal model with the highest total matching score as the terminal model with the highest priority. The third determining module is used to determine the seal that matches the terminal model with the highest priority based on a decision tree algorithm; The fourth determining module is used to determine abnormal coating data on the seal based on the seal. The update module is used to update the preset matching rule set based on the abnormal coating data; The fifth determining module is used to determine the final matching terminal type and seal based on the updated preset matching rule set.

Images