Process retrieval and matching method based on process fingerprints

Abstract

The invention provides a process retrieval and matching method based on a process fingerprint. The method comprises the following steps: S1, acquiring current part information of a target new part and generating a current process fingerprint; s2, traversing a process library based on the current process fingerprint, and performing process class matching on each initial process class to find the most similar process class; s3, performing intra-class process matching on each initial standard process procedure in the most similar process class according to the current process fingerprint so as to find out the most similar process procedure and calculate feature similarity; s4, whether the feature similarity is larger than a preset threshold value or not is judged, and if yes, the current part information is distributed to the category of the most similar process category for process grouping; and if not, creating a new process class different from each initial process class in the process library, and distributing the current part information to the category of the new process class for process grouping. The method has the beneficial effects that automatic, precise and self-adaptive process retrieval and matching can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of process analysis, in particular to a process retrieval and matching method based on process fingerprints. BACKGROUND

[0002] In traditional derivative computer-aided process planning (CAPP) systems, process retrieval and matching mainly rely on process grouping techniques based on coding or key features. Specifically, the system first needs to rely on process personnel to manually prepare complex classification codes for parts, or manually extract limited part features (such as geometric shapes, materials, etc.). Subsequently, the system divides the parts into pre-set, relatively fixed part families according to the manually input codes or features. When a new part needs to be planned for process, the system will match its code or feature with each part family, find the closest family, and retrieve a similar existing process plan from the family for process personnel to modify and use.

[0003] However, this traditional method has significant drawbacks: High dependence on manual experience and subjective judgment: the quality of part coding and feature extraction is heavily dependent on the personal experience and subjective judgment of process personnel, which is prone to introduce inconsistencies and errors, leading to inconsistent standards for subsequent classification and retrieval; Low retrieval efficiency and accuracy: the similarity calculation model based on simple coding or limited features is relatively rough, and it is difficult to fully and finely reflect the essential similarity of the process, which often leads to inaccurate retrieval results and the inability to quickly locate the most similar process, resulting in low process reuse rate; System rigidity, lack of adaptability: traditional part family division is usually static and pre-defined. For new parts that cannot be classified into any existing part family, the system lacks effective dynamic processing mechanisms, often requiring manual re-creation of the family, making it difficult for the system to intelligently evolve and expand with the accumulation of process knowledge.

[0004] The above shortcomings seriously restrict the automation level and application effect of derivative CAPP systems, and there is an urgent need for a method that can automatically, accurately and adaptively implement process retrieval and matching. SUMMARY

[0005] The technical problem to be solved by the present application is how to achieve automatic, accurate and adaptive process retrieval and matching. To overcome the defects of the above prior art (or related art), the present application provides a process retrieval and matching method based on process fingerprints.

[0006] The application provides a process retrieval and matching method based on process fingerprints, a plurality of historical part information corresponding to historical parts is pre-divided into different initial process categories according to process similarity clustering for process grouping, and each initial part family is integrated into a process library, and the process retrieval and matching method comprises the following steps: Step S1, obtaining current part information corresponding to a target new part, performing feature extraction on the current part information to obtain current process feature data, and generating corresponding current process fingerprints according to the current process feature data; Step S2, based on the current process fingerprints, traversing the process library to match each initial process category to find the most similar process category; Step S3, according to the current process fingerprints, in the most similar process category, each pre-stored initial standard process procedure is in-process matched to find the most similar process procedure, and the feature similarity is obtained according to the current process fingerprints and the most similar process procedure; Step S4, judging whether the feature similarity is greater than a preset threshold: If yes, the current part information is assigned to the category of the most similar process category for process grouping; If not, a new process category different from each initial process category is created in the process library, and the current part information is assigned to the category of the new process category for process grouping.

[0007] Compared with the prior art, the process retrieval and matching method based on process fingerprints has the following advantages: In the application, process category matching and in-process matching are performed through process feature data, which significantly improves the efficiency and accuracy of process retrieval, avoids subjective errors compared with traditional manual coding and feature extraction, realizes fast and accurate feature similarity calculation, and automatically determines whether to create a new process category through a preset threshold, ensures dynamic updating and adaptability of the process library, supports efficient operation of the derived CAPP system, realizes automatic, accurate and adaptive process retrieval and matching; and the similarity calculation based on process fingerprints is used for clustering and classification, which can match from the deep features of the process, thereby significantly improving the accuracy of process retrieval and similar process recommendation.

[0008] In a possible implementation, before the step S1 is performed, the process library construction process further comprises: Step A1, for each historical part information, performing feature extraction on the historical part information to obtain historical process feature data; Step A2, generating a corresponding initial process fingerprint according to each of the historical process feature data, each of the initial process fingerprints being associated with a corresponding initial standard process flow, and performing clustering processing on each of the initial process fingerprints to obtain a corresponding initial process class and storing into the process library.

[0009] Compared with the prior art, the above technical scheme can automatically process historical part information through a systematic process library construction process, and classify historical process feature data through a clustering algorithm, which reduces manual intervention, improves the construction efficiency and quality of the process library, and provides a reliable foundation for subsequent process retrieval.

[0010] In a possible implementation, in the step A1, the feature extraction on the historical part information obtains initial topological feature, initial shape feature, initial material type feature, and initial material state feature to form the historical process feature data.

[0011] In a possible implementation, in the step A2, according to the initial topological feature, the initial shape feature, the initial material type feature, and the initial material state feature, initial directory, initial process hash, initial process information, and initial work step information are obtained to form the initial process fingerprint.

[0012] In a possible implementation, in the step S1, the corresponding real-time topological feature, real-time shape feature, real-time material type feature, and real-time material state feature are extracted from the current part information to form the current process feature data.

[0013] In a possible implementation, in the step S1, according to the real-time topological feature, the real-time shape feature, the real-time material type feature, and the real-time material state feature, real-time directory, real-time process hash, real-time process information, and real-time work step information are obtained to form the current process fingerprint.

[0014] Compared with the prior art, after the above technical scheme is adopted, by comprehensively extracting topological features, shape features, material type features, and material state features, a multi-dimensional process fingerprint is constructed, which can comprehensively and deeply depict the essence of a process, and provides a reliable data foundation for subsequent high-precision similarity comparison.

[0015] In a possible implementation, in the step S4, after the new process class is created, the step further includes: output the current process characteristic data to a process modification and editing interface of an external host computer, so that an operator adds a process path including a process sequence and a step sequence to the current process characteristic data, forms a corresponding new standard process flow and associates the new standard process flow with the current part information, and then distributes the new standard process flow to a category of the new process category for process grouping.

[0016] Compared with the prior art, the above technical scheme can automatically create a new process category and trigger a manual editing process for a new part that cannot match an existing initial process category, thereby ensuring continuous improvement and expansion of the process library, which enhances adaptability and practicality and supports accumulation and reuse of process knowledge. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 is a step flowchart of the overall process of the present application; Figure 2 is a step flowchart of the process library construction process of the present application. DETAILED DESCRIPTION

[0018] First of all, those skilled in the art should understand that these embodiments are only used to explain the technical principles of the embodiments of the present application and are not intended to limit the protection scope of the embodiments of the present application. Those skilled in the art can adjust them as needed to adapt to specific application occasions.

[0019] The present application will be further described in detail below in conjunction with the drawings and specific embodiments.

[0020] Referring to Figure 1 The embodiments of the present application disclose a process retrieval and matching method based on process fingerprints, a plurality of historical parts corresponding to historical part information are pre-clustering divided into different initial process categories according to process similarity for process grouping, and each of the initial part families is incorporated into a process library. The process retrieval and matching method comprises the following steps: Step S1, obtaining current part information corresponding to a target new part, and performing feature extraction on the current part information to obtain current process characteristic data, and generating a corresponding current process fingerprint according to the current process characteristic data; Step S2, based on the current process fingerprint, traversing the process library to match each of the initial process categories to find the most similar process category; Step S3, according to the current process fingerprint, in the most similar process category, each of the pre-stored initial standard process procedures is in-class process matching to find the most similar process procedure, and the feature similarity is obtained according to the current process fingerprint and the most similar process procedure; Step S4, judging whether the feature similarity is greater than a preset threshold: If yes, the current part information is assigned to the category of the most similar process class for process grouping; If no, a new process class is created in the process library which is different from each initial process class, and the current part information is assigned to the category of the new process class for process grouping.

[0021] Referring to Figure 2 Before the step S1, a process library construction process is further included, and the process library construction process comprises: Step A1, for each historical part information, feature extraction is performed on the historical part information to obtain historical process feature data; Step A2, according to each historical process feature data, a corresponding initial process fingerprint is generated, each initial process fingerprint is associated with a corresponding initial standard process flow, and each initial process fingerprint is clustered to obtain a corresponding initial process class and stored in the process library.

[0022] In the embodiment of the application, in the step A1, the historical part information is extracted to obtain initial topological features, initial shape features, initial material type features and initial material state features, and in the step A2, according to the initial topological features, initial shape features, initial material type features and initial material state features, initial directory, initial process hash, initial process information and initial process step information are processed to obtain the initial process fingerprint, wherein the initial topological features include initial process step number, initial process step length, initial process step number, initial process length and initial process number sequence composed of each initial process number, the initial shape features include initial part shape number, the initial material type features include initial part material type number, and the initial material state features include initial part material state number, and in the step A2, the initial part shape number, the initial part material type number, the initial part material state number and the initial process length are used to form the initial directory, the initial process number and the initial process step length are used to form the initial process information, the initial process step number is used to form the initial process step information, and the initial process number sequence is used for fuzzy hash processing to obtain the initial process hash.

[0023] In this embodiment of the invention, in step S1, the corresponding real-time topology features, real-time shape features, real-time material type features, and real-time material status features are extracted from the current part information. The real-time topology features, real-time shape features, real-time material type features, and real-time material status features are then processed to obtain a real-time catalog, a real-time process hash, real-time process information, and real-time step information to form the real-time process fingerprint. The real-time topology features include the real-time process number, real-time step length, real-time step number, real-time process length, and a real-time process number sequence composed of all the real-time process numbers. The real-time shape features include real-time component shape numbers. The real-time material type features include real-time component material type numbers. The real-time material status features include real-time component material status numbers. In step S1, the real-time catalog is formed based on the real-time component shape number, the real-time component material type number, the real-time component material status number, and the real-time process length. The real-time process information is formed based on the real-time process number and the real-time step length. The real-time step information is formed based on the real-time step number. The real-time process hash is obtained by performing fuzzy hashing on the real-time process number sequence.

[0024] In this embodiment of the invention, the feature similarity calculation between the current process fingerprint and the most similar process specification is actually the feature similarity calculation between the current process fingerprint and the initial process fingerprint to which the most similar process specification belongs. There are many algorithms for calculating the feature similarity between process fingerprints; one implementation is given here: The process fingerprint consists of four parts: <directory, process hash, process information, and process step information>. Although the directory, process information, and process step information can also be used to calculate the similarity between two process fingerprints, the essence of this invention is not to calculate similarity, but to use process similarity to cluster and thus group processes. Therefore, this embodiment can provide a very simple method for calculating the similarity of process fingerprint features without affecting the overall idea of ​​process grouping technology. This algorithm only uses process hashing to calculate similarity, assuming the process number sequence is C1C2C3...C n Its process hash is a string H(C1C2C3) H(C2C3C4)…H(C n-2 C n-1 C n H(s) is a hash function that converts 6 characters into 4 characters, transforming the process hash into a string set where each member is a four-byte character derived from H(C). k C k+1 C k+2 ), where the size of the string set is n-2, and the calculation formula is: wherein, denotes the feature similarity, and denotes two process fingerprints, respectively, a real-time process fingerprint and an initial process fingerprint, and denotes two process hash transformations resulting in a set of strings.

[0025] In the embodiment of the present application, a K-Medoids clustering algorithm that can be used in the clustering process is listed, K process fingerprints are randomly selected as initial Medoids, each process fingerprint is assigned to a cluster in which the Medoid is the most similar (the SIM value is the largest), for each cluster, an attempt is made to exchange the Medoid and another process fingerprint, the total similarity after the exchange is calculated, if the total similarity increases, the exchange is retained; if it does not increase, it is deleted, the above actions are repeated until the Medoids no longer change.

[0026] In the embodiment of the present application, each process is numbered, the number of which consists of two bytes; each process step is numbered, the number of which also consists of two bytes, the process sequence is the process route, for example, blanking -> rough turning -> heat treatment -> fine turning -> rough milling -> fine milling -> outer circle grinding -> inspection, the process step sequence is the process, the shape features include shafts, disc sleeves, boxes, supports, gears, sheet metal parts, castings, forgings, injection molded parts, etc.; the material type features include steel (carbon steel, alloy steel), cast iron, aluminum alloy, copper alloy, plastic, composite material, etc.; the material state features include bar stock, forgings, castings, profiles, plates, etc.

[0027] In the embodiment of the present application, in step A1, the historical part information of a plurality of historical parts is batch imported from the existing product data management (PDM) or enterprise resource planning (ERP) system of the enterprise, and the historical part information generally includes a three-dimensional model, a two-dimensional drawing, a technical parameter table and the like of the historical part. The specific implementation process of feature extraction is to automatically extract key process features from the historical part information through an integrated feature recognition module. In the embodiment, the extracted features include a historical shape number, a historical material type number and a historical material state number. The above three numbers together constitute historical process feature data of a historical part. The historical shape number is a classification code based on the geometric shape and processing features of the part. For example, according to whether the historical part is a rotary body, a box body, a support or a thin-walled part, and the main features such as holes, grooves, threads and curved surfaces thereon, a unique string or numerical sequence is generated according to a preset coding rule (such as a simplified or variant of Opitz coding). For example, a simple shaft part can be coded as ROTARY_SHAFT_01. The historical material type number refers to the type of material used by the historical part, such as 45 steel, aluminum alloy 6061, stainless steel 304 and the like. Each type of material corresponds to a predefined number, such as MAT_STEEL_45. The historical material state number refers to the initial state or heat treatment requirement of the material, such as a forged part, a cast part, a bar stock, quenching + tempering and the like, which is also coded as STATUS_BAR or STATUS_QUENCHED_TEMPERED.

[0028] In the embodiment of the present application, steps A1 and A2 are a clustering process, aiming to automatically merge the initial process fingerprints of similar historical parts into the same family, i.e., the initial process class. In step A2, all initial process fingerprints are traversed. If a certain initial process fingerprint is unique and unclassified in the current process library (i.e., it cannot be directly associated with any existing data in the process library), it will be created as a new initial process class, for example, the first processed initial process fingerprint will trigger the creation of a new initial process class named Class_001. For non-unique or to-be-classified initial process fingerprints, the class similarity between them and all other initial process fingerprints in the process library will be calculated in turn. The class similarity can be calculated using algorithms such as cosine similarity or Jaccard similarity. The three numbers are treated as a feature vector for processing, for example, after vectorizing the shape, material type, and material state, the cosine value between the two historical part fingerprint vectors is calculated. The closer the value is to 1, the more similar they are. A second preset threshold (e.g., 0.7) is set. If the class similarity of two initial process fingerprints is greater than the second preset threshold, they are considered to belong to the same process family, a new initial process class is created, and all initial process fingerprints with a class similarity higher than the second preset threshold are assigned to the category of the initial process class. This process is repeated until all initial process fingerprints are classified.

[0029] In the embodiment of the present application, after the initial process class is automatically created, a human-computer interaction process is started to form a standard process plan. The specific implementation is as follows: the classified historical process feature data (single or batch) is output to the process modification and editing interface of the external host computer. The process engineer adds a process path containing process sequence and step sequence for the historical process feature data of each historical part in the process modification and editing interface, for example, for the shaft part with historical process feature data (SHAFT_01, MAT_STEEL_45, STATUS_BAR), the engineer may define its standard process path as:

cutting

rough turning

heat treatment

finishing turning

grinding

inspection

rough turning

[0030] In the embodiment of the present application, step S1 is similar to step A1, the drawing or model of the target new part is obtained, and the same feature extraction rule is used to generate the current process feature data thereof, in step S2, the pre-constructed process library is traversed, the similarity between the current process fingerprint and the initial process fingerprint (which can be the center point of the class or the average vector of all member fingerprints) of each initial process class is calculated, and finally, the class with the highest similarity is selected as the most similar process class; in step S3, after the most similar process class is found, the current process fingerprint is further compared with the historical process fingerprint data corresponding to each initial standard process procedure stored in the class.

[0031] In the embodiment of the present application, the specific calculation of the feature similarity will be performed through the following three-level comparison: The first similarity: comparing the similarity of the current shape number (SHAFT_02) and the historical shape number (SHAFT_01) in the initial standard process procedure, since the number is structured, a string similarity algorithm such as edit distance (Levenshtein Distance) can be used here; The second similarity: comparing the similarity of the current material type number and the historical material type number, since the material type is predefined, if they are completely the same, the similarity is 1, if they belong to the same category (such as both being carbon steel), the similarity can be 0.8, otherwise, the similarity is 0; The third similarity: comparing the similarity of the current material state number and the historical material state number; Finally, the three similarities are weighted and averaged or arithmetically averaged to obtain the final feature similarity.

[0032] In the embodiment of the present application, step S4 is used to judge whether the calculated feature similarity is greater than a preset threshold (for example, 0.85), if yes (feature similarity>0.85), it means that there is a highly similar process in the existing process library, and the current part information of the target new part is automatically assigned to the category of the most similar process class, and the process engineer can directly retrieve and use the most similar process procedure matched therewith as a template for fine tuning, greatly improving the efficiency; if not (feature similarity≤0.85), it means that the target new part has uniqueness, and the existing initial process class cannot cover it well, at this time, a new process class different from the initial process classes will be created in the process library, and the current part information will be assigned to the new process class, and the current process feature data of the target new part will be pushed to the process modification and editing interface, prompting the process engineer to create a completely new corresponding new standard process flow and associate it to the new process class, which ensures that the process library can be continuously self-expanded and self-improved with production practice.

[0033] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "in this embodiment", "specific example", or "some examples" and the like means that the specific features, mechanisms, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, mechanisms, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine the different embodiments or examples described in the specification and the features of the different embodiments or examples without contradiction.

[0034] The above description is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A process retrieval and matching method based on process fingerprints, characterized in that, The historical part information corresponding to multiple historical parts is pre-grouped into different initial process categories according to process similarity, and each initial part family is incorporated into the process library. The process retrieval and matching method includes the following steps: Step S1: Obtain the current part information corresponding to the target new part, extract features from the current part information to obtain current process feature data, and generate the corresponding current process fingerprint based on the current process feature data; Step S2: Based on the current process fingerprint, traverse the process library to perform process class matching on each of the initial process classes to find the most similar process class; Step S3: Based on the current process fingerprint, perform intra-class process matching on each of the pre-stored initial standard process procedures in the most similar process class to find the most similar process procedure, and obtain the feature similarity based on the current process fingerprint and the most similar process procedure. Step S4: Determine whether the feature similarity is greater than a preset threshold. If so, the current part information is assigned to the category of the most similar process class for process grouping; If not, a new process class, distinct from each of the initial process classes, is created in the process library, and the current part information is assigned to the category of the new process class for process grouping.

2. The process retrieval and matching method according to claim 1, characterized in that, Before performing step S1, a process library construction process is also included, which includes: Step A1: For each piece of historical part information, perform feature extraction on the historical part information to obtain historical process feature data; Step A2: Generate corresponding initial process fingerprints based on the historical process feature data. Each initial process fingerprint is associated with a corresponding initial standard process flow. Cluster the initial process fingerprints to obtain the corresponding initial process class and store it in the process library.

3. The process retrieval and matching method according to claim 2, characterized in that, In step A1, feature extraction is performed on the historical part information to obtain initial topological features, initial shape features, initial material type features, and initial material state features, which together form the historical process feature data.

4. The process retrieval and matching method according to claim 3, characterized in that, In step A2, the initial process fingerprint is formed by processing the initial topology features, the initial shape features, the initial material type features, and the initial material state features to obtain the initial directory, the initial process hash, the initial process information, and the initial process step information.

5. The process retrieval and matching method according to claim 1, characterized in that, In step S1, the corresponding real-time topology features, real-time shape features, real-time material type features, and real-time material state features are extracted from the current part information to form the current process feature data.

6. The process retrieval and matching method according to claim 5, characterized in that, In step S1, the real-time catalog, real-time process hash, real-time process information, and real-time process step information are processed according to the real-time topology features, the real-time shape features, the real-time material type features, and the real-time material state features to form the current process fingerprint.

7. The process retrieval and matching method according to claim 1, characterized in that, In step S4, after creating the new process class, the following is also included: The current process feature data is output to the process modification and editing interface of an external host computer, so that the operator can add process paths containing process sequence and step sequence to the current process feature data for association, form a corresponding new standard process flow, and then associate it with the current part information and assign it to the category of the new process class for process grouping.

Images