A method and system for automatically searching for relevant work tickets according to maintenance instructions

Abstract

This invention discloses a method and system for automatically finding relevant work orders based on maintenance instructions, relating to the field of intelligent manufacturing technology. The method includes acquiring maintenance instruction content and work order information, and preprocessing the maintenance instruction content and work order information; constructing text feature vector representations of the maintenance instruction content and work orders, calculating the text similarity score of the maintenance instruction content and work orders; identifying potentially affected work orders based on the text similarity score, and sending the work order information to the responsible person. This invention avoids repetitive manual work by automatically acquiring maintenance instructions and work order information. Considering the professional domain characteristics of the maintenance instruction and work order corpus, it adopts a domain-adaptive preprocessing strategy, integrating a bag-of-words model and a pre-trained language model to construct rich text feature vector representations, effectively capturing the key semantic information of the maintenance instruction and work order texts.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing technology, and in particular to a method and system for automatically searching for relevant work tickets according to maintenance instructions. Background Art

[0002] With the continuous improvement of industrial automation, production equipment and processes are becoming more and more complex, and the management of equipment overhaul, maintenance and other operations has become increasingly important. In modern factories, overhaul operations and work ticket systems are key links to ensure safe and efficient production operations. Overhaul instructions, as an instruction document for overhaul operations, usually contain detailed information such as equipment name, overhaul content, and schedule. The work ticket is a written description of the safety protection measures taken at the production site, and is an important management tool for the production site.

[0003] Traditional maintenance operation management methods often rely on manual operations, which are inefficient and prone to errors. In recent years, with the development of artificial intelligence and natural language processing technologies, more and more intelligent methods have been applied in the industrial field. Some scholars have proposed maintenance operation management optimization methods based on text mining, which automatically extract key information from maintenance records and work tickets to provide support for decision-making. These methods mainly rely on constructing feature vector representations of texts and performing text matching and clustering based on vector similarity.

[0004] However, the existing technology still has some shortcomings when applied to the matching of maintenance instructions and work tickets. First, the domain corpus involved in maintenance instructions and work tickets often has a certain degree of professionalism, and directly using general text preprocessing and feature construction methods may not achieve the desired results. Secondly, when calculating text similarity, the existing methods lack effective modeling of domain semantics and rely too much on literal text matching, resulting in limited accuracy of the results. Furthermore, most methods simply give a similarity score and lack an effective subsequent processing flow, making it difficult to truly apply the calculation results to actual work scenarios. Summary of the invention

[0005] In view of the problems existing in the matching of existing maintenance instructions and work orders, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is that the reliance on literal text matching results in limited accuracy.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, an embodiment of the present invention provides a method for automatically searching for relevant work tickets according to a maintenance brief, which comprises:

[0009] Obtaining maintenance instructions and work ticket information, and preprocessing the maintenance instructions and work ticket information;

[0010] Constructing a text feature vector representation of the maintenance briefing content and the work ticket, and calculating a text similarity score between the maintenance briefing content and the work ticket;

[0011] Potentially affected work tickets are identified based on the text similarity scores, and the work ticket information is sent to the person in charge.

[0012] As a preferred solution of the method for automatically searching for relevant work tickets according to the maintenance instructions of the present invention, the preprocessing adopts segmented logic processing, and different preprocessing operations are performed for different data, as shown in the following formula:

[0013] ,

[0014] Among them, x represents the input sequence, y represents the labeled sequence, is the corresponding weight, is the normalization factor, is the weight parameter, and are the mean and standard deviation of the data, respectively. Fill in missing values. It is a format conversion function.

[0015] As a preferred solution of the method for automatically searching for relevant work tickets according to maintenance instructions of the present invention, the text feature vector representation is constructed including:

[0016] Pre-process the maintenance instructions and work ticket information;

[0017] Build vocabulary for maintenance briefing and work ticket information;

[0018] Calculate the frequency of each text vocabulary entry;

[0019] Calculating a weighted value of the word frequency;

[0020] constructing a feature vector representation according to the weighted values;

[0021] The word vector is obtained according to the word vector model and is concatenated with the feature vector to form the final text vector.

[0022] As a preferred solution of the method for automatically searching for relevant work tickets according to the maintenance instructions of the present invention, the weighted value is calculated to reduce the weight of common words and increase the weight of rare words, as shown in the following formula:

[0023] ,

[0024] Among them, w is a word, d is a text, N is the total number of texts, and n(w) is the number of texts containing w.

[0025] As a preferred solution of the method for automatically searching for relevant work tickets according to the maintenance instructions of the present invention, the text similarity score is calculated by merging different information sources by weighted average, as shown in the following formula:

[0026] S ( A → , B → ) = 1 2 [ A → bow ⋅ B → bow | A → bow || B → bow | + a A → emb ⋅ B → emb | A → emb || B → emb | + β 2 | A → bow − B → bow | 1 | A → bow | 1 + | B → bow | 1 ⋅ 1 1 + e − γ ( A → bow ⋅ B → bow ) ] ,

[0027] in, To check the text feature vector, To check the word vector, is the feature vector of the work ticket text, is the word vector of the work ticket, and Represents the number of words in the maintenance briefing and work ticket, is the weight parameter.

[0028] As a preferred solution of the method for automatically searching for relevant work tickets according to the maintenance instructions of the present invention, the method of identifying potentially affected work tickets according to the text similarity score includes:

[0029] If the text similarity score is higher than the preset threshold, the detailed information of the current work ticket is obtained and the current work ticket is added to the list of potentially affected tickets. If the text similarity score is lower than the preset threshold, the current work ticket is placed in the list of unaffected tickets. If the list of potentially affected tickets is not empty, each worksheet in the list of potentially affected tickets is traversed, the similarity score corresponding to it is recalculated, and the worksheets related to the highest similarity score are packaged and sent to the person in charge. After receiving the information, the person in charge checks the specific content to confirm whether it is actually affected by this maintenance. If affected, no other operation is required. If not affected, feedback is given to the system to optimize the similarity model. If the recalculated similarity score is lower than the preset threshold, the relevant worksheet is added to the list of unaffected tickets.

[0030] As a preferred solution of the method for automatically searching for relevant work tickets according to maintenance instructions of the present invention, the selection of the threshold is shown in the following formula:

[0031] ,

[0032] in, .

[0033] In a second aspect, an embodiment of the present invention provides a system for automatically searching for related work tickets according to maintenance instructions, which includes:

[0034] Data acquisition module, used to obtain maintenance instructions and work ticket information, and perform pre-processing;

[0035] Feature construction module, used to construct text feature vector representation of maintenance instructions and work ticket information;

[0036] A similarity calculation module is used to calculate the text similarity score between the maintenance instructions and each work ticket information;

[0037] A potential ticket identification module is used to identify potentially affected work ticket information according to the similarity score and send the affected work ticket information to the person in charge.

[0038] In a third aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, it implements any step of the above-mentioned method of automatically searching for related work tickets based on maintenance instructions.

[0039] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the above-mentioned method of automatically searching for relevant work tickets based on maintenance instructions is implemented.

[0040] The beneficial effect of the present invention is that it avoids manual repetitive work by automatically acquiring maintenance instructions and work ticket information. According to the professional field characteristics of the maintenance instructions and work ticket corpus, a domain-adaptive preprocessing strategy is adopted, and the bag-of-words model and the pre-trained language model are integrated to construct a rich text feature vector representation, which can effectively capture the key semantic information of the maintenance instructions and work ticket texts. The semantic similarity and structural distance are integrated to make the similarity calculation more comprehensive. When setting the threshold, the precision and recall rates are balanced to ensure the reliability of the results. An end-to-end recognition and sending process is established, which can be directly applied to actual scenarios. A manual confirmation link is introduced to ensure the accuracy of the results. The model is optimized through manual feedback to form a closed loop and continuously improve performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0042] Figure 1 The present invention is a flowchart of a method for automatically finding related work tickets according to maintenance instructions. DETAILED DESCRIPTION

[0043] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0044] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0045] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0046] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0047] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0048] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0049] Example 1

[0050] Reference Figure 1, which is the first embodiment of the present invention, provides a method for automatically searching for related work tickets according to maintenance instructions, including:

[0051] S1: Obtain maintenance instructions and work ticket information, and pre-process the maintenance instructions and work ticket information.

[0052] The maintenance briefing contents submitted by maintenance personnel are collected from the production information management system. The maintenance briefing contents usually include basic information such as equipment name, maintenance type, maintenance reason, and maintenance content description. This information is often stored in the database of the production information management system in a fixed text format. The system database is connected and the relevant data of the maintenance briefing are extracted according to specific SQL query statements.

[0053] Furthermore, all uncancelled work ticket information is collected. The work ticket is a written description of the safety protection measures taken for production operations and is an important management tool for the production site. Work ticket information usually includes work ticket number, work type, work content description, safety measures requirements, work area, responsible personnel, etc. This information is also stored in the work ticket data table of the production information management system and needs to be extracted by connecting to the database and executing SQL statements.

[0054] The work ticket information may contain industry-specific abbreviations and technical terms, and it is worth noting that only the work tickets that are currently in execution and have not been cancelled need to be collected, and the cancelled historical work tickets can be ignored.

[0055] Furthermore, the contents of maintenance instructions and work tickets are both text data, which need to be preprocessed through the program. For numerical data such as temperature and pressure, normalization or standardization is required so that data of different dimensions can be compared fairly. The following is presented in the form of a formula.

[0056] For example, in a feasible embodiment, the text is preprocessed as follows:

[0057] First, consider the preprocessing of text data, including word segmentation, removal of stop words, and part-of-speech restoration. The commonly used word segmentation algorithm is the maximum entropy hidden Markov model, which is combined with the preprocessing of numerical data, so the weight parameter is introduced:

[0058] ,

[0059] Among them, x represents the input sequence, y represents the labeled sequence, is the characteristic function, is the corresponding weight, is the normalization factor, is the weight parameter. By adjusting By changing the value of , we can control the importance of the MEMM model in the overall preprocessing. =1, it is equivalent to the pure MEMM model; when =0, the MEMM model is completely abandoned.

[0060] Furthermore, we combine standardization and min-max normalization and introduce a weight parameter to balance between the two methods:

[0061] ,

[0062] Among them, x represents the original data, and are the mean and standard deviation of the data, respectively. is the standardized data. = 1, which is equivalent to pure minimum-maximum normalization. = 0, which is equivalent to pure standardization. There is a balance between these two approaches.

[0063] For missing values ​​and data that need format conversion, two functions are introduced and ,in, It is a missing value filling function, which is implemented by decision tree algorithm in different situations. is a format conversion function, which is used to convert data into a unified format and introduce weight parameters for control. =1 only fills in missing values. =0 only performs format conversion:

[0064] ,

[0065] Combining the above three parts, we get a unified preprocessing:

[0066] ,

[0067] Integrate the preprocessing of different types of data into a unified program to avoid the tediousness of separate processing, adjust the respective weight parameters, and adjust the importance of different preprocessing operations for different situations. Through this part of the logical processing, the format of inconsistent and missing data is unified and supplemented. The above formula 1 and formula 2 are parallel logical relationships, and formula 3 is based on formula 1 and formula 2 to uniformly complete the missing and format unification.

[0068] S2: constructing a text feature vector representation of the maintenance briefing content and the work ticket, and calculating a text similarity score between the maintenance briefing content and the work ticket;

[0069] Traverse all the preprocessed maintenance instructions and work ticket safety measures text data, extract all the different words that appear in these texts, remove common stop words, optionally perform word form restoration, and form a vocabulary with the remaining word set. For example, suppose the vocabulary we extract from the text is: {"maintenance", "work ticket", "safety", "measures", "attention", "time", "procedures",...}.

[0070] Furthermore, for each text, the frequency of each word in the vocabulary in the text is counted. Usually, a one-hot vector corresponding to the word frequency value is used to represent each text. For example, if a text is "Pay attention to safety measures when inspecting work tickets" and the vocabulary size is 1000, the one-hot word frequency vector corresponding to the text may be: [1,1,1,1,1,0,0,...,0] (1000 dimensions).

[0071] Furthermore, we calculate TF-IDF to reduce the weight of common words and increase the weight of rare words, where TF is the word frequency calculated in the previous step and IDF is the inverse document frequency (rare words have higher IDF values). The specific calculation formula is:

[0072] ,

[0073] Among them, w is a word, d is a text, N is the total number of texts, and n(w) is the number of texts containing w.

[0074] It should be noted that, taking a word w in text d as an example, if TF(w,d)=2, N=10000, n(w)=100, then TF-IDF(w,d) = 2 * log(10000 / 100) = 2 * 2 = 4.

[0075] Furthermore, the TF-IDF value of each word is directly substituted into the previous one-hot vector, where the m at the position of the one-hot vector subscript n represents the frequency of the corresponding word in the text d, and the format of the weighted Bag-of-Words feature vector is obtained:

[0076] [ tfidf _ 1 , tfidf _ 2 , … , tfidf _ V ] ,

[0077] Where V represents the vocabulary size and tfidf_i represents the TF-IDF value of the i-th word in the text.

[0078] Through the above three steps, the Bag-of-Words model was successfully used to construct the initial feature vector representation from the maintenance instructions and work ticket text data. These feature vectors will encode the weighted frequency information of different words in each text, laying the foundation for the subsequent text similarity calculation.

[0079] Furthermore, the pre-trained Word Embedding model is used to traverse each word token in the maintenance briefing and work ticket text, and the corresponding API of the pre-trained model is used to find the word vector corresponding to the word.

[0080] If the pre-trained Word Embedding model cannot fully adapt to the domain corpus of maintenance instructions and work tickets, the model can be further fine-tuned using the corpus data of the present invention (such as historical maintenance records, etc.) on the basis of the pre-trained model to obtain domain-specific word vector representations. The obtained word vectors are then concatenated with the feature vectors to obtain the corresponding text feature vectors.

[0081] Preferably, the maintenance briefing text is preprocessed as described above to obtain the maintenance briefing text feature vector A, and for each work ticket, the text content of its safety measures part is obtained, and the feature vector B of the safety measures text of the work ticket is constructed in the same way as the maintenance briefing text.

[0082] Initialize Bag-of-Words and Word Embedding text vector representations:

[0083] ,

[0084] Compute the cosine similarity between two vector representations:

[0085] ,

[0086] In order to measure the difference between two vectors, the L1 distance term is introduced to scale the distance:

[0087] ,

[0088] Furthermore, when the semantic similarity is high, the contribution of the distance term will be amplified; when the semantic similarity is low, the contribution of the distance term will be compressed. In order to better balance the relationship between the distance term and the similarity term, the Sigmoid function is introduced, taking the semantic similarity as its input:

[0089] ,

[0090] Finally, the three different information sources are combined in a weighted average manner to calculate a more accurate and reliable text similarity score:

[0091] S ( A → , B → ) = 1 2 [ A → bow ⋅ B → bow | A → bow || B → bow | + a A → emb ⋅ B → emb | A → emb || B → emb | + β 2 | A → bow − B → bow | 1 | A → bow | 1 + | B → bow | 1 ⋅ 1 1 + e − γ ( A → bow ⋅ B → bow ) ] ,

[0092] Among them, a is the weight parameter that controls the contribution of the semantic vector, is the weight parameter controlling the L1 distance term, where To check the text feature vector, To check the word vector, is the feature vector of the work ticket text, is the word vector of the work ticket, and Represents the number of words in the maintenance briefing and work ticket, is the weight parameter.

[0093] The similarity scores of all instructions and work ticket safety measures are calculated to obtain the score distribution. Multiple different thresholds are set respectively. For each threshold, the number of work tickets above the threshold and the number of work tickets that are actually affected are counted.

[0094] Among them, the calculated precision rate = true number of affected cases / total number of cases above the threshold, and the recall rate = true number of affected cases / all affected cases. Check the precision rate and recall rate under different thresholds and select the appropriate threshold as the final threshold.

[0095] Preferably, for each threshold value, define:

[0096] ,

[0097] Then, the precision and recall are:

[0098] ,

[0099] Select the optimal threshold t that balances precision and recall.

[0100] S3: Identify potentially affected work tickets based on the text similarity score, and send the work ticket information to the person in charge.

[0101] Furthermore, a list of potentially affected tickets and a list of unaffected tickets are initialized to temporarily store potentially affected tickets and unaffected tickets related to the current briefing, and each work ticket safety measure is traversed to calculate the text similarity score of the maintenance briefing text feature vector and the feature vector of the work ticket safety measure text.

[0102] If the text similarity score is higher than the preset threshold, the detailed information of the current work ticket is obtained and the current work ticket is added to the list of potentially affected tickets. If the text similarity score is lower than the preset threshold, the current work ticket is placed in the list of unaffected tickets. If the list of potentially affected tickets is not empty, each worksheet in the list of potentially affected tickets is traversed, the similarity score corresponding to it is recalculated, and the worksheets related to the highest similarity score are packaged and sent to the person in charge. After receiving the information, the person in charge checks the specific content to confirm whether it is actually affected by this maintenance. If affected, no other operation is required. If not affected, feedback is given to the system to optimize the similarity model. If the recalculated similarity score is lower than the preset threshold, the relevant worksheet is added to the list of unaffected tickets.

[0103] Furthermore, the system will send the affected maintenance instructions to the person in charge via text message or instant messaging software.

[0104] Furthermore, this embodiment also provides a system for automatically searching for related work tickets according to maintenance instructions, including:

[0105] Data acquisition module, used to obtain maintenance instructions and work ticket information, and perform pre-processing;

[0106] Feature construction module, used to construct text feature vector representation of maintenance instructions and work ticket information;

[0107] A similarity calculation module is used to calculate the text similarity score between the maintenance instructions and each work ticket information;

[0108] A potential ticket identification module is used to identify potentially affected work ticket information according to the similarity score and send the affected work ticket information to the person in charge.

[0109] This embodiment also provides a computer device, which is suitable for the method of automatically searching for related work tickets based on maintenance instructions, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the method of automatically searching for related work tickets based on maintenance instructions proposed in the above embodiment.

[0110] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used 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 and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0111] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method proposed in the above embodiment for automatically searching for relevant work tickets based on maintenance instructions is implemented.

[0112] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0113] Example 2

[0114] Reference Figure 1 , which is the second embodiment of the present invention, and this embodiment provides a method for automatically searching for related work tickets according to maintenance instructions. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0115] To verify the effectiveness of the invented method, the researchers selected data from the production information management system of a petrochemical enterprise for testing. First, 100 maintenance instructions and 1,000 work ticket information currently in execution were extracted from the system database. The extracted maintenance instructions mainly involved equipment maintenance, replacement, maintenance, etc., with an average of about 200 words; the work tickets covered various types of work activities at the production site, with an average of about 150 words.

[0116] For text data, the researchers used a MEMM-based word segmentation algorithm to segment the text into word sequences; for numerical data such as temperature and pressure, the z-score standardization method was used for normalization. In the text preprocessing process, according to the formula:

[0117] By adjusting the values ​​of weight parameters α and β, the MEMM word segmentation algorithm and the standardization method are integrated. When α=0.7 and β=0.3, the two methods can be balanced to achieve the best preprocessing effect.

[0118] In addition, for missing values, a missing value filling function f based on a decision tree is introduced; for data with inconsistent formats, a format conversion function g is used to unify them into the same format. The importance of the two functions is controlled by the weight parameter γ. When γ=0.6, a more reasonable filling and conversion result can be obtained.

[0119] After preprocessing, the researchers built a vocabulary from the maintenance instructions and work ticket texts, totaling about 5,000 words. Next, the TF-IDF method was used to convert each text into a weighted bag-of-words vector with a dimension of 5,000. At the same time, the pre-trained Word2Vec model was used to obtain a 100-dimensional word vector representation for each word. The bag-of-words vector and the word vector were concatenated to obtain the final 5,100-dimensional text feature vector.

[0120] According to the above process, the text similarity of 100 maintenance instructions and 1000 work tickets was calculated. Five different thresholds of 0.5, 0.6, 0.7, 0.8, and 0.9 were set respectively, and the number of actual affected tickets, the total number of potentially affected tickets, and the precision and recall rates under each threshold were counted. The experimental results are recorded in the following table:

[0121] Table 1 Experimental data table

[0122] ,

[0123] As the threshold increases, precision increases and recall decreases. This is consistent with the trade-off relationship between precision and recall. When the threshold is set to 0.5, the recall reaches the highest value of 0.2778, but the precision is only 0.0704, which means that a large number of irrelevant work tickets are mistakenly marked as affected tickets. When the threshold is increased to 0.9, the precision reaches 1.0, that is, all tickets marked as affected are truly affected tickets, but correspondingly, the recall is only 0.2778, and 27.78% of the affected tickets are missed.

[0124] It is worth mentioning that the method of the present invention combines multiple information sources such as semantic vectors and structural distances to calculate text similarity, rather than relying solely on literal matching. This allows maintenance instructions and work tickets to be accurately identified even if they use industry-specific terms or abbreviations. In addition, the integration of the MEMM word segmentation algorithm with the numerical standardization method to perform unified domain-adaptive preprocessing of text and numerical data is also an innovation of this method.

[0125] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A method for automatically searching for relevant work tickets according to maintenance instructions, characterized in that: include, Obtaining maintenance instructions and work ticket information, and preprocessing the maintenance instructions and work ticket information; Constructing a text feature vector representation of the maintenance briefing content and the work ticket, and calculating a text similarity score between the maintenance briefing content and the work ticket; identifying potentially affected work tickets based on the text similarity scores and sending the work ticket information to the responsible person; Constructing the text feature vector representation of the maintenance instructions and work tickets includes the following steps: Extract the pre-processed maintenance instructions and worksheet information to form a vocabulary list; For each text, counting the frequency of occurrence of the vocabulary to form a word frequency vector; Constructing the text feature vector representation includes: Pre-process the maintenance instructions and work ticket information; Build vocabulary for maintenance briefing and work ticket information; Calculate the frequency of each text vocabulary entry; Calculating a weighted value of the word frequency; constructing a feature vector representation according to the weighted values; Obtain a word vector according to the word vector model and concatenate it with the feature vector to form a final text vector; The weighted values ​​are calculated to reduce the weight of common words and increase the weight of rare words, as shown in the following formula: Where w is a word, d is a text, N is the total number of texts, and n(w) is the number of texts containing w; Calculating the text similarity score between the maintenance instructions and the work ticket includes: Initialize two text vector representations: Bag-of-Words and Word Embedding; Compute the cosine similarity between two vector representations; The distance term is used to scale the distance between two text vectors, and the semantic similarity term is used to balance the relationship between the distance term and the similarity term; Combining the cosine similarity, the distance term and the semantic similarity term to obtain the text similarity score; The cosine similarity between two vector representations is calculated as follows: The distance term is expressed as follows: The semantically similar terms are shown in the following formula: The text similarity score is calculated as follows: Among them, a is the weight parameter that controls the contribution of the semantic vector, is the weight parameter controlling the L1 distance term, where To check the text feature vector, To check the word vector, is the feature vector of the work ticket text, is the word vector of the work ticket, and Represents the number of words in the maintenance briefing and work ticket, is the weight parameter.

2. The method for automatically searching for relevant work tickets according to maintenance instructions as claimed in claim 1, characterized in that: The preprocessing adopts segmented logic processing, and performs different preprocessing operations on different data, as shown in the following formula: Among them, x represents the input sequence, y represents the labeled sequence, is the characteristic function, is the corresponding weight, is the normalization factor, is the weight parameter for text data preprocessing, and are the mean and standard deviation of the data, respectively. Fill in missing values. is the format conversion function, Weight parameters for missing values ​​and data that need format conversion.

3. The method for automatically searching for relevant work tickets according to maintenance instructions as claimed in claim 2, characterized in that: Potentially affected work tickets identified based on the text similarity scores include, If the text similarity score is higher than the preset threshold, the detailed information of the current work ticket is obtained and the current work ticket is added to the list of potentially affected tickets. If the text similarity score is lower than the preset threshold, the current work ticket is placed in the list of unaffected tickets. If the list of potentially affected tickets is not empty, each worksheet in the list of potentially affected tickets is traversed, the similarity score corresponding to it is recalculated, and the worksheets related to the highest similarity score are packaged and sent to the person in charge. After receiving the information, the person in charge checks the specific content to confirm whether it is actually affected by this maintenance. If affected, no other operation is required. If not affected, feedback is given to the system to optimize the similarity model. If the recalculated similarity score is lower than the preset threshold, the relevant worksheet is added to the list of unaffected tickets.

4. The method for automatically searching for related work tickets according to maintenance instructions as claimed in claim 3, characterized in that: The selection of the threshold includes calculating the similarity scores of all the confession contents and the work ticket safety measures, obtaining the score distribution, and setting a plurality of different thresholds respectively. For each threshold, the number of work tickets above the threshold and the number of work tickets actually affected are counted; the precision and recall rates under different thresholds are calculated, and the threshold that balances the precision and recall rates is selected as the final threshold; The calculation of precision and recall is as follows: in, .

5. A system for automatically searching for related work tickets according to maintenance instructions, based on the method for automatically searching for related work tickets according to maintenance instructions according to any one of claims 1 to 4, characterized in that: include, Data acquisition module, used to obtain maintenance instructions and work ticket information, and perform pre-processing; Feature construction module, used to construct text feature vector representation of maintenance instructions and work ticket information; A similarity calculation module is used to calculate the text similarity score between the maintenance instructions and each work ticket information; A potential ticket identification module is used to identify potentially affected work ticket information according to the similarity score and send the affected work ticket information to the person in charge.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for automatically searching for related work tickets based on maintenance instructions as described in any one of claims 1 to 4 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the method for automatically searching for relevant work tickets according to maintenance instructions as described in any one of claims 1 to 4 are implemented.

Images