An elevator fault diagnosis and handling system and method

Abstract

The elevator fault diagnosis and handling system and method of this invention, through comprehensive analysis of multimodal data, combined with a large elevator model and knowledge base, and an AI model, uses a causal rule reasoning model to dynamically match new faults, generate and rank solutions, and automatically generate a visual fault navigation for the optimal solution. Based on user rating data, it is optimized and improved, significantly enhancing maintenance efficiency and quality. It has the following advantages: short fault response time, greatly reducing the time from fault occurrence to solution generation; high first-time handling success rate, as the causal rule reasoning model dynamically matches new faults and automatically generates a visual fault navigation for the optimal solution, significantly improving the fault handling success rate, especially the first-time fault handling success rate; and user participation in optimization, driving continuous improvement in service quality through a rating mechanism.

Description

Technical Field

[0002] The present invention relates to the field of elevator maintenance technology, and in particular to an elevator fault diagnosis and treatment system and method. Background Technology

[0004] With the continuous increase in the number of elevator maintenance personnel, the high turnover rate and varying skill levels of maintenance staff mean that troubleshooting based on personal experience, consulting fault manuals, or requesting remote assistance is often time-consuming and laborious. Analyzing the causes of faults and finding solutions takes a considerable amount of time, resulting in low efficiency and prolonged troubleshooting times, which affects customers' use of elevators. The main problems include:

[0005] Reliance on human experience: Maintenance personnel need to rely on personal experience or consult paper / electronic fault manuals, which is inefficient and easily affected by the technical level and turnover rate of personnel.

[0006] Inefficient remote assistance: Frequent requests for remote expert assistance during troubleshooting increase response time and reduce service timeliness.

[0007] Data silo problem: Elevator operating status, fault codes, video behavior and other data are scattered and lack a unified analysis logic, making it difficult to quickly locate the root cause of the fault.

[0008] The solution lacks versatility: existing methods are poorly adaptable to complex or atypical faults and cannot dynamically optimize the handling strategy. Summary of the Invention

[0010] The elevator fault diagnosis and handling system provided in this embodiment of the invention includes:

[0011] AI client C1, mobile phone A1, elevator E1, DTU terminal T1, elevator knowledge base system Y1, elevator AI large model X1, camera V1.

[0012] The configuration method of the elevator fault diagnosis and handling system provided in this embodiment of the invention includes:

[0013] Elevator E1 and camera V1 upload multimodal data to elevator knowledge base system Y1.

[0014] The elevator knowledge base system Y1 transmits the received multimodal data and stored elevator knowledge base system data to the elevator AI big model X1.

[0015] The elevator AI big data model X1 will preprocess and train the received multimodal data and elevator knowledge base data.

[0016] User P1 installs client C1 via mobile phone A1.

[0017] The specific steps and procedures described above can be adjusted.

[0018] In the elevator fault diagnosis and handling system provided in this embodiment of the invention, when elevator E1 malfunctions, the handling process includes:

[0019] S1: Fault information and other information are transmitted from the elevator knowledge base system Y1 to the elevator AI big model X1. The elevator AI big model X1 uses a causal rule reasoning model to dynamically match new faults with rules, combines the received multimodal data, generates and sorts solutions, and automatically generates a visual fault navigation for the optimal solution. Other information includes other relevant information used to assist decision-making besides fault information, such as elevator status data, human feedback data, etc. Fault information includes direct fault information, such as fault codes, fault status data, etc.

[0020] S2: The elevator AI big model X1 transmits relevant information to the elevator knowledge base system Y1. The relevant information includes fault information, fault causes, solutions, and visual fault navigation. The elevator knowledge base system Y1 then pushes this information to the AI ​​client C1.

[0021] S3: After receiving the relevant information pushed to the AI ​​client C1, the AI ​​client engine is automatically started to further analyze and confirm the fault information, and the optimal solution and its visual fault navigation are displayed to the maintenance personnel P1 for confirmation.

[0022] S4: After the maintenance personnel P1 confirm, they execute the visual fault navigation process through the AI ​​client C1, and complete the fault handling W1 task through voice recognition, intelligent question and answer, intelligent search and intelligent recommendation.

[0023] S5: After the task is completed, the maintenance personnel P1 scores the handling process and forms a case. The mobile phone A1 automatically uploads the fault case U4 to the elevator knowledge base system Y1 and saves it, and then transfers it to the elevator AI big model X1.

[0024] S6: The elevator AI large model X1 optimizes solution recommendations, retrains the causal rule reasoning model, and dynamically updates the elevator knowledge base system.

[0025] The elevator fault diagnosis and handling method provided in this embodiment of the invention includes:

[0026] S101: Multimodal data acquisition.

[0027] In this embodiment, the multimodal data includes: elevator status data, fault code data, video data, historical case data, etc.

[0028] S102: Construct an elevator knowledge base system.

[0029] An elevator knowledge base system is constructed based on the collected and processed data. In this embodiment, the elevator knowledge base may include a structured knowledge base and an unstructured knowledge base. The structured knowledge base can store standardized fault-to-solution mapping relationships, while the unstructured knowledge base can store text data such as maintenance manuals, technical documents, and expert experience, which are then transformed into searchable semantic vectors using NLP technology.

[0030] S103: Multimodal data and elevator knowledge base system data preprocessing.

[0031] In this embodiment, data preprocessing includes:

[0032] S1031: Elevator status data cleaning. Used to remove invalid or missing data, including abnormal data such as data collected during sensor offline periods, data collected under abnormal sensor conditions, and data lost during collection.

[0033] S1032: Video Data Feature Extraction. Used to extract behavioral features from video data, including spatiotemporal features acquired through methods such as motion trajectory and keyframe recognition, and time-frequency domain features obtained through video analysis. These can be extracted using existing video feature extraction methods, which will not be elaborated here.

[0034] S1033: Data Normalization. This is used to unify data from different sources into a standard format for subsequent data use and analysis. It can be accomplished through data analysis or machine learning methods such as mapping fault codes to a unified coding table and standardizing structured data, which will not be elaborated here. The data includes multimodal data and elevator knowledge base system data.

[0035] S1034: Multimodal data alignment. Used to associate different modal data streams by timestamps and automatically construct multidimensional feature vectors in the time domain.

[0036] S104: Fault diagnosis and recommendation model construction: causal rule reasoning model is used to dynamically match new faults with rules.

[0037] In this embodiment, constructing a causal rule-based reasoning model includes:

[0038] S1041: Automatically label the causes and solutions of failures based on historical case data to form a causal relationship map.

[0039] S1042: Construct a causal rule reasoning model based on causal relationship graph.

[0040] Causal relationship graphs can perform fault cause analysis and are a preliminary stage for causal rule reasoning models to process data. Thus, solutions can be derived through causal rule reasoning models. When a new fault occurs, the system completes fault analysis through dynamic rule matching.

[0041] In this embodiment, dynamic rule matching includes:

[0042] S1043: Preliminary fault code identification. Match the fault code with data in the elevator knowledge base to generate a preliminary fault category.

[0043] S1044: Multimodal data fusion analysis of fault occurrence probability. The processed multimodal data includes abnormal operating conditions (such as motor overload), video anomalies (such as wire rope misalignment), and historical similar cases, etc. The probability of fault occurrence is calculated and analyzed based on the multimodal data.

[0044] S1045: Causal rule activation to locate the root cause of the fault. Dynamically activate the rules in the corresponding causal rule reasoning model, and match and output the most likely cause of the fault and its confidence level.

[0045] S105: Solution sorting and automatic generation of visual fault navigation steps.

[0046] In this embodiment, it includes:

[0047] S1051: Solution matching priority ranking. Based on the root cause of the failure, possible solutions are ranked according to multiple dimensions such as confidence level, historical processing success rate, and user rating.

[0048] S1052: Optimal solution automatically generates visual fault navigation steps. The optimal solution is broken down into a visual fault navigation step operation process.

[0049] S106: User rating feedback.

[0050] S107: Solution recommendation optimization, causal rule reasoning model retraining, and dynamic updating of the elevator knowledge base system.

[0051] If the score is high, adjust the solution based on the user rating, increase the weight of the solution, and raise the priority of the solution in similar faults.

[0052] If the score is low, the causal rule reasoning model is retrained, the weights of the causal rules are adjusted, historical failure cases with expert participation and scoring are called up, and the high-scoring cases are selected to regenerate the processing flow.

[0053] The elevator fault diagnosis and handling system and method of this invention, through comprehensive analysis of multimodal data and combined with a large elevator model and knowledge base, achieves automatic judgment and accurate location of fault causes, replacing the manual analysis process by experts. The system uses an AI large model and a causal rule reasoning model to dynamically match new faults, generate and prioritize solutions, and automatically generate a visual fault navigation based on the optimal solution. Based on user rating data, it is optimized and improved, significantly improving maintenance efficiency and quality. It has the following advantages: short fault response time, which can greatly reduce the time from fault occurrence to solution generation; high first-time handling success rate, which greatly improves the fault handling success rate, especially the first-time fault handling success rate, by using a causal rule reasoning model to dynamically match new faults and automatically generate a visual fault navigation based on the optimal solution; and user participation in optimization, which can drive continuous improvement of service quality through a rating mechanism. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the elevator fault diagnosis and handling system provided in an embodiment of the present invention;

[0056] Figure 2 A flowchart illustrating a method for diagnosing and handling elevator faults according to another embodiment of the present invention; Detailed Implementation

[0058] The technical solution of the present invention will be described in detail below with reference to specific embodiments and accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention and not the entire structure.

[0059] Figure 1 This is a schematic diagram of an elevator fault diagnosis and handling system shown in Embodiment 1 of this application, including the following: AI client C1, mobile phone A1, elevator E1, DTU terminal T1, elevator knowledge base system Y1, elevator AI large model X1, and camera V1.

[0060] In this embodiment, mobile phone A1 has a built-in AI client C1. Mobile phone A1 can be other electronic terminals such as mobile phones, tablets, and PCs. The elevator knowledge base system Y1 includes a communication module Z1, a database module B1, and a service module S1. The elevator AI big model X1 includes a status and fault model M1, a video behavior analysis model M2, a speech recognition model, and a fault diagnosis and recommendation model M4.

[0061] The system configuration methods include:

[0062] Elevator E1 and camera V1 upload multimodal data to elevator knowledge base system Y1. In this embodiment, the multimodal data includes: elevator status data, fault code data, video data, historical case data, etc.

[0063] Elevator status data includes: elevator operating parameters collected by sensors, such as elevator speed, acceleration, voltage during startup, operation, or stationary states, current during startup, operation, or stationary states, and overall or component temperatures; fault code data includes: fault code data obtained by parsing fault codes output by the elevator control system; video data includes: video image data obtained by collecting visual image information such as vibration capture, foreign object jamming, personnel operation in the car, personnel operation in the machine room, and shaft conditions or personnel operation from intelligent cameras or image sensors deployed in the elevator car, machine room, and components; historical case data includes: structured data obtained by integrating historical maintenance records, fault types, fault repair time, and user feedback on fault repairs, as well as unstructured data obtained by integrating maintenance reports and on-site maintenance photos.

[0064] The elevator knowledge base system Y1 transmits the received multimodal data and stored elevator knowledge base system data to the elevator AI big model X1.

[0065] The elevator knowledge base system includes both structured and unstructured data. The structured knowledge base can include standardized fault-solution mappings, such as fault codes combined with features, showing fault causes and repair procedures. The unstructured knowledge base can include textual data such as maintenance manuals, technical documents, and expert experience, which are transformed into searchable semantic vectors using NLP technology.

[0066] The elevator AI big data model X1 will preprocess and train the received multimodal data and elevator knowledge base data.

[0067] In this embodiment, data preprocessing may include: elevator status data cleaning, video data feature extraction, data normalization, and multimodal data alignment.

[0068] Specifically, elevator status data cleaning is used to remove invalid or missing data, including abnormal data such as data collected during sensor offline periods, data collected during abnormal sensor states, and data lost during collection. Video data feature extraction is used to extract behavioral features from video data, including spatiotemporal features collected through methods such as motion trajectory and keyframe recognition, and time-frequency domain features obtained through video analysis. These can be achieved using existing video feature extraction methods, which will not be elaborated here. Data normalization is used to unify data from different sources into a standard format for subsequent data use and analysis. This can be accomplished through data analysis or machine learning methods such as mapping fault codes to a unified encoding table and standardizing structured data, which will not be elaborated here. The data includes multimodal data and elevator knowledge base system data. Multimodal data alignment is used to associate different modal data streams through timestamps and automatically construct multidimensional feature vectors in the time domain. In this embodiment, data training can be performed using machine learning algorithms, deep learning models, big data processing technologies, knowledge graphs, etc.

[0069] User P1 installs client C1 via mobile phone A1.

[0070] Specifically, the user can be maintenance personnel or other relevant personnel or users; the mobile phone A1 can be a mobile phone, tablet, PC, or other electronic terminal. The client C1 has a built-in AI engine, including an MCP client, an MCP server, and an AI client LLM model, which has AI functions such as speech recognition, video analysis and playback, intelligent question answering, intelligent retrieval, and intelligent recommendation.

[0071] The specific steps and procedures described above can be adjusted.

[0072] In the system of this embodiment, when elevator E1 malfunctions, the processing flow includes:

[0073] S1: Fault information and other information are transmitted from the elevator knowledge base system Y1 to the elevator AI big model X1. The elevator AI big model X1 uses a causal rule reasoning model to dynamically match new faults with rules, combines the received multimodal data, generates and sorts solutions, and automatically generates a visual fault navigation for the optimal solution. Other information includes other relevant information used to assist decision-making besides fault information, such as elevator status data, human feedback data, etc. Fault information includes direct fault information, such as fault codes, fault status data, etc.

[0074] In this embodiment, maintenance personnel P1 can take and upload images of components using the AI ​​client C1 on mobile phone A1, and the system can use image recognition to assist in obtaining information. Maintenance personnel P1 can also input natural language questions, such as "The elevator suddenly stopped and displayed E12 code," using natural language processing to assist in obtaining feedback data.

[0075] S2: The elevator AI big model X1 transmits relevant information to the elevator knowledge base system Y1. The relevant information includes fault information, fault causes, solutions, and visual fault navigation. The elevator knowledge base system Y1 then pushes this information to the AI ​​client C1.

[0076] S3: After receiving the relevant information pushed to the AI ​​client C1, the AI ​​client engine is automatically started to further analyze and confirm the fault information, and the optimal solution and its visual fault navigation are displayed to the maintenance personnel P1 for confirmation.

[0077] S4: After the maintenance personnel P1 confirm, they execute the visual fault navigation process through the AI ​​client C1, and complete the fault handling W1 task through voice recognition, intelligent question and answer, intelligent search and intelligent recommendation.

[0078] In this embodiment, maintenance personnel P1 can interact with the AI ​​client C1 on mobile phone A1. The AI ​​client C1 can provide prompts, supplements, or inquiries to maintenance personnel P1, or the maintenance personnel can proactively supplement relevant information.

[0079] Among them, the AI ​​client C1 in mobile phone A1 performs real-time analysis of the video stream and prompts abnormal operations, such as incorrect component replacement order.

[0080] S5: After the task is completed, the maintenance personnel P1 scores the handling process and forms a case. The mobile phone A1 automatically uploads the fault case U4 to the elevator knowledge base system Y1 and saves it, and then transfers it to the elevator AI big model X1.

[0081] S6: The elevator AI large model X1 optimizes solution recommendations, retrains the causal rule reasoning model, and dynamically updates the elevator knowledge base system.

[0082] In this embodiment, mobile phone A1 can be a mobile phone, tablet computer, PC or other electronic terminal, and maintenance personnel P1 can be elevator, property management, user or other relevant personnel.

[0083] The elevator fault diagnosis and handling system described in this embodiment achieves automatic judgment and accurate location of fault causes by comprehensively analyzing multimodal data and combining a large elevator model and knowledge base. This replaces the manual analysis process by experts. The system uses an AI large model and a causal rule reasoning model to dynamically match new faults, generate solutions and sort them. The optimal solution is automatically generated into a visual fault navigation logic, which is optimized and improved based on user rating data, significantly improving maintenance efficiency and quality.

[0084] The system described in this embodiment has the following advantages:

[0085] Short fault response time can significantly reduce the time from fault occurrence to solution generation; high first-time success rate, through causal rule reasoning model, dynamic rule matching is performed on new faults, and the solution is automatically generated to generate a visual fault navigation logic, which greatly improves the fault handling success rate, especially the first-time fault handling success rate; user participation optimization can be achieved, and the service quality can be continuously improved through a scoring mechanism.

[0086] Figure 2 This is a flowchart illustrating an elevator fault diagnosis and handling method according to Embodiment 2 of this application, which specifically includes the following steps:

[0087] S101: Multimodal data acquisition.

[0088] In this embodiment, the multimodal data includes: elevator status data, fault code data, video data, historical case data, etc.

[0089] Elevator status data includes: elevator operating parameters collected by sensors, such as elevator speed, acceleration, voltage during startup, operation, or stationary states, current during startup, operation, or stationary states, and overall or component temperatures; fault code data includes: fault code data obtained by parsing fault codes output by the elevator control system; video data includes: video image data obtained by collecting visual image information such as vibration capture, foreign object jamming, personnel operation in the car, personnel operation in the machine room, and shaft conditions or personnel operation from intelligent cameras or image sensors deployed in the elevator car, machine room, and components; historical case data includes: structured data obtained by integrating historical maintenance records, fault types, fault repair time, and user feedback on fault repairs, as well as unstructured data obtained by integrating maintenance reports and on-site maintenance photos.

[0090] S102: Construct an elevator knowledge base system.

[0091] An elevator knowledge base system is constructed based on the collected and processed data. In this embodiment, the knowledge base may include a structured knowledge base and an unstructured knowledge base.

[0092] A structured knowledge base can store standardized fault-to-solution mappings, such as fault codes and feature combinations, which include fault causes and repair steps.

[0093] Unstructured knowledge bases can store text data such as maintenance manuals, technical documents, and expert experience, which can be transformed into searchable semantic vectors through NLP technology.

[0094] S103: Multimodal data and elevator knowledge base system data preprocessing.

[0095] In this embodiment, data preprocessing includes:

[0096] S1031: Elevator status data cleaning.

[0097] Used to remove invalid or missing data, which includes abnormal data, such as data collected during sensor offline periods, data collected under abnormal sensor conditions, and data lost during collection.

[0098] S1032: Feature extraction from video data.

[0099] This is used to extract behavioral features from video data, including spatiotemporal features collected through methods such as motion trajectory and keyframe recognition, and time-frequency domain features obtained through video analysis. These can be achieved using existing video feature extraction methods, which will not be elaborated here.

[0100] S1033: Data normalization.

[0101] This is used to unify data from different sources into a standard format for subsequent data use and analysis. This can be accomplished through data analysis or machine learning methods such as mapping fault codes to a unified coding table and standardizing structured data, which will not be elaborated here. The data includes multimodal data and elevator knowledge base system data.

[0102] S1034: Multimodal data alignment.

[0103] It is used to associate different modal data streams by timestamp and automatically construct multidimensional feature vectors in the time domain.

[0104] S104: Fault diagnosis and recommendation model construction: causal rule reasoning model is used to dynamically match new faults with rules.

[0105] Specifically, models can be built using multimodal data and knowledge base system data, including machine learning algorithms, deep learning models, big data processing technologies, and knowledge graphs.

[0106] In this embodiment, constructing a causal rule-based reasoning model includes:

[0107] S1041: Automatically label the causes and solutions of failures based on historical case data to form a causal relationship map.

[0108] S1042: Construct a causal rule reasoning model based on causal relationship graph.

[0109] Specifically, methods such as Bayesian networks or graph neural networks (GNNs) can be used to construct a causal rule inference model between fault codes, operating status, and video behavior.

[0110] Causal relationship graphs can perform fault cause analysis and are a preliminary stage for causal rule reasoning models to process data. Thus, solutions can be derived through causal rule reasoning models. When a new fault occurs, the system completes fault analysis through dynamic rule matching.

[0111] In this embodiment, dynamic rule matching includes:

[0112] S1043: Preliminary fault code identification. Match the fault code with data in the elevator knowledge base to generate a preliminary fault category.

[0113] S1044: Multimodal data fusion analysis of fault occurrence probability. The processed multimodal data includes abnormal operating conditions (such as motor overload), video anomalies (such as wire rope misalignment), and historical similar cases, etc. The probability of fault occurrence is calculated and analyzed based on the multimodal data.

[0114] S1045: Causal rule activation to locate the root cause of the fault. Dynamically activate the rules in the corresponding causal rule reasoning model, and match and output the most likely cause of the fault and its confidence level.

[0115] The rules include: fault code X + voltage fluctuation characteristics + abnormal noise detected in the car, activating the traction machine bearing fault determination rule; fault code Y + abnormal elevator acceleration + abnormal vibration capture, activating the wire rope fault rule; etc.

[0116] In this embodiment, the fault location of the traction machine bearing is determined by fault code X, the abnormal noise detected in the car is determined by video (weight: 80), and the voltage fluctuation characteristics are determined by voltage fluctuation characteristics (weight: 15, the voltage shows a regular increase every hour). Based on these factors, it is concluded that there is a 95% probability that the fault is caused by insufficient lubrication of the traction machine.

[0117] S105: Solution sorting and automatic generation of visual fault navigation steps.

[0118] Based on the cause of the fault, possible solutions are generated. By automatically generating navigation faults, the optimal solution is broken down into a visual fault navigation step operation process.

[0119] In this embodiment, it includes:

[0120] S1051: Solution matching priority ranking. Based on the root cause of the failure, possible solutions are ranked according to multiple dimensions such as confidence level, historical success rate, and user rating. Specifically, ranking based on multiple dimensions such as confidence level, historical success rate, and user rating can be achieved through various methods such as manual algorithm settings, machine learning, and data analysis.

[0121] S1052: Automatically generate visual fault navigation steps. Break down the optimal solution into a visual fault navigation step operation process.

[0122] In this embodiment, the fault rules for the wire rope can be broken down as follows: First step: check the lubrication of the traction machine; second step: use a PDA tool to adjust the tension of the wire rope.

[0123] S106: User Rating Feedback

[0124] After maintenance personnel complete the process according to the visual fault navigation steps, they score the visual fault navigation steps. In this embodiment, after maintenance or other personnel complete the process, they can score it numerically in the system, or they can complete the evaluation score by marking it with "resolved," "partially resolved," "unresolved," or other methods.

[0125] S107: Solution recommendation optimization, causal rule reasoning model retraining, and dynamic updating of the elevator knowledge base system.

[0126] Specifically, if the user rating is high, the solution is adjusted based on the user rating, the weight of the solution is increased, and the priority of the solution in similar faults is improved; if the user rating is low, the system initiates adaptive learning, triggers and adjusts the retraining of the causal rule reasoning model based on the user rating data and processing time, adjusts the weight of the causal rules, calls up historical fault cases with expert participation and ratings, selects high-rated cases to regenerate the processing flow, strengthens the recommendation algorithm, and new cases are automatically added to the knowledge base after system analysis, forming an iterative optimization of the elevator knowledge base system.

[0127] The criteria for high and low scores can be set manually or automatically. For example, in a maximum score of 10, "higher" is greater than 8 points and "lower" is less than 3 points.

[0128] The method described in this embodiment is applied to the system provided in this embodiment of the invention, so that the method described in this embodiment has corresponding beneficial effects.

[0129] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. An elevator fault diagnosis and handling system, characterized in that, This includes AI clients, mobile phones, elevators, DTU terminals, elevator knowledge base systems, large-scale elevator AI models, and cameras.

2. The elevator fault diagnosis and handling system according to claim 1, characterized in that, The system configuration method includes: The elevator and the camera upload multimodal data to the elevator knowledge base system. The elevator knowledge base system transmits the received multimodal data and the stored knowledge base system data to the elevator AI big model. The elevator AI big data model will perform data preprocessing and training on the received multimodal data and elevator knowledge base data. The user installs the AI ​​client on the mobile phone.

3. The elevator fault diagnosis and handling system according to claim 1, characterized in that, When the elevator malfunctions, the handling procedure includes: S1 fault information and other information are transmitted from the elevator knowledge base system to the elevator AI big model. The elevator AI big model uses a causal rule reasoning model to dynamically match new faults, combines the received multimodal data, generates and sorts solutions, and automatically generates a visual fault navigation based on the optimal solution. The elevator AI model described in S2 transmits relevant information to the elevator knowledge base system. After receiving the relevant information pushed to the AI ​​client, the AI ​​client engine described in S3 automatically starts to further analyze and confirm the fault information, and displays the optimal solution and its visual fault navigation to the maintenance personnel for confirmation. After the maintenance personnel in step S4 confirm the issue, they execute the visualized fault navigation process through the AI ​​client to complete the fault handling task. After the task described in S5 is completed, the maintenance personnel will rate the processing procedure. The elevator AI big model described in S6 optimizes the solution recommendation, the causal rule reasoning model is retrained, and the elevator knowledge base system is dynamically updated.

4. A method for diagnosing and handling elevator faults, characterized in that, S101 multimodal data acquisition, S102 constructs an elevator knowledge base system. S103 Multimodal Data and Elevator Knowledge Base System Data Preprocessing The S104 fault diagnosis and recommendation model constructs a causal rule-based reasoning model to dynamically match new faults using rules. S105 solution sorting and automatic generation of visual fault navigation steps S106 user rating feedback, S107 solution recommendation optimization, causal rule reasoning model retraining, and elevator knowledge base system dynamic updates.

5. The elevator fault diagnosis and handling method according to claim 4, characterized in that, The multimodal data acquired in S101 includes: Elevator status data, fault code data, video data, and historical case data.

6. The elevator fault diagnosis and handling method according to claim 4, characterized in that, The knowledge base in the elevator knowledge base system constructed in S102 includes: Structured knowledge bases and unstructured knowledge bases The structured knowledge base can store standardized fault-to-solution mappings. The unstructured knowledge base can store text data such as maintenance manuals, technical documents, and expert experience, which can be transformed into searchable semantic vectors through NLP technology.

7. The elevator fault diagnosis and handling method according to claim 4, characterized in that, The preprocessing of the S103 multimodal data and elevator knowledge base system data includes: S1031 elevator status data cleaning, Feature extraction from S1032 video data. S1033 data normalization, S1034 describes multimodal data alignment.

8. The elevator fault diagnosis and handling method according to claim 4, characterized in that, The S104 fault diagnosis and recommendation model constructs a causal rule reasoning model to perform dynamic rule matching for new faults, including: S1041 automatically labels the causes and solutions to failures based on historical case data, forming a causal relationship graph. S1042 completes the construction of the causal rule reasoning model based on the causal relationship graph. S1043 fault code preliminary identification: The fault code is matched with data in the elevator knowledge base system to generate a preliminary fault category. The multimodal data fusion analysis described in S1044 describes the probability of fault occurrence. S1045 causal rules are activated, locating the root cause of the fault.

9. The elevator fault diagnosis and handling method according to claim 4, characterized in that, The S105 solution sorting and automatic generation of visual fault navigation steps include: The solution described in S1051 matches priority ordering. The S1052 optimal solution automatically generates the visualized fault navigation steps.

10. The elevator fault diagnosis and handling method according to claim 4, characterized in that, The S107 solution recommendation includes optimization, retraining of the causal rule reasoning model, and dynamic updating of the elevator knowledge base system, including: If the score is high, the solution is adjusted based on the user score, increasing the weight of the solution and raising its priority among similar faults. If the score is low, the causal rule reasoning model is retrained, the weights of the causal rules are adjusted, historical failure cases with expert participation and scoring are called up, and the high-scoring cases are selected to regenerate the processing flow.

Images