Elevator troubleshooting method and system

Abstract

The invention discloses an elevator troubleshooting method and system, and belongs to elevator system design. The intelligent level and diagnosis accuracy of elevator troubleshooting are improved through deep learning, a graph neural network and a causal reasoning technology. The system can efficiently process multi-source heterogeneous data, including operation logs, sensor time sequence data and fault declaration records, from an elevator, and pre-processes different data sources through a timestamp alignment algorithm to generate a structured data set. According to the elevator troubleshooting system, through integrated application of the AI technology, the speed and accuracy of fault diagnosis are improved, the continuous self-learning and optimizing capacity is achieved, and the elevator troubleshooting system is suitable for various elevator systems including high-end villa elevators. Through the system, the elevator maintenance process becomes more intelligent and efficient, and powerful guarantee is provided for elevator operation safety.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of elevator system design, and in particular to an elevator fault troubleshooting method and system. Background Art

[0002] In the field of elevator troubleshooting, traditional fault detection systems often rely on manual judgment or simple rule bases, resulting in the system only being able to detect some common faults and a relatively slow diagnostic process. Especially in troubleshooting high-end elevators (such as villa elevators), due to the complexity of the equipment and the diversity of fault types, traditional methods are unable to promptly and accurately identify and address all potential problems. With the increasing intelligence level of elevator equipment, various heterogeneous data such as sensor data, operation logs, and fault reports have become valuable resources for elevator fault diagnosis. However, how to effectively integrate this multi-source heterogeneous data and accurately identify potential fault modes is a major challenge in existing technologies.

[0003] Currently, troubleshooting systems face the following challenges: Elevator operation data comes from diverse sources, including sensor time series data, operation logs, and fault reporting records. Efficiently integrating this heterogeneous data to build an accurate fault diagnosis model is a technical challenge. Elevator failures are not only affected by time but are also closely related to spatial location (the operating status of each elevator component). Traditional methods fail to effectively integrate temporal and spatial features. Failure modes can be complex, nonlinear relationships. Traditional troubleshooting methods based on rules or simple statistical analysis cannot reveal the underlying causes of failures and can easily overlook deeper systemic problems. Existing systems generally lack self-optimization capabilities and are unable to update models based on actual maintenance feedback, resulting in a lack of sustained improvement in the accuracy and efficiency of troubleshooting.

[0004] How to effectively integrate multi-source heterogeneous data, accurately explore potential failure modes in elevator operation, and improve the intelligence and real-time performance of fault detection through deep integration of spatiotemporal features, thereby achieving efficient and accurate elevator fault diagnosis. Summary of the Invention

[0005] The present invention provides an elevator fault troubleshooting method and system. The technical solution is as follows.

[0006] According to one aspect of the present application, there is provided an elevator fault troubleshooting system, the system comprising:

[0007] The data acquisition and preprocessing module is used to receive the elevator's operation log, sensor time series data, and fault reporting records through a computer-executed program. It uses a timestamp alignment algorithm to time-align heterogeneous data from different sources to generate a structured data set that includes sensor data and log data with consistent timestamps.

[0008] The feature fusion and pattern mining module is used to extract features from multi-source data. It uses a convolutional neural network to extract the temporal features of sensor data and a graph convolutional network to model the spatial correlation of elevator components. It then fuses the spatiotemporal feature vectors and outputs them for fault mode mining.

[0009] The fault knowledge graph module is used to query and expand the elevator fault knowledge graph based on spatiotemporal feature vectors, update the nodes and relationships in the fault graph using a graph neural network, and return a set of candidate fault causes, including the confidence level of the fault cause and the associated evidence chain;

[0010] The causal reasoning module is used to analyze the causal chain of the fault based on the candidate fault cause set using a causal reasoning algorithm and generate a list of root cause hypotheses, each of which contains the causal chain of the fault cause and the basis for elimination;

[0011] The solution generation module is used to generate a structured solution with operation steps, safety specifications, and a parts list based on the root cause hypothesis list, and to check the legality and safety of the solution through a compliance verification algorithm;

[0012] The self-optimization module is used to adjust the troubleshooting model based on maintenance personnel feedback data using reinforcement learning algorithms and update the knowledge graph and reasoning model.

[0013] Optionally, the data acquisition and timing alignment module also receives log files and sensor data from different devices and in different formats by executing a data normalization algorithm, and uses a dynamic time warping algorithm to align timestamps of data from different time sources.

[0014] Optionally, the feature fusion and pattern mining module extracts the time series data features of the elevator equipment and fuses them with the spatial features by executing an algorithm combining a convolutional neural network and a recurrent neural network on a computer, thereby generating an accurate spatiotemporal feature vector.

[0015] Optionally, the fault knowledge graph module dynamically expands the elevator fault knowledge graph according to the spatiotemporal feature vector by executing a graph neural network, automatically adds new fault modes and associated nodes, and optimizes the relationship between fault causes and equipment components to ensure the timeliness and accuracy of the graph.

[0016] Optionally, the causal reasoning module analyzes the causal relationship of elevator fault data by executing a causal reasoning algorithm on a computer, generates a list of root cause hypotheses based on historical fault data, and provides a causal chain and exclusion basis for each hypothesis.

[0017] Optionally, the solution generation module generates a series of fault repair solutions by executing a generative adversarial network algorithm through a computer, and verifies whether each solution complies with the safety standards and operating specifications for elevator maintenance through a compliance check algorithm.

[0018] Optionally, the self-optimization module continuously learns feedback data from maintenance personnel by executing a reinforcement learning algorithm, and automatically adjusts parameters of the fault troubleshooting model based on the feedback data.

[0019] In another aspect, a method for troubleshooting an elevator fault is provided, the method comprising:

[0020] Receive and process elevator operation logs, sensor data, and fault reporting records, and generate structured data sets through data standardization and dynamic time warping algorithms;

[0021] Use convolutional neural networks to extract temporal features, and use graph convolutional networks to fuse spatial features to generate spatiotemporal feature vectors;

[0022] Perform fault knowledge graph expansion operations, update the elevator fault knowledge graph through the graph neural network, generate a set of candidate fault causes, and return the confidence level and evidence chain of the fault cause;

[0023] Perform root cause reasoning operations, use causal reasoning algorithms to analyze candidate fault causes, generate a list of root cause hypotheses, and provide a basis for elimination;

[0024] Execute solution generation operations to generate structured solutions including operation steps, safety specifications, and parts lists based on the root cause hypothesis list, and perform compliance verification;

[0025] Perform self-optimization operations and optimize the system's troubleshooting capabilities through reinforcement learning algorithms based on maintenance personnel feedback.

[0026] Optionally, the method combines deep learning with causal reasoning to automatically extract spatiotemporal features and mine fault patterns.

[0027] Optionally, the method expands the elevator fault knowledge graph in real time through a graph neural network, dynamically adding new fault modes and evidence chains for dealing with new fault types.

[0028] Through deep learning, graph neural networks, and causal reasoning techniques, the intelligence level and diagnostic accuracy of elevator fault troubleshooting are improved. The system can efficiently process multi-source heterogeneous data from elevators, including operation logs, sensor time series data, and fault reporting records, and pre-process different data sources using a timestamp alignment algorithm to generate a structured data set. The elevator fault troubleshooting system of the present invention, through the integrated application of AI technology, not only improves the speed and accuracy of fault diagnosis, but also has the ability to continuously self-learn and optimize, and is suitable for various elevator systems, including high-end villa elevators. Through this system, the elevator maintenance process becomes more intelligent and efficient, providing a strong guarantee for the safety of elevator operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The figure is a flow chart of an elevator fault troubleshooting method provided by an illustrative embodiment of the present application. DETAILED DESCRIPTION

[0030] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0031] In this document, "plurality" refers to two or more. "And / or" describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can mean: A exists alone, A and B exist simultaneously, or B exists alone. The character " / " generally indicates an "or" relationship between the associated objects.

[0032] Example 1

[0033] According to one aspect of the present application, there is provided an elevator fault troubleshooting system, the system comprising:

[0034] The data acquisition and preprocessing module is used to receive the elevator's operation log, sensor time series data and fault reporting records through a program executed by a computer, and use the timestamp alignment algorithm to time-align heterogeneous data from different sources to generate a structured data set, which includes sensor data and log data with consistent timestamps.

[0035] In one possible implementation, data is collected from various sensors (temperature, pressure, vibration, etc.) of the elevator. These data usually have different time formats and need to be aligned and standardized through a computer program.

[0036] The feature fusion and pattern mining module is used to extract features from multi-source data, use convolutional neural networks to extract the temporal features of sensor data, and use graph convolutional networks to model the spatial correlation of various elevator components. It then fuses the spatiotemporal feature vectors and outputs the spatiotemporal feature vectors for fault mode mining.

[0037] In a possible implementation, the sensor data from different sources are time-aligned using a DTW algorithm, thereby ensuring that the multi-source data can be associated in chronological order and avoiding calculation errors caused by time errors.

[0038] The fault knowledge graph module is used to query and expand the elevator fault knowledge graph based on spatiotemporal feature vectors, update the nodes and relationships in the fault graph using a graph neural network, and return a set of candidate fault causes, which includes the confidence level of the fault cause and the associated evidence chain.

[0039] In one possible implementation, programming languages ​​such as Python and R are used in conjunction with standardized tools to generate a structured dataset containing timestamp-aligned sensor data, log data, and fault reporting data for subsequent deep learning and pattern recognition.

[0040] The causal reasoning module is used to analyze the causal chain of the fault based on the set of candidate fault causes through the causal reasoning algorithm, and generate a list of root cause hypotheses, each of which contains the causal chain of the fault cause and the basis for exclusion.

[0041] The solution generation module is used to generate a structured solution with operation steps, safety specifications and parts list based on the root cause hypothesis list, and check the legality and safety of the solution through a compliance verification algorithm.

[0042] The self-optimization module is used to adjust the troubleshooting model based on maintenance personnel feedback data using reinforcement learning algorithms and update the knowledge graph and reasoning model.

[0043] In summary, the core of this elevator troubleshooting system lies in its data acquisition and preprocessing module. The system uses a computer program to receive and process multi-source, heterogeneous data from elevator operations, including elevator operation logs (typically stored in text or binary format), sensor time series data (such as accelerometer and temperature sensor data), and fault report records submitted by users or maintenance personnel. All of this data is timestamped using a dynamic time warping (DTW) algorithm to ensure time consistency across different data sources. The system automatically converts data from different devices and formats into a unified format using a data normalization algorithm, generating a structured dataset where each data point includes precise timestamp information, providing high-quality input for subsequent processing. This enables efficient acquisition and unified processing of multi-source elevator data, ensuring data consistency and efficiency in subsequent analysis, and providing high-quality data support for feature extraction, pattern mining, and fault inference. This approach enables the system to collect real-time data on elevator operating status and provide accurate input for fault diagnosis.

[0044] Example 2

[0045] Optionally, the data acquisition and timing alignment module also receives log files and sensor data from different devices and in different formats by executing a data normalization algorithm, and uses a dynamic time warping algorithm to align timestamps of data from different time sources.

[0046] In one possible implementation, the system uses the DTW algorithm to time-align data from different elevator sensors. This method effectively handles data that is not completely synchronized on the timeline. Data from different devices may have different formats, such as binary sensor data and text log data. Using programming tools like Python, the system converts all data into a unified structured format for easier processing.

[0047] In the data acquisition and timing alignment module of the embodiment of the present application, the system receives log files and sensor data from different devices and in different formats through a data standardization algorithm, and aligns the data in combination with the dynamic time warping (DTW) algorithm. Taking the acceleration sensor data and elevator log record data of the elevator as an example, these data come from different devices and are in inconsistent formats. Through the DTW algorithm, the system can align these two data sources to the same time base, avoid errors caused by time asynchrony, and achieve efficient time series data fusion, which can effectively solve the alignment problem of various heterogeneous data in the elevator, ensure the consistency of all data on the time axis, and provide an accurate data basis for subsequent analysis, thereby improving the accuracy of troubleshooting.

[0048] Example 3

[0049] Optionally, the feature fusion and pattern mining module extracts the time series data features of the elevator equipment and fuses them with the spatial features by executing an algorithm combining a convolutional neural network and a recurrent neural network on a computer, thereby generating an accurate spatiotemporal feature vector.

[0050] In one possible implementation, a CNN is used to extract features from elevator time series data (such as vibration and temperature) to identify trends, periodicity, and abnormal fluctuations. A GCN is used to model the spatial relationships between elevator components (e.g., the car, door, and elevator shaft), thereby identifying potential fault modes in the elevator system. The time series and spatial features are combined through a fusion network to generate accurate spatiotemporal feature vectors, supporting subsequent fault diagnosis.

[0051] In summary, in the feature fusion and pattern mining module, the system uses a convolutional neural network (CNN) to extract temporal features from elevator sensor data and a graph convolutional network (GCN) to model spatial features. Ultimately, these spatial and temporal feature vectors are fused for in-depth fault pattern mining. For example, after extracting temporal features from the elevator's vibration and temperature sensor data using a CNN, the GCN models the spatial correlations between elevator components to identify potential fault modes in the elevator equipment. This effectively fuses the temporal and spatial features of the elevator equipment and uses deep learning techniques to mine potential fault modes. The combination of CNN and GCN enables in-depth mining of complex fault patterns, thereby improving the accuracy of elevator fault diagnosis.

[0052] Example 4

[0053] Optionally, the fault knowledge graph module dynamically expands the elevator fault knowledge graph according to the spatiotemporal feature vector by executing a graph neural network, automatically adds new fault modes and associated nodes, and optimizes the relationship between fault causes and equipment components to ensure the timeliness and accuracy of the graph.

[0054] In one possible implementation, GNNs are used to dynamically expand the elevator fault knowledge graph, adding new nodes and relationships in real time based on new fault data to ensure the graph's timeliness and accuracy. GNNs automatically identify the chain of evidence surrounding elevator faults and dynamically add nodes and associated evidence related to the current fault to the graph.

[0055] In summary, the fault knowledge graph module uses a graph neural network (GNN) to expand and update the elevator fault knowledge graph in real time. When the system receives data from different sensors and fault reporting records, the GNN automatically identifies the cause of the elevator fault and dynamically adds new fault nodes and evidence chains to the graph. For example, if the system detects an abnormal correlation between the elevator's acceleration sensor data and the elevator door opening time, the GNN will add this information to the fault graph, expanding the knowledge base of elevator door faults. GNN technology enables the elevator fault knowledge graph to have dynamic expansion capabilities, ensuring that the system can handle ever-changing fault types, updating the knowledge base in real time, and improving the adaptability and accuracy of the troubleshooting system.

[0056] Example 5

[0057] Optionally, the causal reasoning module analyzes the causal relationship of elevator fault data by executing a causal reasoning algorithm on a computer, generates a list of root cause hypotheses based on historical fault data, and provides a causal chain and exclusion basis for each hypothesis.

[0058] In one possible implementation, a Bayesian inference algorithm is used to analyze the fault chain of various elevator components, inferring the root cause of the fault by calculating the conditional probability of each node. Each fault cause is combined with relevant elimination evidence to form a causal chain, ensuring a clear and reliable reasoning process.

[0059] In summary, the causal reasoning module uses causal reasoning algorithms (such as Bayesian networks) to analyze the causal relationships of elevator failures. For example, if an elevator door fails to close properly, the system uses causal reasoning to analyze sensor data, historical failure records, and the door's open and closed status to infer the root cause of the elevator door failure and generate a list of root cause hypotheses. Using causal reasoning algorithms, it can generate a detailed causal chain for each failure, accurately identifying the root cause and providing maintenance personnel with a clear basis for repair.

[0060] Example 6

[0061] Optionally, the solution generation module generates a series of fault repair solutions by executing a generative adversarial network algorithm through a computer, and verifies whether each solution complies with the safety standards and operating specifications for elevator maintenance through a compliance check algorithm.

[0062] In one possible implementation, multiple repair solutions are generated using a GAN, and each solution is evaluated for feasibility and safety using a computer program. After the repair solution is generated, the solution is checked for safety and compliance using an automated compliance verification algorithm.

[0063] In summary, the solution generation module uses a generative adversarial network (GAN) to generate multiple fault repair solutions. For example, if an elevator door fault is diagnosed as a motor failure, the GAN automatically generates multiple repair solutions, including operating procedures, required parts, and safety specifications. A compliance verification algorithm then ensures that these solutions meet safety requirements for elevator maintenance. This automated generation of repair solutions and the assurance that all solutions comply with safety standards significantly improves maintenance efficiency and elevator operational safety.

[0064] Example 7

[0065] Optionally, the self-optimization module continuously learns feedback data from maintenance personnel by executing a reinforcement learning algorithm, and automatically adjusts parameters of the fault troubleshooting model based on the feedback data.

[0066] In one possible implementation, a reinforcement learning algorithm is used to adjust the weights and parameters of the troubleshooting model based on feedback data, allowing the model to continuously optimize over time. Feedback data from each repair becomes an important basis for the system to update the model, thus enabling the system to self-learn.

[0067] In summary, the self-optimization module, based on a reinforcement learning (RL) algorithm, continuously learns from maintenance personnel feedback and adjusts the troubleshooting model based on this feedback. For example, if the system discovers a low diagnostic efficiency for a particular fault mode through feedback data, the RL algorithm automatically adjusts the model parameters to improve the diagnostic accuracy of that mode. Through the self-optimization capabilities of reinforcement learning, the system can continuously improve the accuracy and efficiency of troubleshooting based on feedback data, giving the elevator troubleshooting system the ability to self-improve.

[0068] Example 8

[0069] In another aspect, a method for troubleshooting an elevator fault is provided, the method comprising:

[0070] Step 201: Receive and process the elevator's operation log, sensor data, and fault reporting records, and generate a structured data set through data standardization and dynamic time warping algorithms.

[0071] Preprocessing of the elevator's multi-source heterogeneous data (including sensor data, operation logs, and fault reporting records) ensures data consistency and reliability. Time series data is aligned using the Dynamic Time Warping (DTW) algorithm, overcoming the problem of synchronizing data from different devices and in different formats. The data standardization algorithm ensures that sensor data from different sources (such as accelerometers, temperature sensors, etc.) can be processed in a unified format, thereby reducing errors caused by format conversion and timestamp misalignment. The Dynamic Time Warping (DTW) algorithm is used to optimize the time alignment process of data from different devices and sensors, providing more accurate support for time series data fusion. The introduction of the data standardization algorithm solves the problem of inconsistent data formats generated by multiple devices, avoids the cumbersome data conversion process in traditional methods, and enables the system to handle more complex and diverse data sources.

[0072] Step 202: Use a convolutional neural network to extract temporal features, and use a graph convolutional network to fuse spatial features to generate a spatiotemporal feature vector.

[0073] This step combines two deep learning models, convolutional neural networks (CNN) and graph convolutional networks (GCN), to process the elevator's time series data and spatial data, respectively. CNN excels at extracting trends, periodic changes, and abnormal fluctuations from time series data, while GCN is used to model the spatial relationships between elevator components (such as the car, doors, elevator shaft, etc.). By fusing spatiotemporal features, a spatiotemporal feature vector is generated that can effectively describe the operating status and potential failure modes of the elevator equipment. By deeply fusing the time series features extracted by the convolutional neural network (CNN) with the spatial features extracted by the graph convolutional network (GCN), the multi-dimensional analysis capabilities of the fault diagnosis system are enhanced, and the spatiotemporal behavioral characteristics of the elevator equipment can be captured.

[0074] By modeling the spatial relationships among elevator equipment components using GCN, the system can discover complex spatial fault patterns and improve the accuracy of fault diagnosis, especially in complex elevator systems with multiple interconnected components.

[0075] Step 203: Execute the fault knowledge graph expansion operation, update the elevator fault knowledge graph through the graph neural network, generate a set of candidate fault causes, and return the confidence and evidence chain of the fault cause.

[0076] Graph Neural Networks (GNNs) are used to dynamically expand and update the elevator fault knowledge graph in real time. This graph contains information on various elevator fault patterns, their causes, and influencing factors. Based on spatiotemporal feature vectors, the system analyzes elevator operating data in real time. Using the GNN, it identifies and updates the chain of evidence leading to the fault cause and returns a confidence score for the cause. This enables the elevator fault diagnosis system to quickly adapt to emerging fault types, enhancing its flexibility and timeliness.

[0077] By dynamically expanding the elevator fault knowledge graph through GNN, the system can process new fault modes in real time and automatically identify the associated evidence chain. This real-time update capability, unprecedented in traditional knowledge graphs, significantly improves the system's adaptability and accuracy. By combining spatiotemporal features with graph updates, the system can provide a confidence level for each candidate fault cause, helping to prioritize the most likely fault modes and improve troubleshooting efficiency.

[0078] Step 204 : Execute root cause reasoning operations, use causal reasoning algorithms to analyze candidate fault causes, generate a list of root cause hypotheses, and provide a basis for elimination.

[0079] Using causal inference algorithms (such as Bayesian networks or structural equation models), the system analyzes candidate fault causes and generates a list of root cause hypotheses. Each hypothesis consists of a causal chain and its rationale for elimination, allowing the system to infer the root cause of the fault. This approach not only eliminates unlikely fault causes but also provides maintenance personnel with detailed reasoning, enhancing transparency in troubleshooting.

[0080] Application of causal reasoning algorithms: Applying causal reasoning algorithms to the troubleshooting process allows for a comprehensive analysis of the causal relationships between potential fault causes. This approach transcends traditional statistical or empirical troubleshooting methods, improving the scientific nature and accuracy of fault analysis.

[0081] Each root cause hypothesis not only provides a causal chain for the fault cause, but also includes the basis for excluding irrelevant factors, providing maintenance personnel with a clear and scientific fault diagnosis path.

[0082] Step 205 , executing a solution generation operation, generating a structured solution of operation steps, safety specifications and parts list based on the root cause hypothesis list, and performing compliance verification.

[0083] Based on the root cause hypothesis generated in step 204, the system automatically generates a detailed fault repair plan, including operational steps, safety specifications, and a list of required parts. These plans are verified by a compliance check algorithm to ensure they meet safety standards and operational specifications for elevator maintenance. By automatically generating and verifying repair plans, the system significantly improves maintenance efficiency and safety.

[0084] AI-generated operating procedures, safety regulations, and parts lists not only improve maintenance efficiency but also reduce human error. Compliance verification algorithms ensure that each generated solution adheres to industry standards and safety requirements for elevator maintenance, a feat not possible with traditional troubleshooting systems.

[0085] Step 206 , performing self-optimization operations, based on the feedback from maintenance personnel, and optimizing the system's troubleshooting capabilities through reinforcement learning algorithms.

[0086] Reinforcement learning algorithms enable the system to continuously learn and optimize from feedback from maintenance personnel. When maintenance personnel resolve a problem, the system collects this feedback and uses the reinforcement learning algorithm to update the parameters of the troubleshooting model. This self-optimization capability allows the system to continuously improve the accuracy and efficiency of fault diagnosis based on historical experience.

[0087] Using reinforcement learning algorithms, the system continuously optimizes itself over time, improving the accuracy of elevator fault diagnosis. This innovation enables the system to adapt to changing elevator failure patterns without human intervention.

[0088] The feedback data from maintenance personnel is not only supplementary information for diagnosis, but also the driving force for continuous learning and optimization of the entire system, which forms a closed-loop feedback loop in the troubleshooting process and improves the system's adaptability.

[0089] This method combines deep learning with causal reasoning to not only improve the accuracy of fault pattern recognition but also accurately infer the root cause of the fault, improving the efficiency and accuracy of troubleshooting. By combining deep learning technology and causal reasoning algorithms, this method uses multi-source data to intelligently troubleshoot elevator faults. It extracts temporal features using a convolutional neural network (CNN) and integrates spatial features using a graph convolutional network (GCN) to generate spatiotemporal feature vectors for in-depth fault pattern analysis.

[0090] Optionally, the method automatically extracts spatio-temporal features and fault mode mining through the combination of deep learning and causal reasoning. The method expands the elevator fault knowledge graph through GNN, and the system can adapt to new fault types in real time, improving the real-time and accuracy of fault troubleshooting. The fault cause and evidence chain are updated in real time to ensure that the elevator fault troubleshooting system can cope with the emergence of new fault types.

[0091] Optionally, the method dynamically adds new fault modes and evidence chains for coping with new fault types by expanding the elevator fault knowledge graph in real time through the graph neural network. This method dynamically expands the elevator fault knowledge graph using the graph neural network and adds new fault modes and their evidence chains, improving the intelligence and adaptability of the fault troubleshooting system. By updating the fault knowledge graph in real time, the system can adapt to changes in elevator fault types, ensuring continuous intelligent improvement of the diagnosis system.

[0092] In summary, the present application proposes an intelligent elevator fault troubleshooting system and method through the innovative combination of deep learning, graph neural network and causal reasoning algorithm. The system can not only process multi-source heterogeneous data and identify spatio-temporal features, but also dynamically update the fault knowledge graph and use reinforcement learning algorithm to realize self-optimization, thereby greatly improving the fault troubleshooting efficiency and accuracy, especially suitable for high-end elevator systems such as villa elevators.

[0093] The above description is only optional embodiments of the present application and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An elevator fault troubleshooting system, characterized in that: The system comprises: The data acquisition and preprocessing module is used to receive the elevator's operation log, sensor time series data, and fault reporting records through a computer-executed program. It uses a timestamp alignment algorithm to time-align heterogeneous data from different sources to generate a structured data set that includes sensor data and log data with consistent timestamps. The feature fusion and pattern mining module is used to extract features from multi-source data. It uses a convolutional neural network to extract the temporal features of sensor data and a graph convolutional network to model the spatial correlation of elevator components. It then fuses the spatiotemporal feature vectors and outputs them for fault mode mining. The fault knowledge graph module is used to query and expand the elevator fault knowledge graph based on spatiotemporal feature vectors, update the nodes and relationships in the fault graph using a graph neural network, and return a set of candidate fault causes, including the confidence level of the fault cause and the associated evidence chain; The causal reasoning module is used to analyze the causal chain of the fault based on the candidate fault cause set using a causal reasoning algorithm and generate a list of root cause hypotheses, each of which contains the causal chain of the fault cause and the basis for elimination; The solution generation module is used to generate a structured solution with operation steps, safety specifications, and a parts list based on the root cause hypothesis list, and to check the legality and safety of the solution through a compliance verification algorithm; The self-optimization module is used to adjust the troubleshooting model based on maintenance personnel feedback data using reinforcement learning algorithms and update the knowledge graph and reasoning model.

2. The elevator fault troubleshooting system according to claim 1, characterized in that: The data acquisition and timing alignment module also receives log files and sensor data from different devices and in different formats by executing a data normalization algorithm, and uses a dynamic time warping algorithm to align timestamps on data from different time sources.

3. The elevator fault troubleshooting system according to claim 1, characterized in that: The feature fusion and pattern mining module extracts the time series data features of the elevator equipment and fuses them with the spatial features by executing an algorithm combining convolutional neural networks and recurrent neural networks on a computer, thereby generating an accurate spatiotemporal feature vector.

4. The elevator fault troubleshooting system according to claim 1, characterized in that: The fault knowledge graph module dynamically expands the elevator fault knowledge graph based on the spatiotemporal feature vector by executing a graph neural network, automatically adding new fault modes and associated nodes, and optimizing the relationship between fault causes and equipment components to ensure the timeliness and accuracy of the graph.

5. The elevator fault troubleshooting system according to claim 1, characterized in that: The causal reasoning module uses a computer to execute a causal reasoning algorithm to analyze the causal relationship of elevator fault data, generates a list of root cause hypotheses based on historical fault data, and provides a causal chain and exclusion basis for each hypothesis.

6. The elevator fault troubleshooting system according to claim 1, characterized in that: The solution generation module generates a series of fault repair solutions by executing a generative adversarial network algorithm on a computer, and verifies whether each solution complies with the safety standards and operating specifications for elevator maintenance through a compliance check algorithm.

7. The elevator fault troubleshooting system according to claim 1, characterized in that: The self-optimization module continuously learns feedback data from maintenance personnel by executing a reinforcement learning algorithm, and automatically adjusts the parameters of the fault troubleshooting model based on the feedback data.

8. An elevator fault troubleshooting method, characterized in that: The method comprises: Receive and process elevator operation logs, sensor data, and fault reporting records, and generate structured data sets through data standardization and dynamic time warping algorithms; Use convolutional neural networks to extract temporal features, and use graph convolutional networks to fuse spatial features to generate spatiotemporal feature vectors; Perform fault knowledge graph expansion operations, update the elevator fault knowledge graph through the graph neural network, generate a set of candidate fault causes, and return the confidence level and evidence chain of the fault cause; Perform root cause reasoning operations, use causal reasoning algorithms to analyze candidate fault causes, generate a list of root cause hypotheses, and provide a basis for elimination; Execute solution generation operations to generate structured solutions including operation steps, safety specifications, and parts lists based on the root cause hypothesis list, and perform compliance verification; Perform self-optimization operations and optimize the system's troubleshooting capabilities through reinforcement learning algorithms based on maintenance personnel feedback.

9. The elevator fault troubleshooting method according to claim 8, characterized in that: The method combines deep learning with causal reasoning to automatically extract spatiotemporal features and mine fault patterns.

10. The elevator fault troubleshooting method according to claim 8, characterized in that: The method uses a graph neural network to expand the elevator fault knowledge graph in real time, dynamically adding new fault modes and evidence chains to deal with new fault types.

Images