Fault diagnosis and safety early warning system of automatic driving carrying robot

Abstract

The invention discloses a fault diagnosis and safety early warning system for an automatic driving carrying robot, and the system comprises the steps: collecting panoramic operation tense data, carrying out the threshold judgment, cross validation, performance and trend analysis, and task execution result verification, and outputting a preliminary diagnosis result; performing time anchor point diffusion analysis, locking a core event range, extracting related synchronization fragments, analyzing an association relationship, calculating association strength of different data exceptions and core faults, and locking an exceptional operation data range; constructing an abnormal event graph structure, identifying a root cause node and a key propagation path, generating a macroscopic abnormal event graph, performing hypothesis verification, and outputting an abnormal operation data chain; performing scene reproduction through digital twinning to position a root cause, and outputting a fault analysis result; and performing hierarchical security early warning and generating a fault file, and storing the fault file in a block chain. The method aims at solving the problems that complex faults of an automatic driving carrying robot are difficult to position quickly, and the root cause diagnosis efficiency is low.

Description

Technical Field

[0001] This invention relates to the field of robot fault diagnosis technology, and in particular to a fault diagnosis and safety early warning system for an autonomous driving transport robot. Background Technology

[0002] Currently, fault diagnosis of autonomous driving robots largely relies on preset threshold alarms or isolated analysis of single data sources, which has significant limitations. First, the system consists of deeply coupled multiple modules; an initial anomaly can propagate rapidly along dependency chains, forming a complex "network of fault phenomena," making it difficult for traditional methods to identify the core root cause from numerous concurrent alarms. Second, existing technologies lack in-depth mining of multimodal and spatiotemporal data correlations, failing to effectively construct a complete causal path from symptoms to results, resulting in superficial diagnosis. Furthermore, the diagnostic process heavily relies on expert experience for manual backtracking, which is inefficient and prone to misjudgment. Although some solutions introduce digital twin or graph technologies, they are mostly limited to local applications, failing to form a closed-loop, self-consistent system from data perception and causal reasoning to virtual verification. Moreover, diagnostic knowledge cannot be effectively accumulated and reused, resulting in a lack of continuous system evolution capabilities. Summary of the Invention

[0003] This invention provides a fault diagnosis and safety early warning system for an autonomous driving transport robot, comprising: Preliminary diagnosis module: Collects panoramic runtime data, and after threshold judgment, cross-validation, performance and trend analysis, and task execution result verification, outputs preliminary diagnosis results; Anomaly Range Locking Module: Based on the preliminary diagnostic results, time anchor point diffusion analysis is performed to lock the range of core events, extract synchronization fragments of relevant data sources within the range, analyze the correlation in different data streams, and calculate the correlation strength between different data anomalies and core faults to lock the range of abnormal running data. Anomaly Graph Construction Module: Based on the range of abnormal operation data, it mines the causal and temporal relationships between events, constructs the anomaly event graph structure, identifies root cause nodes and key propagation paths, and generates a macro-level anomaly event graph; Anomaly data chain verification module: drills down layer by layer according to the macro-anomaly event map and combines it with historical fault files to verify hypotheses and output anomaly operation data chain; Fault Analysis Module: Based on the abnormal data chain, the module uses digital twins to reproduce the scene, locate the root cause, output the fault analysis results, and save the records. Tiered early warning and file storage module: Based on the fault analysis results, it provides tiered security early warnings and generates fault files, which are stored in an immutable blockchain.

[0004] The aforementioned fault diagnosis and safety early warning system for an autonomous driving transport robot includes an anomaly range locking module, comprising: Anomaly data fusion submodule: Using the anomaly time point of the initial diagnosis as the anchor point, it traces back the origin of the symptoms and covers the scope of influence backward to determine the core analysis time period. Within this time period, it extracts multimodal data of all upstream and downstream dependent modules of the suspected module located by the initial results to form a multidimensional anomaly data cube. Anomaly Propagation Analysis Submodule: Analyzes the order and dependencies of anomalies in a multidimensional anomaly data cube, identifies key anomaly propagation paths, calculates the correlation strength between each anomaly data and the core fault, and, in conjunction with the anomaly propagation paths, pinpoints the set of anomaly running data.

[0005] The aforementioned fault diagnosis and safety early warning system for an autonomous driving transport robot includes an anomaly propagation analysis submodule, comprising: Causal path construction submodule: Based on the data cube, the propagation order of abnormal data is determined by a graph propagation model based on causal reasoning algorithm, and the key abnormal propagation path from potential root cause to final failure phenomenon is constructed. The correlation strength assessment submodule calculates the correlation strength between each data anomaly and the core fault indicators based on the anomaly data cube, prioritizes the data based on the key anomaly propagation paths, and locks down the set of abnormal running data. The aforementioned fault diagnosis and safety early warning system for an autonomous driving transport robot includes an abnormal event graph construction module, comprising: The graph structure construction submodule constructs an abnormal event graph structure based on the abnormal operation data set, with abnormal events as nodes and causal relationships between events as edges. Root cause node identification submodule: Calculates node importance based on the anomaly event graph structure, identifies root cause nodes, and uses community detection algorithms to divide anomaly event clusters to generate a macro-level anomaly event graph. The aforementioned fault diagnosis and safety early warning system for an autonomous driving transport robot includes an abnormal data link verification module, comprising: Fault Hypothesis Generation Submodule: Based on the root cause candidate nodes of the macro-abnormal event map, and combined with historical fault archives, it generates multiple specific fault mechanism hypotheses and prioritizes them according to causal confidence and impact scope. Evidence collection submodule: For high-priority hypotheses, drill down from the top to the bottom of the system hierarchy to collect quantitative and qualitative evidence that supports the hypothesis; Hypothesis verification submodule: comprehensively evaluate the strength of evidence, verify the hypothesis, and reconstruct a complete causal chain with data support from the root cause to the failure phenomenon, forming an abnormal operation data chain.

[0006] The aforementioned fault diagnosis and safety early warning system for an autonomous driving transport robot includes a fault analysis module comprising: Virtual Model Construction Submodule: Based on the physics engine and machine learning model, it constructs a virtual model of the robot to accurately simulate the dynamics, sensors, and behavior of the actual robot; Fault Reproduction Submodule: Input the abnormal data chain, run the reproduction scenario in a virtual environment, perform parameter adjustments, isolate the root cause, use sensitivity analysis to quantify the impact of each factor, and generate a detailed fault analysis report.

[0007] The aforementioned fault diagnosis and safety early warning system for an autonomous driving transport robot includes a graded early warning and file storage module, comprising: Tiered early warning submodule: Based on the fault analysis structure, the severity is classified and a safety warning is triggered and displayed through multiple channels; The file storage submodule generates a complete fault file based on the fault analysis report and stores the fault file in a standardized format on the blockchain network for hypothesis verification.

[0008] The beneficial effects achieved by this invention are as follows: By constructing a multi-dimensional anomaly data cube and analyzing its propagation paths, we can deeply mine the correlations between multi-source data and accurately pinpoint the core fault scope and propagation chain from concurrent anomalies. Utilizing anomaly event mapping and hypothesis testing mechanisms, we systematically reveal the root causes and evolutionary logic of faults, significantly improving the depth and interpretability of diagnosis. Leveraging digital twin technology for safe and controllable scenario reproduction enables direct verification and localization of the root cause. The resulting fault profile is stored on the blockchain and fed back to the historical archive, forming a closed-loop learning and continuous optimization capability for the system, greatly improving the level of intelligent operation and maintenance and system reliability. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0010] Figure 1 This is a schematic diagram of a fault diagnosis and safety early warning system for an autonomous driving transport robot provided in Embodiment 1 of this application. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] Example 1 like Figure 1 As shown, Embodiment 1 of this application provides a fault diagnosis and safety early warning system for an autonomous driving transport robot, comprising: Preliminary diagnosis module: Collects panoramic runtime data, and after threshold judgment, cross-validation, performance and trend analysis, and task execution result verification, outputs preliminary diagnosis results; The system comprehensively collects multi-dimensional dynamic data from the transport robot, covering the perception layer, positioning and navigation layer, control layer, system state layer, and task layer, constructing a complete data foundation. It then compares real-time data with preset static / dynamic thresholds to quickly identify obvious anomalies. Through cross-validation, it leverages information redundancy to compare the consistency of multi-source data, identifying data inconsistencies caused by the failure of a single sensor or module, such as cross-source verification of positioning, perception, and control systems. By using a sliding time window, it performs statistical analysis and trend fitting on time-series data to capture soft faults or performance degradation symptoms that are not exceeding thresholds but are continuously worsening, such as increased control error and abnormal motor temperature rise. Next, it verifies the task execution results, checking whether the robot completes the task as expected at the functional level, and discovering systemic and logical faults that are difficult to reach through lower-level analysis, such as task execution timeouts and excessive path deviations. Finally, it comprehensively analyzes the output anomaly alarms, generating a structured preliminary diagnostic result containing a list of abnormal events and preliminary suspects.

[0013] Anomaly Range Locking Module: Based on the preliminary diagnostic results, time anchor point diffusion analysis is performed to lock the range of core events, extract synchronization fragments of relevant data sources within the range, analyze the correlation in different data streams, and calculate the correlation strength between different data anomalies and core faults to lock the range of abnormal running data. The anomaly range locking module includes the following sub-modules: Anomaly data fusion submodule: Using the anomaly time point of the initial diagnosis as the anchor point, it traces back the origin of the symptoms and covers the scope of influence backward to determine the core analysis time period. Within this time period, it extracts multimodal data of all upstream and downstream dependent modules of the suspected module located by the initial results to form a multidimensional anomaly data cube. Based on the abnormal time points and the preliminary suspect module list in the preliminary diagnosis module, the core analysis time period is determined through a change point detection algorithm. When tracing backward, the start time is calculated by combining the system inertia of the suspect modules (such as response time constant and data buffer cycle) with the average latency of similar historical faults, capturing the earliest symptom signals. When covering backward, the end time is determined based on the downstream influence range and processing delay of the preliminary suspect modules in the system dependency topology, fully covering the anomaly propagation process and forming a dynamically adapted core analysis time period.

[0014] Next, based on a predefined module dependency topology graph, data from suspected modules is collected, and their upstream and downstream modules are automatically associated to ensure the completeness of root cause tracing and impact analysis. Then, a unified high-precision master timeline is created to unify the heterogeneity of different data streams in terms of timestamps, sampling frequencies, and formats. A multidimensional anomaly data cube is then constructed, using time, module (e.g., sensing / forward-facing cameras, control / steering), and data metrics (e.g., lateral tracking error) as three dimensions. At each intersection point, cleaned and aligned data values ​​and anomaly confidence scores are stored. Through lightweight cross-dimensional correlation scanning, Pearson correlation coefficients and mutual information are pre-calculated, and the correlation strength matrix is ​​embedded as metadata into the cube.

[0015] Anomaly Propagation Analysis Submodule: Analyzes the order and dependencies of anomalies in a multidimensional anomaly data cube, identifies key anomaly propagation paths, calculates the correlation strength between each anomaly data and the core fault, and, in conjunction with the anomaly propagation paths, pinpoints the set of anomaly running data.

[0016] The anomaly propagation analysis submodule includes the following submodules: Causal path construction submodule: Based on the data cube, the propagation order of abnormal data is determined by a graph propagation model based on causal reasoning algorithm, and the key abnormal propagation path from potential root cause to final failure phenomenon is constructed. The anomaly confidence sequences of <module, index> in the data cube are defined as candidate nodes. Candidate edges are constructed using improved temporal Granger causal inference. A minimum system response delay threshold δ is introduced, and only directed edges that can significantly predict the node state at time t+δ are retained. Combined with a pre-defined dependency topology graph, associations without physical paths are eliminated, forming a candidate causal directed graph. Subsequently, using the weighted geometric mean of temporal causal significance score and data association strength as edge weights, a root cause candidate set with low in-degree and early anomaly initiation is identified. An improved Dijkstra algorithm is used to find the optimal path with the highest cumulative causal flow strength for each candidate. Finally, the overall explanatory power is calculated by combining the cumulative causal flow strength of the path with the number of covered anomaly nodes / path length, and 1-3 key anomaly propagation paths with the highest explanatory power are selected.

[0017] The correlation strength assessment submodule calculates the correlation strength between each data anomaly and the core fault indicators based on the anomaly data cube, prioritizes the data based on the key anomaly propagation paths, and locks down the set of abnormal running data.

[0018] The core of the correlation strength assessment submodule lies in a dual-view fusion assessment algorithm. Channel 1 calculates the global statistical correlation between each abnormal data sequence and the core fault indicator within the complete core time period, regardless of time sequence, using weighted mutual information. Anomaly confidence is used as the weight to filter out normal fluctuation noise and capture signals co-variant with the core fault. Channel 2 decomposes the contribution based on the propagation path, calculates the causal driving force of nodes through conditional transition entropy or residual analysis, and combines it with a hierarchical attenuation factor (the closer the root cause, the higher the weight) to derive the path causal correlation strength. Nodes on key propagation paths are included in the high-priority dataset, and their final score is calculated primarily based on path causal correlation strength and secondarily on global statistical correlation. High thresholds are set for non-path nodes, and only associated or competing root cause signals with global statistical correlation strength exceeding the threshold are included as candidate datasets. Finally, the datasets are merged to form an abnormal operation data set, with each data point accompanied by a correlation strength score and a reason for inclusion.

[0019] Anomaly Graph Construction Module: Based on the range of abnormal operation data, it mines the causal and temporal relationships between events, constructs the anomaly event graph structure, identifies root cause nodes and key propagation paths, and generates a macro-level anomaly event graph; The abnormal event graph construction module includes the following sub-modules: The graph structure construction submodule constructs an abnormal event graph structure based on the abnormal operation data set, with abnormal events as nodes and causal relationships between events as edges. The core task of the graph structure construction submodule is to transform the set of abnormal operational data with labeled correlation strengths, output by the preceding modules, into a computable, semantically rich abnormal event graph structure. Specifically, each abnormal data point in the set is treated as an abnormal event node based on its associated physical module and specific indicators. Each node is assigned rich attributes, including but not limited to: the time window for anomaly triggering, duration, the final confidence level inherited from the correlation strength assessment module, and the magnitude of its abnormal data value deviation. Based on this, key anomaly propagation paths are used as the backbone, and they are directly instantiated as directed edges with the highest weights in the graph. These backbone edges are also assigned a causal flow strength attribute, the value of which comes from the cumulative causal flow strength in the upstream path analysis. Nodes not on the backbone paths but with extremely strong global statistical correlations are also introduced and connected by dashed edges to represent their related rather than dominant causal relationships. A graph complexity-based pruning algorithm is used to remove redundant edges with minimal impact on the overall structure, forming the abnormal event graph structure.

[0020] Root cause node identification submodule: Calculates node importance based on the anomaly event graph structure, identifies root cause nodes, and uses community detection algorithms to divide anomaly event clusters to generate a macro-level anomaly event graph.

[0021] A multi-dimensional analysis is conducted based on the anomaly event graph structure. An improved directed PageRank algorithm is used to calculate node importance. Causal flow strength is introduced as an edge weight in addition to traditional link analysis, and a time decay factor is assigned to each node, giving higher root cause weights to earlier-appearing anomalies in the propagation chain. Subsequently, modular dynamic community detection is performed using a causal flow-oriented Louvain optimization algorithm to classify anomalies into fault clusters with strong internal causal relationships. Finally, a macroscopic anomaly event graph is generated, presenting a clear hierarchical structure. The top layer displays core root cause nodes and main propagation paths, the middle layer shows the correlation strength of each fault cluster, and the bottom layer retains the detailed attributes of the original anomalies.

[0022] Anomaly data chain verification module: drills down layer by layer according to the macro-anomaly event map and combines it with historical fault files to verify hypotheses and output anomaly operation data chain; The abnormal data chain verification module includes the following sub-modules: Fault Hypothesis Generation Submodule: Based on the root cause candidate nodes of the macro-abnormal event map, and combined with historical fault archives, it generates multiple specific fault mechanism hypotheses and prioritizes them according to causal confidence and impact scope. Based on the root cause candidate nodes identified in the macro-anomaly event graph, deep semantic analysis is performed on each root cause node to extract its key attributes, including its subsystem, equipment type, anomaly pattern, and occurrence time context. Based on these features, a graph neural network-based similarity matching algorithm is used to search for historical failure cases with similar root cause attributes and similar propagation path patterns in the archive. Subsequently, the retrieved historical failure mechanisms are used as basic templates and parameterized to adapt to the current specific context, generating multiple specific failure mechanism hypotheses. Each hypothesis clearly describes the complete causal chain from the root cause to the phenomenon and is labeled with its similarity to historical cases. Finally, a multi-index fusion priority evaluation model is adopted, comprehensively considering the causal confidence of each hypothesis (based on the causal flow strength of the path in the current graph and the number of verifications in historical cases) and the scope of influence (based on the number of downstream anomaly nodes that the hypothesis can explain). A weighted scoring algorithm is used to rank all hypotheses, outputting a list of failure hypotheses arranged in descending order of resolution priority.

[0023] Evidence collection submodule: For high-priority hypotheses, drill down from the top to the bottom of the system hierarchy to collect quantitative and qualitative evidence that supports the hypothesis; High-priority hypotheses are analyzed in depth to identify causal nodes and propagation paths requiring verification. Based on the system's hierarchical topology and fault propagation model, a three-dimensional forensic framework covering the hardware, software, and control layers is constructed, automatically generating targeted data collection lists. During the forensic process, multi-dimensional data collection is carried out from top to bottom along the system dependencies: from application-layer task execution logs and decision records, to middleware service call chains and data flow states, and then to the underlying sensor raw readings and control signals. In particular, the system activates an association retrieval engine based on a multi-dimensional anomaly data cube to extract direct evidence related to the hypotheses and proactively discover indirect evidence with spatiotemporal correlations. Natural language processing techniques are used to analyze anomaly descriptions and alarm information in system logs to collect qualitative evidence; time-series data analysis methods are used to verify the consistency between data patterns and hypothesis predictions to collect quantitative evidence. All evidence is weighted and integrated through a unified confidence assessment model to form a structured evidence set.

[0024] Hypothesis verification submodule: comprehensively evaluate the strength of evidence, verify the hypothesis, and reconstruct a complete causal chain with data support from the root cause to the failure phenomenon, forming an abnormal operation data chain.

[0025] The propagation path described by the fault hypothesis is instantiated in the event graph, forming a target causal chain to be verified. Then, real-time operational data corresponding to each node in the chain is extracted from the anomaly data cube, providing raw material for verification. Next, the critical stress testing phase begins, calculating the causal transmission efficiency between adjacent nodes. This composite metric simultaneously evaluates the accuracy and logical consistency of state transmission—that is, whether the anomalous output of the preceding node is sufficient in both time and magnitude to trigger the anomalous input of subsequent nodes, and whether the actual data correlation strength reaches the expected value when the path was constructed. The specific formula for calculating the causal transmission efficiency is as follows: For adjacent nodes A (preceding) and B (following), their abnormal time series are defined as follows: and And extract the corresponding normal baseline time series from historical data. and ; The dependency between A and B is measured by mutual information. If the causal chain of the failure holds, the correlation strength under abnormal conditions should be significantly higher than the normal baseline; this factor > 1 is strong evidence that the hypothesis holds. It is the mutual information between A and B under abnormal conditions, capturing possible nonlinear causal relationships; It is the mutual information between A and B under normal baseline conditions; Calculate the similarity between the abnormal waveforms of A and B; the smaller the ratio and the larger the exponent, the higher the morphological consistency. It is the DTW distance between sequence A and sequence B under abnormal conditions, used to evaluate the consistency of morphological transmission, and is insensitive to time delay and scaling; DTW distance at normal baseline; Verify whether the actual delay matches the system's expectations to determine if a causal triggering relationship exists; It is the actual time delay, calculated. and The peak value of the cross-correlation function is obtained; This is the expected time delay, based on historical data.

[0026] Finally, a path resilience score is calculated based on the overall link's throughput performance. When the score exceeds a threshold, the hypothesis is verified.

[0027] For a line of Anomaly propagation path composed of nodes Its pathway resilience score The calculation is as follows: It is the first Causal propagation efficiency of a stripe; It is a dynamic topology weight, used to assign higher weights to critical connections on the critical path of the system. It is a weighted average efficiency. Physical meaning: Penalizes paths with large efficiency fluctuations. A stable path consisting of moderately efficient edges is better than a fluctuating path containing both extremely high and extremely low efficiency edges. This represents the actual number of edges on the path. Indicates the maximum possible propagation path length; This represents a complexity sensitivity coefficient, which, under the same efficiency, favors shorter, more direct propagation paths; to further improve accuracy, a correction based on data quality is introduced: ,in It is the first The standard error of the causal propagation efficiency of the stripe; This engineered testing method automatically generates an abnormal operation data chain composed of multi-source data fragments linked together according to causal logic. This empirically verified data chain fully presents the entire process of fault evolution.

[0028] Fault Analysis Module: Based on the abnormal data chain, the module uses digital twins to reproduce the scene, locate the root cause, output the fault analysis results, and save the records. The fault analysis module includes the following sub-modules: Virtual Model Construction Submodule: Based on the physics engine and machine learning model, it constructs a virtual model of the robot to accurately simulate the dynamics, sensors, and behavior of the actual robot; The core task of the virtual model construction submodule is to create a digital baseline image that is highly consistent with the real autonomous driving robot. Specifically, based on the robot's precise geometric dimensions, mass attributes, joint constraints, and transmission system parameters, its multibody dynamics skeleton is constructed in the physics engine to ensure that the virtual model's response to physical effects such as inertia, friction, and collisions is indistinguishable from the physical robot. Next, high-fidelity sensor simulation units are integrated into the model: the camera module integrates ray tracing and noise models to generate realistic images; the lidar simulates the physical processes of beam transmission and reception; and the IMU is injected with bias and noise that conform to its real characteristics. Finally, by injecting a behavior clone model trained with a large amount of actual operating data, the typical navigation, decision-making, and control behavior logic of the physical robot in complex environments is reproduced, constructing a baseline digital twin without introducing any faults.

[0029] Fault Reproduction Submodule: Input the abnormal data chain, run the reproduction scenario in a virtual environment, perform parameter adjustments, isolate the root cause, use sensitivity analysis to quantify the impact of each factor, and generate a detailed fault analysis report.

[0030] In the fault reproduction submodule, the abnormal data chain is transformed into driving instructions for the virtual model, and the root cause is accurately located through systematic parameter space exploration. Specifically, the abnormal data chain is deeply analyzed to identify the abnormal parameter sets and their spatiotemporal distribution patterns. Then, a layered perturbation test is initiated in the virtual environment: the first layer runs benchmark tests within the normal parameter range to establish a behavioral baseline; the second layer, based on the temporal characteristics of the abnormal data chain, gradually injects quantified parameter deviations to reproduce the fault phenomenon; the third layer performs targeted parameter isolation, using the controlled variable method to identify the key parameter combinations that lead to the fault.

[0031] Specifically, each test generates a fault impact matrix, recording the degree to which changes in each parameter affect system behavior. Regarding the parameters... Disturbance test Chinese system behavior indicators degree of influence The calculation formula is as follows: Indicates parameters In the test After the disturbance, the system behavior indicators Observed values; Indicators in benchmark testing The normal value; Used to quantify the relative extent to which an indicator deviates from the normal benchmark; the larger the value, the more significant the impact. Used to assess changes in the distribution pattern of indicators caused by parameter perturbations; among which... Indicates the distribution of indicators after disturbance Compared with the baseline distribution KL divergence between them; This represents the normalization coefficient of the KL divergence, which is the standard deviation of the KL divergence in historical tests. Measuring this indicator Fault detection discrimination capability; Indicators The area under the ROC curve on the historical fault dataset; This represents the largest AUC value among all indicators, used for normalization. Indicates parameters The relative amplitude of the disturbance; This represents the fault trigger threshold, used to determine the critical point at which the system begins to produce a significant response; This indicates the transition speed, used to control the intensity of the image response near a threshold. It is used to simulate common nonlinear responses in real systems. Small perturbations may have no effect, but the effect increases sharply after exceeding the threshold.

[0032] Through multiple rounds of iterative testing, a parameter sensitivity map was constructed to precisely quantify the contribution of each parameter to the fault phenomenon. For each parameter... Its overall impact score on all indicators across all tests The calculation is as follows: in, This indicates the number of indicators and the number of tests. Indicators Importance weights; Indicates test The order of disturbance intensity; This represents the attenuation coefficient, which controls the sensitivity to extreme test conditions.

[0033] In addition, real-time comparisons are performed with historical fault records to verify the uniqueness or prevalence of parameter combinations. Finally, a detailed fault analysis report is generated, including root cause parameter location, quantitative analysis of impact, and priority ranking of repair recommendations.

[0034] Tiered early warning and file storage module: Based on the fault analysis results, it provides tiered security early warnings and generates fault files, which are stored in an immutable blockchain.

[0035] The tiered early warning and archive storage module includes the following sub-modules: Tiered early warning submodule: Based on the fault analysis structure, the severity is classified and a safety warning is triggered and displayed through multiple channels; A hierarchical matrix is ​​established based on three levels: functional safety, task interruption, and equipment wear and tear. When a fatal risk such as a low-level hardware failure or a broken causal chain is identified, a red emergency alert is triggered, forcing the system into a safe mode. For intermediate risks involving localized performance degradation but where the system is still operational, a yellow alert is activated, and a degraded operation plan is generated. For intermittent soft failures, a blue alert is recorded. The system simultaneously builds a panoramic early warning dashboard, highlighting key alarms on terminals, sending structured early warning reports to the monitoring center, and pushing simplified warning information to mobile devices.

[0036] The file storage submodule generates a complete fault file based on the fault analysis report and stores the fault file in a standardized format on the blockchain network for hypothesis verification.

[0037] The module encapsulates elements such as fault analysis reports, anomaly data chains, and causal graphs into traceable knowledge units. It employs a lightweight blockchain architecture and innovatively introduces a smart contract-driven retrieval protocol when storing archive hash values. This protocol ensures that the archives are not only immutable but also automatically invoked by the hypothesis testing submodule—when the system detects a similar fault pattern, the smart contract proactively pushes causal paths and solutions from historical archives, forming a closed-loop application of diagnostic knowledge.

[0038] Corresponding to the above embodiments, the present invention provides a computer storage medium, including: at least one memory and at least one processor; The memory is used to store one or more program instructions; A processor is used to run one or more program instructions to execute a fault diagnosis and safety warning method for an autonomous driving transport robot.

[0039] Corresponding to the above embodiments, this embodiment of the invention provides a computer-readable storage medium containing one or more program instructions, which are executed by a processor to provide a fault diagnosis and safety warning method for an autonomous driving transport robot.

[0040] The embodiments disclosed in this invention provide a computer-readable storage medium storing computer program instructions. When the computer program instructions are executed on a computer, the computer performs the aforementioned method for fault diagnosis and safety warning of an autonomous driving transport robot.

[0041] In this embodiment of the invention, the processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0042] The various methods, steps, and logic diagrams disclosed in the embodiments of this invention can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The processor reads information from the storage medium and, in conjunction with its hardware, completes the steps of the above methods.

[0043] The storage medium can be memory, such as volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.

[0044] Among them, non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.

[0045] Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (Synchlink DRAM, SLDRAM), and direct memory bus RAM (DRRAM).

[0046] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0047] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in this invention can be implemented using a combination of hardware and software. When applied as software, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of computer programs from one place to another. Storage media can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0048] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. A fault diagnosis and safety early warning system for an autonomous driving transport robot, characterized in that, include: Preliminary Diagnostic module: Collects panoramic runtime data, performs threshold judgment, cross-validation, performance and trend analysis, and verifies task execution results, and outputs preliminary diagnostic results; Anomaly Range Locking Module: Based on the preliminary diagnostic results, time anchor point diffusion analysis is performed to lock the range of core events, extract synchronization fragments of relevant data sources within the range, analyze the correlation in different data streams, and calculate the correlation strength between different data anomalies and core faults to lock the range of abnormal running data. Anomaly Graph Construction Module: Based on the range of abnormal operation data, it mines the causal and temporal relationships between events, constructs the anomaly event graph structure, identifies root cause nodes and key propagation paths, and generates a macro-level anomaly event graph; Anomaly data chain verification module: drills down layer by layer according to the macro-anomaly event map and combines it with historical fault files to verify hypotheses and output anomaly operation data chain; Fault Analysis Module: Based on the abnormal data chain, the module uses digital twins to reproduce the scene, locate the root cause, output the fault analysis results, and save the records. Tiered early warning and file storage module: Based on the fault analysis results, it provides tiered security early warnings and generates fault files, which are stored in an immutable blockchain.

2. The fault diagnosis and safety early warning system for an autonomous driving transport robot according to claim 1, characterized in that, The anomaly range locking module includes: Anomaly data fusion submodule: Using the anomaly time point of the initial diagnosis as the anchor point, it traces back the origin of the symptoms and covers the scope of influence backward to determine the core analysis time period. Within this time period, it extracts multimodal data of all upstream and downstream dependent modules of the suspected module located by the initial results to form a multidimensional anomaly data cube. Anomaly Propagation Analysis Submodule: Analyzes the order and dependencies of anomalies in a multidimensional anomaly data cube, identifies key anomaly propagation paths, calculates the correlation strength between each anomaly data and the core fault, and, in conjunction with the anomaly propagation paths, pinpoints the set of anomaly running data.

3. The fault diagnosis and safety early warning system for an autonomous driving transport robot according to claim 2, characterized in that, The anomaly propagation analysis submodule includes: Causal path construction submodule: Based on the data cube, the propagation order of abnormal data is determined by a graph propagation model based on causal reasoning algorithm, and the key abnormal propagation path from potential root cause to final failure phenomenon is constructed. The correlation strength assessment submodule calculates the correlation strength between each data anomaly and the core fault indicators based on the anomaly data cube, prioritizes the data based on the key anomaly propagation paths, and locks down the set of abnormal running data.

4. The fault diagnosis and safety early warning system for an autonomous driving transport robot according to claim 1, characterized in that, The abnormal event graph construction module includes: The graph structure construction submodule constructs an abnormal event graph structure based on the abnormal operation data set, with abnormal events as nodes and causal relationships between events as edges. Root cause node identification submodule: Calculates node importance based on the anomaly event graph structure, identifies root cause nodes, and uses community detection algorithms to divide anomaly event clusters to generate a macro-level anomaly event graph.

5. The fault diagnosis and safety early warning system for an autonomous driving transport robot according to claim 1, characterized in that, The abnormal data chain verification module includes: Fault Hypothesis Generation Submodule: Based on the root cause candidate nodes of the macro-abnormal event map, and combined with historical fault archives, it generates multiple specific fault mechanism hypotheses and prioritizes them according to causal confidence and impact scope. Evidence collection submodule: For high-priority hypotheses, drill down from the top to the bottom of the system hierarchy to collect quantitative and qualitative evidence that supports the hypothesis; Hypothesis verification submodule: comprehensively evaluate the strength of evidence, verify the hypothesis, and reconstruct a complete causal chain with data support from the root cause to the failure phenomenon, forming an abnormal operation data chain.

6. The fault diagnosis and safety early warning system for an autonomous driving transport robot according to claim 1, characterized in that, The fault analysis module includes: Virtual Model Construction Submodule: Based on the physics engine and machine learning model, it constructs a virtual model of the robot to accurately simulate the dynamics, sensors, and behavior of the actual robot; Fault Reproduction Submodule: Input the abnormal data chain, run the reproduction scenario in a virtual environment, perform parameter adjustments, isolate the root cause, use sensitivity analysis to quantify the impact of each factor, and generate a detailed fault analysis report.

7. The fault diagnosis and safety early warning system for an autonomous driving transport robot according to claim 1, characterized in that, The tiered early warning and archive storage module includes: Tiered early warning submodule: Based on the fault analysis structure, the severity is classified and a safety warning is triggered and displayed through multiple channels; The file storage submodule generates a complete fault file based on the fault analysis report and stores the fault file in a standardized format on the blockchain network for hypothesis verification.

Images