Multi-source heterogeneous data driven equipment full life cycle health assessment method and system

Abstract

The invention relates to the technical field of industrial equipment health assessment, in particular to an equipment full-life-cycle health assessment method and system driven by multi-source heterogeneous data. The method comprises the following steps: collecting multi-source heterogeneous data of equipment, and constructing a life cycle health state time sequence diagram; constructing an equipment standard health path based on the standard data of the equipment; constructing a historical health path of the equipment based on the multi-source heterogeneous data; constructing an equipment forward prediction health path based on the historical health path and the multi-source heterogeneous data of the current life cycle stage; defining a virtual decommissioning final-state node of the equipment, and constructing a reverse causal inverse deduction health path of the equipment; using a four-path health deviation mode evaluation mechanism to analyze deviation mode distribution and health evaluation results of the equipment; the invention realizes a method for comprehensively evaluating the health state of the equipment in each stage of the full life cycle and the influence of the health state on the retirement result.

Description

Technical Field

[0001] This invention relates to the field of industrial equipment health assessment technology, specifically to a method and system for equipment lifecycle health assessment driven by multi-source heterogeneous data. Background Technology

[0002] In modern industrial production systems, large-scale critical equipment continuously generates multi-source heterogeneous data from sensing and monitoring systems, control systems, production execution systems, and operation and maintenance management systems during design, manufacturing, installation, commissioning, and long-term operation and maintenance. This includes continuous monitoring data such as vibration, temperature, current, and voltage; process operation data such as operating parameters and control commands; event-based data such as fault alarm records and maintenance records; and textual records from some operation and maintenance management systems. With the increasing complexity and automation level of equipment, how to quantitatively characterize and evaluate the health status of equipment across its entire life cycle based on multi-source heterogeneous data, and construct a full life cycle health assessment system that spans the design, operation, and decommissioning stages, has become one of the important research directions in the fields of equipment reliability engineering, condition monitoring and fault prediction, and health management. In practical engineering applications, the process of conducting full lifecycle health assessments of equipment faces several challenges. Firstly, it is difficult to uniformly present multi-source heterogeneous data and health evolution trajectories across the entire lifecycle, and quantitatively trace each lifecycle stage when equipment experiences decommissioning anomalies. Specifically: Firstly, multi-source heterogeneous data is typically stored across different platforms of monitoring, control, and maintenance systems. Existing methods often employ modeling based on a single time series or several independent feature vectors, making it difficult to uniformly represent and correlate the health status of each stage of the equipment's lifecycle within a structured view. Secondly, existing methods generally use historical monitoring data as input to construct forward prediction models, estimating remaining lifespan or health indicators from the current moment to the future. For decommissioning states or target termination states that have already occurred, only post-hoc statistical analysis is typically performed, making it difficult to answer the question of the degree of impact at each stage of the equipment's lifecycle given a decommissioning outcome within the same model framework. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for health assessment of equipment throughout its entire life cycle driven by multi-source heterogeneous data, in order to solve the technical problems mentioned in the background art, such as the difficulty in uniformly presenting multi-source heterogeneous data and health evolution trajectory at the entire life cycle scale, and the inability to quantitatively trace each stage of the life cycle when equipment decommissioning abnormalities occur.

[0004] To achieve the above objectives, the technical solution of the present invention is: a device lifecycle health assessment method driven by multi-source heterogeneous data, comprising: S1. Collect multi-source heterogeneous data from the equipment, divide the entire life cycle of the equipment into several continuous life cycle stage nodes, and construct a life cycle health status time sequence diagram using the life cycle time sequence relationship and stage transition relationship as directed edges. S2. Based on the standard data of the device, combined with the nodes of each life cycle stage of the life cycle health status time sequence diagram, construct the standard health path of the device; calculate the actual health status value of the device in each historical stage based on multi-source heterogeneous data, and map it to the corresponding historical stage node to construct the historical health path of the device. S3. Using the current lifecycle stage node as the starting node, based on historical health paths and multi-source heterogeneous data of the current lifecycle stage, predict the future health status of the equipment and construct a forward predictive health path for the equipment; define the virtual decommissioning final state node of the equipment and use the virtual decommissioning final state node as a constraint condition, perform reverse time series inference on the health status of each lifecycle stage of the equipment based on the lifecycle health status time series diagram, calculate the reverse health status value of each lifecycle stage and its stage contribution, and construct a reverse causal reverse inference health path for the equipment. Among them, the virtual retirement final state node is a graph node in the life cycle health state time sequence diagram that represents the target health state when the equipment ends its life; the equipment reverse causal inverse health path is a path that infers the health state of each life cycle stage of the equipment in reverse order along the life cycle time sequence, starting from the virtual retirement final state node. S4. In the life cycle health status time sequence diagram, based on the equipment standard health path, equipment historical health path, equipment forward predicted health path and equipment reverse causal inverse health path, the four-path health deviation pattern assessment mechanism is used to analyze the deviation pattern distribution between different path health states and generate the equipment's full life cycle joint health assessment results. The four-path health deviation mode assessment mechanism is used to calculate the degree of mutual deviation between each path and to assess and analyze the health status of the equipment throughout its entire life cycle based on the distribution of deviation modes.

[0005] Preferably, in step S1, the lifecycle health status time sequence graph is a directed time sequence graph structure with lifecycle stage nodes as graph nodes and directed edges representing the temporal order and stage transition relationships between adjacent lifecycle stage nodes. It is used to carry multi-source heterogeneous data and corresponding health status for each lifecycle stage. The graph nodes include a stage identifier field, a multi-source heterogeneous data index field, and a stage health status field. The directed edges are used to represent the sequential order of lifecycle stages and the transition relationships between running stages. The edge weights of the directed edges include the time interval between adjacent lifecycle stages and the intensity of health status changes.

[0006] Preferably, in step S3, the virtual decommissioning final state node is used as a constraint boundary to limit the reverse health status of each life cycle stage during the reverse time series inference process; the data structure of the virtual decommissioning final state node includes: a decommissioning stage identifier field, a decommissioning health status field, and a decommissioning constraint parameter field.

[0007] Preferably, in step S3, the reverse time series inference is a health state inference method that starts from the virtual decommissioning final state node, traverses the directed edges of the life cycle health state time series graph in reverse order along the time sequence, and calculates the reverse health state value and stage contribution for each stage. The reverse time series inference is used to quantify the impact of each life cycle stage on the equipment decommissioning health state under given decommissioning constraints. The reverse time series inference is performed on the health state of each life cycle stage of the equipment, and a reverse causal reverse health path is constructed. The specific method includes: starting from the virtual decommissioning final state node, obtaining the previous life cycle stage node adjacent to the current node, calculating the reverse health state value of the previous life cycle stage node based on the health state value of the current node, the edge weight of the directed edge, and the decommissioning constraint parameter field, and writing the reverse health state value into the stage health state field of the corresponding graph node. At the same time, the stage contribution of the stage is calculated based on the deviation between the reverse health state value and the standard health state of the corresponding stage. The reverse causal reverse health path is formed by traversing forward along the directed edges of the life cycle health state time series graph until all life cycle stage nodes are covered.

[0008] Preferably, in S3, the reverse-engineered health status value refers to the health status value of the equipment at each life cycle stage obtained by inferring from the reverse time series under the constraint of the virtual decommissioning final state node, indicating the health level that each life cycle stage should have under the premise of a given decommissioning health status; the stage contribution is a stage-level contribution index determined based on the deviation relationship between the reverse-engineered health status value of each life cycle stage and the standard health status of the corresponding stage, used to quantify the degree of influence of each life cycle stage on the deviation of the virtual decommissioning final state health status; the stage contribution is limited to a value not less than zero according to the normalization rule, and the sum of the stage contributions of all life cycle stages is equal to the total deviation between the decommissioning health status of the virtual decommissioning final state node and the standard decommissioning health status of the equipment.

[0009] Preferably, in S3, the reverse causal health path of the device is a node sequence path formed by sequentially connecting the virtual decommissioning final state node in the life cycle health state time sequence diagram and the life cycle stage nodes associated with it in reverse time. Each life cycle stage node stores the corresponding reverse health state value and stage contribution in the path, which is used to identify the health impact link from the virtual decommissioning final state node back to the initial operation stage of the device, and to identify the key life cycle stages whose contribution to the decommissioning health state exceeds a preset threshold.

[0010] Preferably, in S4, the four-path health deviation pattern assessment mechanism is an assessment mechanism that obtains the stage health status values ​​corresponding to the equipment's standard health path, historical health path, forward predicted health path, and reverse causal inverse health path at the same life cycle stage node, and constructs a path-level deviation feature vector. Based on the path-level deviation feature vector, a comprehensive analysis is performed on the multidimensional deviation relationship between the four paths to determine the deviation pattern distribution of the equipment at each life cycle stage. The path-level deviation feature vector includes node-level deviation indicators, window-level cumulative deviation indicators, and path trend consistency indicators.

[0011] Preferably, in step S4, the deviation pattern distribution is formed by classifying the path-level deviation feature vectors corresponding to each life cycle stage node in the life cycle health status time sequence diagram according to the deviation pattern determined by the four-path health deviation pattern evaluation mechanism, and forming a sequence distribution of the deviation patterns of each life cycle stage node on the life cycle time axis, which is used to map the health status of the device in different life cycle stages.

[0012] On the other hand, the present invention provides a device lifecycle health assessment system driven by multi-source heterogeneous data, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the aforementioned device lifecycle health assessment method driven by multi-source heterogeneous data.

[0013] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: 1. In this invention, four types of health paths are constructed based on the life cycle health status time sequence diagram: standard health path of equipment, historical health path of equipment, forward predictive health path of equipment, and reverse causal inference health path of equipment. The health changes of equipment from commissioning to decommissioning can be presented simultaneously in the same data structure, making the health status and interrelationships of equipment at different stages of the entire life cycle more intuitive and clarifying the overall health trend and stage deviation of equipment. 2. In this invention, by introducing a virtual decommissioning final state node and combining it with reverse time series inference and a four-path health deviation pattern assessment mechanism, the source of equipment decommissioning results or significant deterioration results can be quantitatively traced at the stage level. The decommissioning health deviation is decomposed into different life cycle stages, and the key stages that have a significant impact on the decommissioning results are identified. Attached Figure Description

[0014] Figure 1 This is a flowchart of one embodiment of the present invention. Detailed Implementation

[0015] Example 1, as Figure 1 As shown, the present invention proposes a device lifecycle health assessment method driven by multi-source heterogeneous data, the specific implementation steps of which are as follows: S1. Collect multi-source heterogeneous data from the equipment, divide the entire life cycle of the equipment into several continuous life cycle stage nodes, and construct a life cycle health status time sequence diagram using the life cycle time sequence relationship and stage transition relationship as directed edges. S2. Based on the standard data of the device, combined with the nodes of each life cycle stage of the life cycle health status time sequence diagram, construct the standard health path of the device; calculate the actual health status value of the device in each historical stage based on multi-source heterogeneous data, and map it to the corresponding historical stage node to construct the historical health path of the device. S3. Using the current lifecycle stage node as the starting node, based on historical health paths and multi-source heterogeneous data of the current lifecycle stage, predict the future health status of the equipment and construct a forward predictive health path for the equipment; define the virtual decommissioning final state node of the equipment and use the virtual decommissioning final state node as a constraint condition, perform reverse time series inference on the health status of each lifecycle stage of the equipment based on the lifecycle health status time series diagram, calculate the reverse health status value of each lifecycle stage and its stage contribution, and construct a reverse causal reverse inference health path for the equipment. Among them, the virtual retirement final state node is a graph node in the life cycle health state time sequence diagram that represents the target health state when the equipment ends its life; the equipment reverse causal inverse health path is a path that infers the health state of each life cycle stage of the equipment in reverse order along the life cycle time sequence, starting from the virtual retirement final state node. S4. In the life cycle health status time sequence diagram, based on the equipment standard health path, equipment historical health path, equipment forward predicted health path and equipment reverse causal inverse health path, the four-path health deviation pattern assessment mechanism is used to analyze the deviation pattern distribution between different path health states and generate the equipment's full life cycle joint health assessment results. The four-path health deviation mode assessment mechanism is used to calculate the degree of mutual deviation between each path and to assess and analyze the health status of the equipment throughout its entire life cycle based on the distribution of deviation modes.

[0016] In this embodiment S1, the lifecycle health status time sequence graph is a directed time sequence graph structure with lifecycle stage nodes as graph nodes and directed edges representing the temporal order relationship and stage transition relationship between adjacent lifecycle stage nodes. It is used to carry multi-source heterogeneous data and corresponding health status of each lifecycle stage. The graph nodes include a stage identifier field, a multi-source heterogeneous data index field, and a stage health status field. The directed edges are used to represent the sequential order of lifecycle stages and the transition relationship between running stages. The edge weights of the directed edges include the time interval between adjacent lifecycle stages and the intensity of health status change.

[0017] In this embodiment S1, a multi-source heterogeneous data acquisition and aggregation environment is constructed for the target equipment. This environment integrates operational data, maintenance data, and environmental data generated at different stages of the equipment's lifecycle, including factory shipment, installation and commissioning, trial operation, stable operation, performance degradation, high failure rate, and decommissioning preparation. Operational data includes online monitoring data such as power, speed, temperature, pressure, vibration amplitude, and number of start-stop cycles. Maintenance data includes inspection records, repair records, spare parts replacement records, fault alarm and shutdown event records, etc. Environmental data includes ambient temperature, humidity, external load condition level, and upstream process conditions. After integration, the multi-source heterogeneous data undergoes preprocessing operations such as timestamp alignment, unit dimension unification, outlier removal, and missing segment completion to establish an original data time series with the equipment timeline as the main line.

[0018] In this embodiment S1, the entire lifecycle of the equipment is considered as a continuous splicing of several stages with relatively stable health and operating conditions. The time axis is divided by preset rules or data-driven methods to obtain a series of lifecycle stage nodes. The division rules of the lifecycle stages can be preset based on the equipment technical data and operation and maintenance strategies, such as dividing them into the factory acceptance stage, commissioning and trial operation stage, stable operation stage, slow performance degradation stage, high failure rate stage, and decommissioning preparation stage. Alternatively, they can be statistically segmented based on the health status change trend, failure event density, and operating condition switching frequency in historical operating data. Each lifecycle stage node corresponds to a time interval. Within this time interval, the dominant operating condition type, health status level, and maintenance strategy of the equipment remain basically consistent, which facilitates the abstraction of the continuous time axis into a set of discrete stage node sequences and provides clear stage boundaries for subsequent multi-source heterogeneous data indexing and health status fields at the node layer.

[0019] In this embodiment S1, a node data structure in the lifecycle health status time sequence diagram is constructed based on lifecycle stage nodes. Each stage diagram node includes at least a stage identifier field, a multi-source heterogeneous data index field, and a stage health status field. The stage identifier field is used to uniquely mark the position and attributes of the stage in the entire lifecycle of the device, including the stage number, stage type label, and start and end timestamps of the stage. The multi-source heterogeneous data index field is used to associate the preprocessed original data storage location, rather than directly storing all the original data in the node. It is achieved by recording the time sequence database index number, file path, dataset identifier, or data shard primary key, etc. This method establishes a mapping relationship between operational data, maintenance data, and environmental data and the corresponding time windows of each stage. This ensures that the original data can be quickly retrieved when a node is queried, while avoiding a large amount of redundancy in the graph structure. The stage health status field is used to store the comprehensive health assessment results of that stage. It can take the form of health score, health level, fault risk level, degradation rate level, or combination of key component health indicators. This field is written by the subsequent health status calculation module after evaluation based on the data index corresponding to that stage. However, the field position is reserved for it in advance during the construction of the life cycle health status sequence graph, so that the node field layout can be completed at the beginning of the graph structure creation.

[0020] In this embodiment S1, the temporal sequence relationship and stage transition relationship between adjacent lifecycle stage nodes are explicitly modeled as directed edges, and edge weight information reflecting the evolution rhythm and health change magnitude is added to the directed edges, thereby forming a complete lifecycle health state temporal graph topology. The starting point of the directed edge is the previous lifecycle stage node, and the ending point is the next lifecycle stage node. The direction is consistent with the actual time progression direction of the device's operation, used to describe the sequence of lifecycle stages and the transition logic from one stage operating state to the next stage operating state. The edge weight of the directed edge includes the time interval between adjacent stages and the intensity of health state change. The interval is calculated based on the difference between the start time of the next stage and the end time of the previous stage. It is used to characterize the tightness of the connection between stages and the relative weight of the stage duration in the overall life cycle. The intensity of health status change is quantified based on the difference in health status fields between the two stages. It can be used by means of health score difference, health level jump amplitude or key health indicator change amplitude, etc., to characterize the degree of health status change and degradation trend of the equipment at the stage switching point. Through the construction of the above nodes and directed edges, the entire life cycle of the equipment is organized into a life cycle health status time sequence diagram with stage nodes as carriers and time sequence and state transition as the skeleton.

[0021] In this embodiment S2, a standard health path for the equipment is constructed based on the equipment's standard data. The standard data includes the design life, rated operating range, allowable deviation range of key performance indicators, performance curves generated by factory testing, health assessment specifications provided by the manufacturer, and industry or enterprise-specific health judgment standards for different lifecycle stages, all given during the equipment design phase. In implementation, firstly, based on the stage type, start and end time, and corresponding operating condition category of each lifecycle stage node in the lifecycle health status time sequence diagram, standard health status information matching the stage type is retrieved from the standard data. For example, for the trial operation stage, stable operation stage, and performance degradation stage, the corresponding target health level, allowable performance degradation ratio, and allowable failure rate range are extracted respectively. Then, this standard health status information is arranged according to the order of the lifecycle stage nodes, and the standard health status value corresponding to each stage is written into the standard health status field or standard health reference field of that stage node, forming a standard health status sequence distributed along the time sequence. This sequence constitutes the equipment's standard health path, which is used to characterize the health level that the equipment should achieve in each lifecycle stage under ideal operating conditions and standardized operation and maintenance conditions, as well as the reasonable range of health status changes between stages.

[0022] In this embodiment S2, based on the multi-source heterogeneous data associated with each historical stage node in the lifecycle health status time sequence diagram, a historical health path for the device is constructed. Within the time interval corresponding to each lifecycle stage node, the operating data, maintenance data, and environmental data within that stage are extracted from the data storage location pointed to by the multi-source heterogeneous data index field. Statistical aggregation and feature extraction operations are performed on key monitoring indicators, such as calculating the mean, extreme values, fluctuation amplitude, and number of limit violations for indicators like temperature, vibration, pressure, and efficiency within that stage; statistically analyzing the occurrence, duration, and severity of fault events within that stage; and extracting representative quantitative indicators of load level and environmental conditions within that stage. After setting the above feature set, a pre-configured health assessment model or health scoring rule is invoked to integrate each feature according to the set weights, thereby obtaining the actual health status value of this life cycle stage. The actual health status value can be in the form of continuous scoring or discrete health level, and the health status value is written into the stage health status field in the life cycle stage node. By repeating the above health status calculation and field writing process on all historical stage nodes, an actual health status sequence consistent with the time order is formed in the life cycle health status time sequence diagram. This sequence is the equipment historical health path, which is used to objectively reflect the true health evolution trajectory of the equipment in each life cycle stage under actual operation and maintenance conditions.

[0023] In this embodiment S2, both the standard health path and the historical health path of the equipment are organized based on the same lifecycle health status time sequence diagram. The two paths correspond to different health status fields in the same set of lifecycle stage nodes. The standard health path focuses on representing the target health level that each stage should have under theoretical design conditions and standardized operation and maintenance conditions, while the historical health path focuses on representing the actual health performance of each stage driven by real operation data. By storing the standard health status value and the actual health status value at the same lifecycle stage node, the deviation between the target health status and the actual health status of the stage can be intuitively compared at the node level, and the changing trend of the deviation of each stage can be observed in the time axis dimension. For example, if the actual health status is significantly lower than the standard health status in a certain stage node, it indicates that there is a problem of insufficient health status or accelerated degradation in that stage. If the two are close in another stage node, it indicates that the operating status of that stage is more consistent with the expectation. This dual-path representation method based on the same graph structure enables the standard health trajectory and historical health trajectory of the equipment throughout its entire lifecycle to be stored and compared in a unified data form, which is convenient for analyzing the deviation of the equipment health status and the health evolution characteristics at different time scales and different stage granularities.

[0024] In this embodiment S3, the virtual decommissioning final state node is used as a constraint boundary to limit the reverse health status of each life cycle stage during the reverse time series inference process; the data structure of the virtual decommissioning final state node includes: a decommissioning stage identifier field, a decommissioning health status field, and a decommissioning constraint parameter field.

[0025] In this embodiment S3, the current lifecycle stage node is used as the starting node for forward prediction. Multi-source heterogeneous data for this stage is combined with historical health paths to construct an input feature sequence for health status prediction. The multi-source heterogeneous data for the current lifecycle stage includes the operating load level, key monitoring indicator trends, maintenance operation frequency, and environmental condition fluctuations within this stage. The historical health path provides the health status evolution trajectory from equipment commissioning to the current stage. By aligning these two parts of information on the time axis and constructing a feature sequence with time-series labels, the inertial trend of health status changes during long-term use of the equipment can be reflected. The cumulative effect of recent changes in operating conditions; based on this, a health status prediction model is invoked to estimate the health status of several future life cycle stages. The health status prediction model can adopt a time series-based degradation trend fitting method, a health level discrimination method based on stage characteristics, or an integrated prediction method combining both. The predicted health status value of each future life cycle stage is written into the corresponding stage node health status prediction field, and these predicted health status values ​​are concatenated in the order of life cycle stages to form a forward predicted health path, which is used to depict the health status evolution trend of the equipment from the current stage to the future stage in the positive direction of the time axis.

[0026] In this embodiment S3, a virtual decommissioning final state node is introduced to explicitly model the termination state of the entire equipment lifecycle. This node serves as the constraint boundary for reverse time series inference, constructing a reverse causal health path for the equipment. In the lifecycle health state time series diagram, the virtual decommissioning final state node is inserted as a new stage node after the last actual lifecycle stage node, connected by a directed edge pointing from the last actual stage to the virtual decommissioning final state node, representing the temporal sequence of the equipment's evolution from the final operating stage to the decommissioning state. During reverse time series inference, the health state field and constraint parameter field of the virtual decommissioning final state node are used as fixed known conditions, and the path is constructed along the lifecycle health sequence. In the time sequence graph of the decommissioning state, the nodes of each life cycle stage are traversed sequentially in the reverse direction of the directed edges. At each stage node, based on the decommissioning health status, the operational characteristics of the stage, and the stage's position on the time axis, the backward health status value that the stage should possess under the premise of meeting the decommissioning conditions is estimated. Based on the change magnitude of the backward health status value relative to the backward health status values ​​of other stages, the degradation rate within the stage, and the density of key events, the stage contribution of the stage to the final decommissioning state is calculated. Through the above reverse traversal and backward calculation, a stage sequence containing the backward health status value and stage contribution is formed in the reverse direction of the time axis, namely the reverse causal backward health path, which is used to characterize the distribution of the influence of each life cycle stage on the formation process of the decommissioning state from the perspective of the decommissioning endpoint.

[0027] In this embodiment S3, the data structure of the virtual decommissioning final state node consists of a decommissioning stage identifier field, a decommissioning health status field, and a decommissioning constraint parameter field. The decommissioning stage identifier field is used to uniquely identify the position and role of the virtual node in the life cycle health status time sequence diagram, including information such as the decommissioning stage number, decommissioning stage type, and expected decommissioning time marker, thereby distinguishing the node from other actual operating stage nodes in the graph structure. The decommissioning health status field is used to record the target health status when the equipment reaches the decommissioning conditions, including health score thresholds, performance index lower limits, and fault risk level upper limits, etc., and is used as a fixed reference for the endpoint health status in reverse time series inference, so that the health status values ​​of each stage obtained by reverse inference are numerically consistent with the final decommissioning status. The state remains consistent or gradually converges to the final state within the allowable deviation range; the decommissioning constraint parameter field is used to centrally store engineering and management constraint information related to decommissioning decisions, including the maximum allowable cumulative runtime, the maximum allowable cumulative number of failures, the maintenance cost ratio threshold, risk control indicators required by safety specifications, and decommissioning strategy parameters formulated internally by the enterprise. In the reverse time series inference process, the decommissioning constraint parameter field limits the feasible combination space between the health state of each stage and factors such as runtime, failure frequency, and maintenance cost during the reverse calculation, ensuring that the obtained reverse health state value and stage contribution meet the decommissioning conditions set, so that the virtual decommissioning final state node in the graph structure serves both as the function of expressing the endpoint health state and as the function of the decommissioning constraint boundary.

[0028] In this embodiment S3, the lifecycle health status sequence diagram serves as a unified carrying structure, simultaneously storing three types of health status information: historical health paths, forward predicted health paths, and reverse causal inverse health paths. In specific implementation, for nodes in the same lifecycle stage, actual health status, predicted health status, and reverse inverse health status fields are set to distinguish health status values ​​from different sources. The actual health status field is written from the historical health path calculation results, reflecting the health evolution that the device has already undergone. The predicted health status field is written from the forward prediction calculation results, describing the health status that may be reached in future stages under the current stage's characteristics. The reverse health status field is written from the reverse time series inference results with the virtual retirement final state node as the boundary. It is used to characterize the target health level of each stage in the causal chain starting from the retirement final state constraint. The stage contribution can be attached and stored as an independent field in the corresponding life cycle stage node, representing the influence weight of that stage in the reverse causal chain. By storing the three types of health status fields and stage contribution in columns in the same graph structure, it can be ensured that historical results, forward prediction results and reverse inference results are independent and comparable at the node level, so that the health status information of the whole life cycle is clearly divided in structure, reducing data coverage and confusion between different paths.

[0029] In this embodiment S3, the reverse time series inference is a health state inference method that starts from the virtual decommissioning final state node, traverses the directed edges of the life cycle health state time series graph in reverse chronological order, and calculates the reverse health state value and stage contribution for each stage. The reverse time series inference is used to quantify the impact of each life cycle stage on the equipment decommissioning health state under given decommissioning constraints. The reverse time series inference is performed on the health state of each life cycle stage of the equipment, and a reverse causal reverse health path is constructed. The specific method includes: starting from the virtual decommissioning final state node, obtaining the previous life cycle stage node adjacent to the current node, calculating the reverse health state value of the previous life cycle stage node based on the health state value of the current node, the edge weight of the directed edge, and the decommissioning constraint parameter field, and writing the reverse health state value into the stage health state field of the corresponding graph node. At the same time, the stage contribution of the stage is calculated based on the deviation between the reverse health state value and the standard health state of the corresponding stage. The reverse causal reverse health path is formed by traversing forward along the directed edges of the life cycle health state time series graph until all life cycle stage nodes are covered.

[0030] In this embodiment S3, the reverse time series inference uses the virtual decommissioning final state node as the starting point for calculation. The health status field and decommissioning constraint parameter field of this node in the life cycle health status time series diagram are used as known boundary conditions. The nodes of each life cycle stage are traversed in the reverse direction of the directed edges in the graph structure. Specifically, the previous life cycle stage node directly connected to the virtual decommissioning final state node is first determined. This previous stage node is regarded as the current calculation object. By reading the edge weight information of the directed edge connecting the two nodes, including the time interval between adjacent stages and the intensity of health status change, and combining the health status value and decommissioning constraint parameters of the virtual decommissioning final state node, the target reverse health level of the previous stage node is estimated. In the case of multiple predecessor paths or multiple stage transition relationships, a representative path can be selected for reverse traversal based on the main time axis order determined when dividing the life cycle or based on the priority of stage type, so as to ensure that the reverse time series inference unfolds along the main evolution link of the equipment life cycle and avoid the situation where the reverse traversal path is inconsistent with the main life cycle sequence.

[0031] In this embodiment S3, whenever a new lifecycle stage node is traversed backwards, the calculation of the reverse health status value is performed based on the mapping relationship between the current node's health status value, the directed edge weights, and the decommissioning constraint parameters. During the calculation, firstly, based on the change magnitude corresponding to the change intensity of the current node's health status value and the health status change intensity in the directed edges, combined with the time interval in the edge weights, the current node's health status is shifted forward along the time axis to the estimated target health status value of the previous stage, obtaining the initial reverse health status value of this stage under the unconstrained assumption. Then, this initial reverse health status value is compared with the decommissioning constraint parameters... The system compares the initial reverse health status value with the constraints given in the section, such as the cumulative runtime limit, the maximum allowable degradation range, and the upper limit of the number of failures. When the initial reverse health status value exceeds the allowable range, the reverse result is corrected by truncation, compression, or weight redistribution. This ensures that the corrected reverse health status value matches the evolution gradient depicted by the current node's health status and edge weights, while not violating the boundary conditions set by the retirement constraint parameters. The corrected reverse health status value is written into the reverse health status field of the node in this lifecycle stage, which is used to record the health level that this stage should bear in the causal chain under the retirement boundary conditions.

[0032] In this embodiment S3, after obtaining the reverse health status value of a certain lifecycle stage node, the stage contribution is calculated by the deviation between the reverse health status value of that stage and the standard health status, so as to quantify the degree of influence of that stage on the equipment retirement health status. Specifically, firstly, the standard health status field stored in the stage node is read, the reverse health status value is compared with the standard health status, and the absolute deviation or relative deviation quantification value is obtained. Then, combined with the equipment type, lifecycle stage attributes and preset weight rules, the deviation quantification value is mapped to the stage contribution index. The stage contribution can be represented by normalized weight form or hierarchical scoring form, which is used to represent the impact of the deviation of the health status of that stage from the standard trajectory on the final retirement state. The impact ratio; after the stage contribution is written into the stage contribution field of the stage node, the current stage node is regarded as the new "current node". The process of searching for the previous life cycle stage node from the node along the directed edge and performing the reverse calculation of health status value and stage contribution is repeated until all life cycle stage nodes in the life cycle health status time sequence diagram are traversed. Through the above reverse traversal and stage-by-stage calculation, an ordered node sequence containing the reverse health status value and corresponding stage contribution of all life cycle stage nodes is formed in the reverse direction of the time axis. This ordered node sequence constitutes the reverse causal reverse health path of the equipment, which is used to comprehensively characterize the cumulative impact distribution of each life cycle stage on the decommissioning health status under the decommissioning constraint condition within a unified path structure.

[0033] In this embodiment S3, the reverse health status value refers to the health status value of the equipment at each life cycle stage obtained by inferring from the reverse time series under the constraint of the virtual decommissioning final state node. It identifies the health level that each life cycle stage should have under the premise of a given decommissioning health status. The stage contribution is a stage-level contribution index determined based on the deviation relationship between the reverse health status value of each life cycle stage and the standard health status of the corresponding stage. It is used to quantify the degree of influence of each life cycle stage on the deviation of the virtual decommissioning final state health status. The stage contribution is limited to a value not less than zero according to the normalization rule, and the sum of the stage contributions of all life cycle stages is equal to the total deviation between the decommissioning health status of the virtual decommissioning final state node and the standard decommissioning health status of the equipment.

[0034] In this embodiment S3, the reverse health status value refers to the target health status value calculated for each life cycle stage through a reverse time series inference process, based on the retirement health status and retirement constraint parameters given by the virtual retirement final state node. It is used to characterize the health level that each life cycle stage should maintain under the condition that the given retirement conditions remain unchanged. To this end, when completing the reverse time series inference, a set of reverse health status data corresponding to each life cycle stage node is generated and written into the reverse health status field of that stage node. This field is independent of the actual health status field recorded in the historical health path, the standard health status field recorded in the standard health path, and the predicted health status field recorded in the forward predicted health path. The reverse health status value is different in meaning from the health status value observed during actual equipment operation and also different from the predicted health status value extrapolated from historical data. It focuses more on the "ought" health level calculated backward from the virtual retirement final state node along the life cycle health status time series diagram, given that the retirement health status and retirement constraint parameters have been determined. It is used to characterize the reasonable health status configuration of each life cycle stage during the whole life cycle degradation process from the perspective of causal analysis.

[0035] In this embodiment S3, the stage contribution is a stage-level contribution index determined based on the deviation relationship between the reverse health status value of each life cycle stage and the standard health status of the corresponding stage. It is used to quantify the degree of influence of each life cycle stage on the deviation of the virtual decommissioning final health status. Specifically, for each life cycle stage node, the reverse health status value and the standard health status value of the stage are first obtained, the deviation between the two is calculated, and then the deviation is converted into a non-negative stage contribution value according to the preset mapping rule. The stage contribution value is written into the stage contribution field of the life cycle stage node to represent the influence share of the stage in the process of the decommissioning health status deviating from the standard decommissioning health status of the equipment. In order to ensure that the stage contribution has a clear physical meaning as a whole, the stage contribution is limited to a value not less than zero according to the normalization rule.

[0036] In this embodiment S3, the sum of the stage contributions of all lifecycle stage nodes is equal to the total deviation between the decommissioning health status of the virtual decommissioning final state node and the standard decommissioning health status of the equipment. The total deviation can be measured in different ways according to actual application needs: In one case, the total deviation can be represented by the direct difference or absolute difference between the decommissioning health status score and the standard decommissioning health score; in another case, when the decommissioning health status consists of multiple sub-indicators, the sub-indicators can be weighted and aggregated according to preset weights, and the difference between the aggregated health index and the standard decommissioning status can be calculated as the total deviation; in yet another case, the decommissioning health status and the standard decommissioning health status can be mapped to a unified risk measurement scale, and the total deviation can be defined by the risk measurement difference. The above different implementations only differ in the specific measurement method of the total deviation, and all meet the definition requirements of the total deviation between the decommissioning health status and the standard decommissioning health status of the equipment. On this basis, the total deviation is proportionally allocated to the stage contribution corresponding to each lifecycle stage node through normalization rules, so that the stage contribution can be numerically interpreted as the share of the deviation between the decommissioning health status and the standard decommissioning health status in that stage.

[0037] In this embodiment S3, the reverse causal health path of the device is a node sequence path formed by sequentially connecting the virtual decommissioning final state node in the life cycle health state time sequence diagram and the life cycle stage nodes associated with it in reverse time. Each life cycle stage node stores the corresponding reverse health state value and stage contribution in the path, which is used to identify the health impact link from the virtual decommissioning final state node back to the initial operation stage of the device, and to identify the key life cycle stage whose contribution to the decommissioning health state exceeds a preset threshold.

[0038] In this embodiment S3, the reverse causal health path is a node sequence path constructed based on the lifecycle health state time sequence diagram. This path starts from the virtual decommissioning final state node and connects the adjacent lifecycle stage nodes in reverse chronological order until it connects to the lifecycle stage node corresponding to the initial operation stage of the equipment. During the path construction process, the previous lifecycle stage node with a direct temporal relationship to the current node is selected sequentially in the opposite direction of the directed edge in the lifecycle health state time sequence diagram. Each selected lifecycle stage node is appended to the end of the node sequence, thereby forming an ordered node sequence starting from the virtual decommissioning final state node and arranged in reverse chronological order of the actual operation time of the equipment. In terms of data organization, the reverse causal health path can be represented as a list of node identifiers arranged in reverse chronological order. Each node identifier corresponds one-to-one with a lifecycle stage node in the lifecycle health state time sequence diagram. The node carries the reverse health state value and stage contribution field written in the aforementioned reverse calculation process in the path, which are used to simultaneously express the target health level of the stage under the decommissioning constraint conditions and the contribution share of the stage to the decommissioning health state deviation under the same path structure.

[0039] In this embodiment S3, the reverse causal health path is used to identify the health impact chain from the virtual decommissioning final state node back to the initial operation stage of the equipment. By associating the reverse health status value with the stage contribution on the node sequence in reverse time, it can intuitively reflect that the decommissioning health status is a causal chain formed by the gradual accumulation of health deviations in each life cycle stage. When analyzing this path, by traversing the stage contribution field of each life cycle stage node in the node sequence, the stage contribution is compared with a preset contribution threshold. When the stage contribution of a certain life cycle stage node is greater than or equal to the preset threshold, the stage is marked as a critical life cycle stage. A critical stage identifier field can be added to the node or a critical stage list can be recorded externally to distinguish between general stages and stages that have a significant impact on the decommissioning health status deviation. The preset contribution threshold can be set according to the equipment operation and maintenance strategy, risk control requirements, or statistical analysis results. For example, the upper quantile of the stage contribution distribution can be selected as the threshold, or the minimum stage contribution corresponding to the cumulative contribution reaching a certain proportion of the total deviation can be used as the threshold.

[0040] In this embodiment S4, the four-path health deviation pattern assessment mechanism obtains the stage health status values ​​corresponding to the equipment's standard health path, historical health path, forward predicted health path, and reverse causal inverse health path at the same life cycle stage node, and constructs an assessment mechanism for path-level deviation feature vectors. Based on the path-level deviation feature vectors, a comprehensive analysis is performed on the multidimensional deviation relationship between the four paths to determine the deviation pattern distribution of the equipment at each life cycle stage. The path-level deviation feature vectors include node-level deviation indicators, window-level cumulative deviation indicators, and path trend consistency indicators.

[0041] In this embodiment S4, for each life cycle stage node in the same life cycle health status time series diagram, the stage health status values ​​corresponding to the standard health path, historical health path, forward predicted health path, and reverse causal inverse health path are uniformly obtained. Based on this set of multi-source health status data, a path-level deviation feature vector is constructed. Specifically, for any life cycle stage node, the standard health status field, actual health status field, predicted health status field, and reverse health status field in the node are read, and the values ​​of these four types of health status are aligned on the same node to form the basic health status set of the node. On this basis, the difference relationship between the above four types of health status is quantified by a preset deviation analysis rule, and node-level deviation index, window-level cumulative deviation index, and path trend consistency index are generated in sequence. The three types of indexes together form the path-level deviation feature vector of the node, which is used to characterize the multidimensional deviation relationship between the four paths in the same feature space. Repeating the above operation for all life cycle stage nodes in the life cycle health status time series diagram can obtain a set of path-level deviation feature vectors distributed along the life cycle time axis.

[0042] In this embodiment S4, the node-level deviation index is used to describe the instantaneous deviation relationship between the four health paths at a single life cycle stage node. In implementation, the difference measures between the historical health state value, the predicted health state value, and the reverse-engineered health state value and the standard health state value can be calculated based on the standard health state value. Additionally, as needed, difference measures between the historical health state value and the predicted health state value, and between the historical health state value and the reverse-engineered health state value, can be provided to form a subset of node-level deviation indices, used to characterize the comprehensive deviation of this stage from four perspectives: "desired level," "measured level," "future trend," and "desired level under retirement constraints." The window-level cumulative deviation index is used to reflect the cumulative effect of node-level deviation within a time window centered on or starting from the current life cycle stage node. In implementation, a time window containing several adjacent life cycle stage nodes can be selected, and the window... The node-level deviation indicators of each node are accumulated, averaged, or extreme valued to form indicators such as window-level deviation intensity and window-level deviation duration, which are used to distinguish between short-term sudden deviations and long-term persistent deviations. The path trend consistency indicator is used to analyze whether the change direction and change pattern of the four paths are consistent in the time dimension near this stage. For example, a trend symbol sequence can be constructed based on the change direction of the health status of several adjacent nodes, or a trend consistency score can be constructed based on the correlation coefficient of the change in health status within adjacent windows. This characterizes the degree of consistency or deviation between the standard health path and the historical health path, the forward predicted health path, and the reverse causal inverse health path at the trend level. The above-mentioned node-level deviation indicators, window-level cumulative deviation indicators, and path trend consistency indicators together form a path-level deviation feature vector, which is used to fully express the multidimensional deviation relationship between the health status of the four paths at the node of this life cycle stage in the feature space.

[0043] In this embodiment S4, the four-path health deviation mode assessment mechanism comprehensively analyzes the multidimensional deviation relationship between the four health paths based on the path-level deviation feature vectors of nodes at each life cycle stage, and determines a deviation mode category for each life cycle stage node in the feature space. In implementation, the path-level deviation feature vectors can be judged according to a pre-set rule base. For example, when both the node-level deviation index and the window-level cumulative deviation index are at low levels and the path trend consistency index indicates that the historical, predicted, and standard path trends are basically consistent, the node is classified into a deviation mode where the health status is basically stable. When the historical health status value has a long-term negative deviation from the standard health status value and the window-level cumulative deviation index is at a high level, the node is classified into a continuous degradation deviation mode. When the historical health status value is close to the standard health status value, but the forward predicted health status value shows a significant unfavorable deviation relative to the standard health status value, the node is classified into the future risk warning deviation mode. When the reversed health status value is significantly higher than the standard health status value, but the historical health status value is consistently low, the node is classified into the retirement pressure concentration deviation mode. Alternatively, cluster analysis or other pattern recognition algorithms can be used to automatically classify several deviation mode categories based on the aggregation of path-level deviation feature vectors in the multidimensional feature space. Through the above rule-based discrimination or cluster analysis mechanism, a deviation mode identifier is assigned to each life cycle stage node in the life cycle health status time series diagram, so that the path-level deviation feature vector of each node corresponds to a deviation mode category under the four-path health deviation mode evaluation mechanism.

[0044] In this embodiment S4, the deviation pattern distribution is to classify the path-level deviation feature vectors corresponding to each life cycle stage node in the life cycle health status time sequence diagram according to the deviation pattern determined by the four-path health deviation pattern evaluation mechanism, and form a sequence distribution of the deviation patterns of each life cycle stage node on the life cycle time axis, which is used to map the health status of the device in different life cycle stages.

[0045] In this embodiment S4, the deviation pattern distribution is achieved by classifying the path-level deviation feature vectors corresponding to each life cycle stage node in the life cycle health status time sequence diagram according to the deviation patterns determined by the four-path health deviation pattern assessment mechanism, and forming a sequence distribution of the deviation patterns belonging to each life cycle stage node on the life cycle time axis. Specifically, each life cycle stage node in the life cycle health status time sequence diagram is traversed in chronological order, the deviation pattern identifier of each node is read, and the deviation pattern identifier is arranged in chronological order to obtain a deviation pattern sequence covering the initial operation stage of the equipment to the virtual retirement final state node. This deviation pattern sequence can be regarded as a patterned mapping form of the equipment's full life cycle health status, used to reflect the health deviation state of the equipment at different life cycle stages. For example, by observing the continuous distribution segment of a certain type of deviation pattern in the deviation pattern sequence, it can be identified that the equipment is in a long-term continuous degradation, future risk warning, or concentrated retirement pressure mode within a specific time interval. Thus, the spatial distribution and temporal evolution law of the equipment's full life cycle health status are intuitively presented in the joint health assessment results, providing an interpretable pattern-level health assessment basis for operation and maintenance decisions, life management, and retirement strategy formulation.

[0046] Example 2: The present invention proposes a device lifecycle health assessment system driven by multi-source heterogeneous data, which is applied to the device lifecycle health assessment method driven by multi-source heterogeneous data proposed in Example 1. It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the device lifecycle health assessment method driven by multi-source heterogeneous data in Example 1.

[0047] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for device lifecycle health assessment driven by multi-source heterogeneous data, characterized in that, Includes the following steps: S1. Collect multi-source heterogeneous data from the equipment, divide the entire life cycle of the equipment into several continuous life cycle stage nodes, and construct a life cycle health status time sequence diagram using the life cycle time sequence relationship and stage transition relationship as directed edges. S2. Based on the standard data of the device, combined with the nodes of each life cycle stage of the life cycle health status time sequence diagram, construct the standard health path of the device; calculate the actual health status value of the device in each historical stage based on multi-source heterogeneous data, and map it to the corresponding historical stage node to construct the historical health path of the device. S3. Using the current lifecycle stage node as the starting node, based on historical health paths and multi-source heterogeneous data of the current lifecycle stage, predict the future health status of the equipment and construct a forward predictive health path for the equipment; define the virtual decommissioning final state node of the equipment and use the virtual decommissioning final state node as a constraint condition, perform reverse time series inference on the health status of each lifecycle stage of the equipment based on the lifecycle health status time series diagram, calculate the reverse health status value of each lifecycle stage and its stage contribution, and construct a reverse causal reverse inference health path for the equipment. Among them, the virtual retirement final state node is a graph node in the life cycle health state time sequence diagram that represents the target health state when the equipment ends its life; the equipment reverse causal inverse health path is a path that infers the health state of each life cycle stage of the equipment in reverse order along the life cycle time sequence, starting from the virtual retirement final state node. S4. In the life cycle health status time sequence diagram, based on the equipment standard health path, equipment historical health path, equipment forward predicted health path and equipment reverse causal inverse health path, the four-path health deviation pattern assessment mechanism is used to analyze the deviation pattern distribution between different path health states and generate the equipment's full life cycle joint health assessment results. The four-path health deviation mode assessment mechanism is used to calculate the degree of mutual deviation between each path and to assess and analyze the health status of the equipment throughout its entire life cycle based on the distribution of deviation modes.

2. The method for device lifecycle health assessment driven by multi-source heterogeneous data according to claim 1, characterized in that: In S1, the lifecycle health status time sequence graph is a directed time sequence graph structure with lifecycle stage nodes as graph nodes and directed edges representing the temporal order and stage transition relationships between adjacent lifecycle stage nodes. It is used to carry multi-source heterogeneous data and corresponding health status for each lifecycle stage. The graph nodes include a stage identifier field, a multi-source heterogeneous data index field, and a stage health status field. The directed edges are used to represent the sequential order of lifecycle stages and the transition relationships between running stages. The edge weights of the directed edges include the time interval between adjacent lifecycle stages and the intensity of health status changes.

3. The method for device lifecycle health assessment driven by multi-source heterogeneous data according to claim 2, characterized in that: In S3, the virtual retirement final state node is used as a constraint boundary to limit the reverse health status of each life cycle stage during the reverse time series inference process. The data structure of the virtual decommissioning final state node includes: a decommissioning stage identifier field, a decommissioning health status field, and a decommissioning constraint parameter field.

4. The method for device lifecycle health assessment driven by multi-source heterogeneous data according to claim 3, characterized in that: In S3, the reverse time series inference is a health state inference method that starts from the virtual decommissioning final state node, traverses the directed edges of the life cycle health state time series graph in reverse chronological order along each life cycle stage node, and calculates the reverse health state value and stage contribution for each stage. The reverse time series inference is used to quantify the impact of each life cycle stage on the equipment decommissioning health state under given decommissioning constraints. The reverse time series inference is performed on the health state of each life cycle stage of the equipment, and a reverse causal reverse health path is constructed. The specific method includes: starting from the virtual decommissioning final state node, obtaining the previous life cycle stage node adjacent to the current node, calculating the reverse health state value of the previous life cycle stage node based on the health state value of the current node, the edge weight of the directed edge, and the decommissioning constraint parameter field, and writing the reverse health state value into the stage health state field of the corresponding graph node. At the same time, the stage contribution of the stage is calculated based on the deviation between the reverse health state value and the standard health state of the corresponding stage. The reverse causal reverse health path is formed by traversing forward along the directed edges of the life cycle health state time series graph until all life cycle stage nodes are covered.

5. The method for device lifecycle health assessment driven by multi-source heterogeneous data according to claim 4, characterized in that: In S3, the reverse health status value refers to the health status value of the equipment at each life cycle stage obtained by inferring the equipment through reverse time series under the constraint of the virtual decommissioning final state node. It identifies the health level that each life cycle stage should have under the premise of a given decommissioning health status. The stage contribution is a stage-level contribution index determined based on the deviation relationship between the reverse health status value of each life cycle stage and the standard health status of the corresponding stage. It is used to quantify the degree of influence of each life cycle stage on the deviation of the virtual decommissioning final state health status. The stage contribution is limited to a value not less than zero according to the normalization rule, and the sum of the stage contributions of all life cycle stages is equal to the total deviation between the decommissioning health status of the virtual decommissioning final state node and the standard decommissioning health status of the equipment.

6. The method for device lifecycle health assessment driven by multi-source heterogeneous data according to claim 5, characterized in that: In S3, the reverse causal health path of the device is a node sequence path formed by connecting the virtual decommissioning final state node in the life cycle health state time sequence diagram and the nodes of each life cycle stage associated with it in reverse time. Each life cycle stage node stores the corresponding reverse health state value and stage contribution in the path, which is used to identify the health impact link from the virtual decommissioning final state node back to the initial operation stage of the device, and to identify the key life cycle stages whose contribution to the decommissioning health state exceeds a preset threshold.

7. The method for device lifecycle health assessment driven by multi-source heterogeneous data according to claim 6, characterized in that: In S4, the four-path health deviation pattern assessment mechanism acquires the stage health status values ​​corresponding to the equipment's standard health path, historical health path, forward predicted health path, and reverse causal inverse health path at the same life cycle stage node, and constructs an assessment mechanism for path-level deviation feature vectors. Based on the path-level deviation feature vectors, a comprehensive analysis is performed on the multidimensional deviation relationship between the four paths to determine the deviation pattern distribution of the equipment at each life cycle stage. The path-level deviation feature vectors include node-level deviation indicators, window-level cumulative deviation indicators, and path trend consistency indicators.

8. The method for device lifecycle health assessment driven by multi-source heterogeneous data according to claim 7, characterized in that: In S4, the deviation pattern distribution is to classify the path-level deviation feature vectors corresponding to each life cycle stage node in the life cycle health status time sequence diagram according to the deviation pattern determined by the four-path health deviation pattern evaluation mechanism, and form a sequence distribution of the deviation patterns of each life cycle stage node on the life cycle time axis, which is used to map the health status of the device in different life cycle stages.

9. A device lifecycle health assessment system driven by multi-source heterogeneous data, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes a computer program to implement the device lifecycle health assessment method driven by multi-source heterogeneous data as described in any one of claims 1-8.

Images