Operation state data monitoring system and method for biomass boiler flexible pipe group structure

Abstract

The application provides a kind of biomass boiler flexible pipe group structure operating state data monitoring system and method.There is:Multi-source perception module, for the flexible pipe group structure is collected including pipe wall temperature distribution data, structural acoustic data and displacement data operating state data;Edge processing module, for generating fusible data object;State estimation module, for constructing coupling state model, and output state quantity associated with flexible pipe group structure;Abnormal identification module, for parallel computing residual abnormal indication of coupling state model and event indication of structural acoustic data;Evidence package generation module, for generating evidence package associated with anomaly;Output interface module, for output target pipe segment identification, state quantity and evidence package to upper system, for the operating state monitoring of flexible pipe group structure is used.The application can improve the consistency of multi-source monitoring data fusion, improve the accuracy of abnormal positioning, reduce false alarm and enhance traceability.

Description

Technical Field

[0001] This application relates to the field of biomass boiler operation status monitoring technology, and in particular to a system and method for monitoring the operation status data of a flexible tube assembly structure of a biomass boiler. Background Technology

[0002] Biomass boilers are prone to slagging, fouling, localized overheating, and corrosion thinning on their heating surfaces due to large fluctuations in fuel composition and high ash and alkali metal content. Simultaneously, the flexible tube assembly structure experiences relative displacement and strain concentration under the influence of thermal expansion compensation, support constraints, and flow-induced vibration, resulting in a thermal, mechanical, and fluid coupling characteristic during operation. Current technologies for monitoring boiler heating surfaces typically employ acoustic or vibration-based leak / tube rupture alarms, or use a limited number of wall temperature measurement points and thermal parameter trend analysis for anomaly identification. There are also solutions that obtain local displacement information through visual measurement.

[0003] The aforementioned existing technical solutions mostly rely on a single signal source or sparse measurement points, making it difficult to achieve stable positioning and verifiable judgment under strong disturbance conditions such as soot blowing, slag falling, and load fluctuations. Furthermore, they lack unified time alignment and fusion modeling of pipe wall temperature distribution, structural acoustic events, and displacement changes, making it difficult to quantify the coupled states of support constraint degradation, slag heat transfer degradation, and strain risks. This results in both false alarms and missed alarms, an incomplete alarm evidence chain, and difficulty in supporting maintenance decisions and operation management. Summary of the Invention

[0004] In view of this, the present application provides a system and method for monitoring the operational status data of a flexible tube assembly structure in a biomass boiler, in order to solve the problems of difficulty in unifying and merging multi-source data in time, difficulty in stably locating anomalies, and lack of verifiable evidence chains in the monitoring results in the prior art.

[0005] The first aspect of this application provides an operational status data monitoring system for a biomass boiler flexible tube assembly structure, comprising: a multi-source sensing module for collecting operational status data including tube wall temperature distribution data, structural acoustic data, and displacement data for the flexible tube assembly structure, and writing a unified time stamp to the operational status data; an edge processing module for performing time alignment on the operational status data based on the unified time stamp, and performing data quality judgment and operating condition event labeling on the operational status data to generate fusionable data objects; and a state estimation module for constructing a coupled state model based on the topological identifiers and constraint relationships of the flexible tube assembly structure, and using the fusionable data... The system performs online calibration of the coupled state model based on observed inputs to output state variables associated with the flexible pipe assembly structure. An anomaly identification module is used to compute residual anomaly indicators from the coupled state model and event indicators from the structural acoustic data in parallel, and maps anomalies to target pipe segment identifiers in the flexible pipe assembly structure based on consistency constraints. An evidence package generation module is used to generate an evidence package associated with the anomaly by aggregating corresponding time window data, operational event annotations, and model version information based on the target pipe segment identifier when the anomaly mapping is valid. An output interface module outputs the target pipe segment identifier, state variables, and evidence package to the upper-level system for monitoring the operational status of the flexible pipe assembly structure.

[0006] The second aspect of this application provides a method for monitoring the operational status data of a flexible tube assembly structure in a biomass boiler based on the system of the first aspect. The method includes: collecting operational status data of the flexible tube assembly structure, including tube wall temperature distribution data, structural acoustic data, and displacement data, and writing a unified time stamp to the operational status data; performing time alignment on the operational status data based on the unified time stamp, and performing data quality judgment and operational event labeling on the operational status data to generate a fusionable data object; constructing a coupled state model based on the topological identifier and constraint relationship of the flexible tube assembly structure, and performing online calibration on the coupled state model using the fusionable data object as observation input, outputting state variables associated with the flexible tube assembly structure; parallel computing the residual anomaly indication of the coupled state model and the event indication of the structural acoustic data, and mapping the anomaly to the target tube segment identifier of the flexible tube assembly structure based on consistency constraints; when the anomaly mapping is valid, aggregating the corresponding time window data, operational event labels, and model version information based on the target tube segment identifier to generate an evidence package associated with the anomaly; and outputting the target tube segment identifier, state variables, and evidence package to the upper-level system for use in monitoring the operational status data of the flexible tube assembly structure.

[0007] The above-described technical solutions adopted in the embodiments of this application can achieve the following beneficial effects: The system employs a multi-source sensing module to collect operational status data for the flexible pipe assembly structure, including pipe wall temperature distribution, structural acoustic data, and displacement data, and writes a unified time stamp to the operational status data. An edge processing module performs time alignment on the operational status data based on the unified time stamp, and assesses data quality and labels operational events to generate fusionable data objects. A state estimation module constructs a coupled state model based on the topological identifiers and constraints of the flexible pipe assembly structure, and performs online calibration of the coupled state model using fusionable data objects as observation input to output state variables associated with the flexible pipe assembly structure. An anomaly identification module calculates residual anomaly indicators from the coupled state model and event indicators from the structural acoustic data in parallel, and maps anomalies to target pipe segment identifiers in the flexible pipe assembly structure based on consistency constraints. An evidence package generation module, when anomaly mapping is valid, aggregates corresponding time window data, operational event labels, and model version information based on the target pipe segment identifier to generate an evidence package associated with the anomaly. An output interface module outputs the target pipe segment identifier, state variables, and evidence package to the upper-level system for use in monitoring the operational status of the flexible pipe assembly structure. This application can improve the consistency of multi-source monitoring data fusion, enhance the accuracy of anomaly location, reduce false alarms and missed alarms, and enhance traceability. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application, 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 of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a schematic diagram of the structural composition of the biomass boiler flexible tube assembly operation status data monitoring system provided in the embodiments of this application; Figure 2 This is a flowchart illustrating the method for monitoring the operational status data of a biomass boiler flexible tube assembly structure provided in this application embodiment. Detailed Implementation

[0010] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0011] In existing technologies, the monitoring of boiler heating surfaces and flexible tube structures typically employs acoustic or vibration signals to alarm for leaks / tube ruptures, or relies on a limited number of wall temperature measuring points and trends in thermal parameters for anomaly identification. Some solutions supplement this with visual or displacement measurements to obtain local motion information. These solutions often rely on a single signal source or sparse measuring points, and data from different sources often lack a unified time base and a unified identification system. This results in insufficient stability in anomaly detection and limited granularity under strong disturbance conditions such as soot blowing, slag shedding, and load fluctuations. Alarm conclusions lack verifiable data evidence chains, making it difficult to meet the continuous monitoring requirements of the "thermal-mechanical-coupling state" of flexible tube structures.

[0012] Based on this, the technical problem to be solved by this application is that: multi-source operating status data is difficult to unify time alignment and fusion modeling, anomalies are difficult to be stably mapped to target pipe segment identifiers, and monitoring results lack verifiable and traceable evidence chains, resulting in both false alarms and missed alarms, and insufficient basis for maintenance decisions.

[0013] To address the aforementioned technical issues, this application proposes a system and method for monitoring the operational status data of a flexible tube assembly structure in a biomass boiler. This application utilizes a multi-source sensing module to collect operational status data, including tube wall temperature distribution data, structural acoustic data, and displacement data, and writes this data into a unified time stamp. An edge processing module performs time alignment based on the unified time stamp and performs data quality assessment and operational event labeling on the operational status data to generate fusionable data objects. A state estimation module constructs a coupled state model based on the topological identifiers and constraint relationships of the flexible tube assembly structure, and performs online calibration on the coupled state model using the fusionable data objects as observation input to output state variables. An anomaly identification module calculates residual anomaly indicators and event indicators in parallel and maps anomalies to target tube segment identifiers based on consistency constraints. When anomaly mapping is valid, an evidence package generation module aggregates corresponding time window data, operational event labels, and model version information around the target tube segment identifier to generate an evidence package, and outputs the target tube segment identifier, state variables, and evidence package to the upper-level system through an output interface module.

[0014] Through the above technical solutions, this application can improve the consistency of multi-source monitoring data fusion under a unified time base and unified identification system, improve the accuracy and stability of anomaly mapping to target pipe segment identification, reduce false alarms and missed alarms, and enhance the verifiability and traceability of monitoring conclusions through the evidence package mechanism, thereby better supporting the operation management and maintenance decision-making of flexible pipe group structures.

[0015] The specific composition and functions of the biomass boiler flexible tube assembly structure operation status data monitoring system provided in this application will be described in detail below with reference to the accompanying drawings and specific embodiments. Figure 1 This is a schematic diagram of the structural composition of the biomass boiler flexible tube assembly operation status data monitoring system provided in the embodiments of this application, as shown below. Figure 1 As shown, the system may specifically include the following components: The multi-source sensing module 101 is used to collect operational status data for the flexible tube assembly structure, including tube wall temperature distribution data, structural acoustic data, and displacement data, and write a unified time stamp for the operational status data. The edge processing module 102 is used to perform time alignment on the running status data based on a unified time stamp, and to perform data quality judgment and working condition event labeling on the running status data to generate fusionable data objects. The state estimation module 103 is used to construct a coupled state model based on the topological identifiers and constraint relationships of the flexible tube group structure, and to perform online calibration of the coupled state model with fusionable data objects as observation input, so as to output the state variables associated with the flexible tube group structure. Anomaly identification module 104 is used to perform parallel calculations of residual anomaly indicators of the coupled state model and event indicators of structural acoustic data, and to map anomalies to target pipe segment identifiers of the flexible pipe group structure based on consistency constraints. The evidence package generation module 105 is used to generate an evidence package associated with the anomaly based on the target pipe segment identifier, the corresponding time window data, the working condition event annotation, and the model version information when the anomaly mapping is established. The output interface module 106 is used to output the target pipe segment identifier, status quantity and evidence package to the upper system for use in monitoring the operation status of the flexible pipe group structure.

[0016] It should be noted that "flexible tube assembly structure" refers to at least one set of tube bundle / tube screen / serpentine tube assembly installed in the heating surface system of a biomass boiler. This assembly allows for thermal expansion and relative displacement through support devices, guide / limiting structures, expansion compensation structures or flexible connection structures, and exhibits observable displacement, vibration and strain changes during operation. Its failure modes include, but are not limited to: local overheating, creep bulging, wear thinning, corrosion thinning, abnormal displacement / friction caused by support point degradation, flow-induced vibration fatigue cracks and eventual leakage / tube rupture.

[0017] In some embodiments, the multi-source sensing module is specifically used for: Obtain the set of pipe segment identifiers associated with the flexible pipe assembly structure and the corresponding measurement point mapping relationship of each pipe segment identifier. The measurement point mapping relationship is used to characterize the correspondence between the pipe wall temperature distribution acquisition channel, the structural acoustic acquisition channel and the displacement acquisition channel and the pipe segment identifier, respectively. Based on the measurement point mapping relationship, pipe wall temperature distribution data, structural acoustic data and displacement data are collected respectively, and a unified time mark corresponding to the same time base is written for each collected data record. When collecting pipe wall temperature distribution data, a distributed measurement method is used to output spatial ranging markers corresponding to the pipe wall temperature distribution data, and the spatial ranging markers are associated with a unified time marker and written into the operating status data.

[0018] Specifically, the pipe segment identifier set refers to the set of unique identifiers assigned to each pipe segment, elbow segment, or weld joint neighboring segment after the flexible pipe assembly structure is segmented. These identifiers are used to uniformly reference pipe segment objects within the system and serve as the spatial index basis for subsequent state estimation and anomaly mapping.

[0019] The measurement point mapping relationship refers to establishing association rules between the acquisition channels and the pipe segment identifiers, with the pipe segment identifier as the primary key. The acquisition channels include pipe wall temperature distribution acquisition channels, structural acoustic acquisition channels, and displacement acquisition channels. The association rules are used to clarify which pipe segment identifier the data acquired by a certain channel should belong to, or which channels' data should be aggregated for a certain pipe segment identifier.

[0020] A unified time stamp refers to a timestamp field generated under the same time base, used to identify the acquisition time of each record of operational status data. The same time base refers to a unified clock reference shared by all acquisition units within the multi-source sensing module. It can be provided by the edge-side time synchronization service and distributed to each acquisition unit to ensure time consistency.

[0021] Distributed measurement refers to a measurement method that forms a continuous or quasi-continuous sequence of measurement points along the spatial direction of a pipe segment. Unlike discrete single-point measurement, its output data naturally carries spatial position parameters along the length direction.

[0022] Spatial ranging markers refer to the distance parameters corresponding to each spatial sampling point in the distributed measurement output. They are used to characterize the path distance of the sampling point relative to the reference zero point. The reference zero point can be set as the fiber optic input end, the starting weld of the pipe segment, or the reference point defined by the pipe segment identifier, thereby realizing the correspondence between the temperature distribution and the spatial position of the pipe segment.

[0023] Operational status data refers to a set of multi-source data records output by the multi-source sensing module and consumed by the edge processing module. Each record carries at least a unified time stamp and may further carry pipe segment identifiers, acquisition channel identifiers, spatial ranging markers, and sampled value sequences depending on the data type.

[0024] In terms of specific implementation, the multi-source sensing module can establish a mapping relationship between pipe segment identifiers and measurement points during the system deployment phase. For example, for the flexible tube panel of the superheater in a biomass boiler, each tube panel is divided into several pipe segments along its length, and each pipe segment is assigned a pipe segment identifier. At the same time, a corresponding pipe wall temperature distribution acquisition channel is set up as a distributed temperature measurement medium laid along the outer wall of the tube panel, a corresponding structural acoustic acquisition channel is set up as a structural acoustic sensor channel arranged near the support points of the tube panel and in the vicinity of the header, and a corresponding displacement acquisition channel is set up as a displacement measurement channel aligned with the edge of the tube panel or the support components.

[0025] In some examples, the measurement point mapping relationship can be embedded as configuration data and sent to the multi-source sensing module, enabling the module to automatically merge data from each channel to the corresponding pipe segment identifier during acquisition. To ensure a common time base, the multi-source sensing module can periodically synchronize its time via an edge-side time synchronization service, or a dedicated time synchronization unit can output synchronization pulses and timestamp references to each acquisition unit. This allows event segments of structural acoustic data, sampling sequences of displacement data, and measurement frames of temperature distribution data to share the same unified time stamp generation rule. During acquisition, the multi-source sensing module writes a unified time stamp, channel identifier, and pipe segment identifier to each data record, enabling subsequent edge processing modules to perform cross-channel time alignment based on the unified time stamp and complete spatial merging based on the pipe segment identifier.

[0026] Regarding the acquisition of pipe wall temperature distribution data, the multi-source sensing module employs a distributed measurement method to form the temperature distribution along the pipe segment. Specifically, a reference zero point for spatial ranging markers can be established on each pipe segment, and the correspondence between the ranging values ​​and temperature values ​​of multiple spatial sampling points can be output within the acquisition frame. When generating operational status data, the multi-source sensing module associates the spatial ranging markers with a unified time marker. For example, the data record corresponding to the same unified time marker contains both the temperature distribution sampling value sequence at that moment and the spatial ranging marker sequence corresponding to each point of the sampling value sequence, thereby ensuring that the "spatial temperature curve at the same moment" can be completely reproduced.

[0027] To improve the stability of spatial markers, the multi-source sensing module can calibrate the spatial ranging markers during the deployment phase. The calibration process can use the actual length of the pipe segment, the reference weld position, or the fixed support position as calibration points to establish a mapping between the ranging value and the actual spatial position. This mapping parameter is then bound and stored with the pipe segment identifier to avoid spatial assignment errors caused by ranging drift after maintenance or replacement.

[0028] The following is a specific example illustrating the operation of this embodiment. Taking the flexible tube panel of a biomass boiler superheater as the monitoring object, the tube panel is divided into several tube segments according to the allowable direction of thermal expansion and assigned a set of tube segment identifiers. Among them, the tube segment identifiers near the header are used to monitor changes in bearing structural constraints, and the tube segment identifiers near the high-temperature flue gas scouring area are used to monitor changes in bearing slagging heat transfer. The multi-source sensing module simultaneously collects temperature distribution, structural acoustic, and displacement data for each pipe segment based on the measurement point mapping relationship: when a soot blowing event occurs during boiler load ramp-up, the structural acoustic channel collects short-term impact signals, the displacement channel collects instantaneous swing amplitude changes of the pipe screen, and the temperature distribution channel collects abrupt changes in local temperature gradients. The multi-source sensing module writes a unified time stamp generated under the same time base to the above data records and synchronously writes a spatial ranging mark into the temperature distribution data, so that the subsequent edge processing module can align the three types of data to the same event time window with a unified time stamp and accurately map the temperature abrupt change segment to the corresponding pipe segment identifier and spatial location segment with the spatial ranging mark, thereby providing a consistent data foundation for the construction of observation inputs for the subsequent coupled state model, residual anomaly calculation, and target pipe segment identifier mapping.

[0029] Through the above embodiment, the multi-source sensing module can achieve unified time-based acquisition and consistent merging of pipe wall temperature distribution data, structural acoustic data and displacement data under the constraints of pipe segment identifier set and measurement point mapping relationship. It can also achieve spatial traceability recording of temperature distribution through spatial ranging markers, thereby providing more complete data support for subsequent time alignment, fusion modeling and anomaly mapping, improving the consistency of multi-source data alignment and reducing misjudgments caused by unclear spatial attribution and time asynchrony.

[0030] In some embodiments, the multi-source sensing module further includes a structured acoustic array unit, which includes a plurality of distributed structured acoustic sensors and array position information corresponding to the structured acoustic sensors, for generating acoustic event indications and sound source orientation parameters based on structured acoustic data, and using the sound source orientation parameters as fields of a fusionable data object.

[0031] In some embodiments, the edge processing module is specifically used for: Based on a unified time stamp, time alignment processing is performed on pipe wall temperature distribution data, structural acoustic data, and displacement data, and an alignment index corresponding to each data channel is generated. For the aligned data of each data channel, extract quality parameters to characterize the validity of the data; Data availability masks are generated based on quality parameters and corresponding to each data channel. The data availability mask is used to indicate the available segment of the running status data within the corresponding time window. Acquire operating condition data associated with the operating conditions of the biomass boiler, and align and associate the operating condition data with the operating status data based on a unified time stamp; Based on the joint features of operating condition data and operational status data, operating condition event identification is performed, operating condition event labels are generated, and event time windows and confidence parameters are written for each operating condition event label. The runtime status data, which carries alignment indexes, data availability masks, and operational event annotations, is encapsulated to generate a mergeable data object.

[0032] Specifically, the alignment index refers to the index structure generated by the edge processing module when performing time alignment processing. It is used to describe the correspondence between sampling points of different data channels on the same aligned time axis. For example, it is used to characterize the alignment position of sampling segments of structural acoustic data with sampling points of displacement data and temperature distribution measurement frames on the time axis.

[0033] Quality parameters refer to a set of parameters used to characterize the validity and usability of data, at least to describe situations such as data loss, saturation, drift, changes in noise floor, or sampling discontinuity.

[0034] Data availability mask refers to the available segment identification information generated based on quality parameters. It is used to indicate which sampling segments of the operating status data can be accepted by subsequent algorithms and which segments should be downweighted or removed within a certain time window.

[0035] Operating data refers to process data related to the operation of biomass boilers, including at least one of the following: load, feedwater parameters, steam parameters, furnace negative pressure, air volume and oxygen content, flue gas temperature, soot blowing status, and fuel batch or blending ratio.

[0036] Operating condition event labeling refers to the identification of key operating condition events during operation as structured labels. The event time window describes the start and end time range of the event, and the confidence parameter describes the reliability of the event identification result.

[0037] Fusionable data objects refer to encapsulated data objects, which include, but are not limited to, alignment indexes, data availability masks, and operating condition event annotations, and retain unified time stamps and pipe segment identifiers associated with each data record, so that the state estimation module can construct observation inputs and perform online calibration.

[0038] First, regarding time alignment, the edge processing module constructs an aligned time axis based on a unified time stamp and maps the pipe wall temperature distribution data, structural acoustic data, and displacement data onto this aligned time axis. Since temperature distribution data is typically output periodically in the form of measurement frames, structural acoustic data is typically output continuously with high-frequency sampling, and displacement data may be output with mid-to-low-frequency sampling, in this embodiment, the edge processing module can set the aligned time axis to a main time granularity aligned with the displacement sampling period or temperature frame period, and establish a segment mapping relationship for the structural acoustic data corresponding to the main time granularity. After completing the mapping, the edge processing module generates an alignment index, which records the temperature frame number, displacement sampling point number, and the start and end positions of the structural acoustic segment corresponding to each main time granularity, thereby ensuring that subsequent modules can obtain consistent multi-source data slices within the same time window.

[0039] Furthermore, regarding data quality assessment, the edge processing module extracts quality parameters and generates a data availability mask for the aligned data of each data channel. Taking structural acoustic data as an example, the edge processing module can extract noise floor parameters and saturation parameters to distinguish abnormal waveforms caused by equipment impact, dust blowing noise, or sensor saturation. Taking displacement data as an example, the edge processing module can extract packet loss parameters and drift parameters to identify communication jitter or measurement reference drift. Taking temperature distribution data as an example, the edge processing module can extract sampling continuity parameters and abnormal jump parameters to identify distributed measurement link interruptions or ranging mapping anomalies. The edge processing module generates a data availability mask based on the quality parameters. The data availability mask identifies available segments with time windows as the granularity and can add a weighted indicator to unavailable segments, enabling subsequent state estimation and anomaly identification to automatically avoid contaminating the conclusions with bad data segments when calculating residuals and confidence parameters.

[0040] Furthermore, regarding operating condition data alignment and operating condition event labeling, the edge processing module acquires operating condition data from distributed control systems or other process data interfaces and aligns and correlates the operating condition data with the operational status data based on a unified time stamp. After alignment and correlation, the edge processing module performs operating condition event identification based on the joint features of the operating condition data and the operational status data. For example, when identifying soot blowing events, the changes in soot blowing status in the operating condition data and the sudden increase in noise energy in the structural acoustic data can be used simultaneously; when identifying load ramping events, the joint features of the load change rate and the overall upward trend of the temperature distribution can be used; when identifying fuel switching events, the joint features of changes in fuel batch identification and furnace combustion characteristic parameters, as well as changes in the spectrum of structural acoustic background noise, can be used. The edge processing module writes the identified operating condition events into operating condition event labels and records the event time window and confidence parameters for each event label. The confidence parameters can be generated based on the consistency of multi-source evidence, allowing the subsequent state estimation module to use the operating condition event labels as calibration gating conditions or anomaly interpretation conditions.

[0041] Furthermore, in terms of generating fusionable data objects, the edge processing module encapsulates operational status data carrying alignment indexes, data availability masks, and operational event annotations. During encapsulation, the edge processing module retains a unified timestamp for each data record and writes key fields such as pipe segment identifiers, acquisition channel identifiers, and spatial ranging markers to maintain the "time-space-channel" referencing foundation. For high-frequency channels such as structural acoustic data, the edge processing module can slice the original waveform into time windows at the edge and reference them using an index, thereby reducing the transmission burden without changing the event windows that can be reconstructed by subsequent modules. The finally generated fusionable data object is output to the state estimation module to construct observation inputs and drive online calibration of the coupled state model. It is also used by the anomaly identification module to determine the consistency between residual anomaly indications and event indications.

[0042] The workflow of this embodiment is illustrated below with a specific example. During the operation of the flexible tube screen of a biomass boiler superheater, the multi-source sensing module continuously outputs temperature distribution measurement frames, structural acoustic continuous waveforms, and displacement sampling points, all of which are written with a unified time stamp. After receiving the above data, the edge processing module establishes an aligned time axis and generates an alignment index using a unified time stamp. When a soot blowing operation occurs during a certain period, the soot blowing status in the operating data changes from not executed to executed, and the structural acoustic channel simultaneously experiences a short-term broadband noise increase. Based on this, the edge processing module generates a soot blowing event label and writes it into the event time window and confidence parameters. At the same time, the structural acoustic channel experiences local saturation at the moment soot blowing begins. The edge processing module marks this segment as an unusable segment using the saturation parameter and writes it into the data availability mask. Within this event time window, the temperature distribution data remains continuous and the displacement data has no packet loss. The edge processing module encapsulates the alignment index, data availability mask, and soot blowing event label together to generate a fusionable data object. This allows the state estimation module to automatically reduce the weight of acoustically unusable segments during subsequent online calibration and select non-soot blowing steady-state segments for parameter updates under gating conditions, thereby avoiding bias in model calibration caused by soot blowing disturbances.

[0043] Through the above embodiment, the edge processing module can achieve stable time alignment of multi-source operating status data based on a unified time stamp, explicitly identify bad data segments using quality parameters and data availability masks, and generate operating condition event labels containing event time windows and confidence parameters after aligning the operating condition data with the operating status data. Finally, it outputs fusionable data objects with reference basis, thereby improving the reliability and consistency of multi-source data fusion processing and reducing the impact of strong disturbance operating conditions and data quality fluctuations on subsequent state estimation and anomaly identification conclusions.

[0044] In some embodiments, a coupled state model is constructed based on the topological identifiers and constraint relationships of the flexible tube assembly structure, including: Obtain the topology identifier of the flexible pipe assembly structure. The topology identifier includes the pipe segment identifier, the connection relationship corresponding to the pipe segment identifier, and the support point identifier associated with the connection relationship. Based on topological identifiers, a structural state network is constructed with pipe segment identifiers as nodes and connection relationships as edges, and constraint relationships are configured for each support point identifier in the structural state network; Define a set of coupled state variables on the structural state network. The set of coupled state variables includes pipe wall temperature distribution state variables, pipe segment displacement state variables, and constraint state variables corresponding to constraint relationships. Establish state evolution relationships for the set of coupled state variables. These relationships include coupled evolution terms that use thermal and mechanical boundary quantities corresponding to the operating data as driving inputs, and writing equivalent thermal resistance parameters and support constraint parameters as parameters to be calibrated into the coupled state model. Establish observation mapping relationships to map pipe wall temperature distribution data and displacement data into observation inputs for the coupled state model.

[0045] Specifically, a structural state network refers to a graph-structured data object consisting of pipe segment identifiers as nodes and connection relationships as edges. It is used to express the geometric connectivity and constraint coupling paths of flexible pipe group structures and serves as a carrier for the organization and propagation calculation of state variables.

[0046] The equivalent thermal resistance parameter refers to the equivalent heat transfer impedance parameter that characterizes the pipe section under the influence of heat exchange and deposit layer on the flue gas side. It is written into the coupled state model in the form of parameters and used as the object to be calibrated to update the heat transfer state characterization of the pipe section during operation.

[0047] Support point constraint parameters refer to the set of equivalent constraint characteristic parameters that characterize the movement of the support point on adjacent pipe segments. They are used to describe at least the constraint characteristics such as allowable displacement direction, limit gap and equivalent constraint strength, and are written into the coupled state model as the object to be calibrated to characterize the kinematic changes caused by constraint degradation or jamming.

[0048] Firstly, regarding topology identification, the state estimation module can read the topology identification of the flexible pipe assembly structure from design drawing data, equipment ledgers, or configuration files. Topology identification includes, but is not limited to: pipe segment identifiers for uniquely identifying each pipe segment; connection relationships characterizing the connection methods between pipe segments, which can reflect connection forms such as welded connections, elbow connections, header connections, or pipe bank connections; and support point identifiers associated with the connection relationships, indicating the support point objects that apply motion constraints to the connection area or adjacent pipe segments. To ensure that the topology identification can be accurately referenced in subsequent observation mappings, the state estimation module can also write a connection type tag and adjacent pipe segment identifier pair for each connection relationship when reading the topology identification, thereby ensuring that the edges of the structural state network can distinguish different connection forms and carry constraint configuration entries.

[0049] Furthermore, in terms of constructing the structural state network, the state estimation module generates a structural state network based on topological identifiers, with pipe segment identifiers as nodes and connection relationships as edges. Nodes are used to carry state variables associated with their corresponding pipe segment identifiers, and edges are used to carry coupling relationships and constraint propagation relationships between adjacent pipe segments. Subsequently, the state estimation module configures constraint relationships for each support point identifier in the structural state network, enabling support point constraints to act parametrically on adjacent nodes and their connecting edges in the graph structure. For example, for guide supports that allow thermal expansion in a certain direction, the constraint relationship can be configured as a combination of "allowed direction + limiting direction"; for limiting structures with gaps and friction, the constraint relationship can be configured as a combination of "gap parameter + equivalent constraint strength parameter". Through the above configuration, the structural state network not only expresses connectivity but also explicitly expresses the constraint application location and constraint characteristics, thereby providing a structured carrier for the subsequent writing of constraint state variables and support point constraint parameters.

[0050] Furthermore, regarding the definition of the coupled state variable set, the state estimation module defines pipe wall temperature distribution state variables and pipe segment displacement state variables for each pipe segment identifier on the structural state network, and defines constraint state variables corresponding to the constraint relationships for each support point identifier or its associated edge. The pipe wall temperature distribution state variable is used to characterize the temperature distribution pattern along the spatial distance marker direction under the same pipe segment identifier, the pipe segment displacement state variable is used to characterize the displacement amplitude or displacement trajectory parameters under the same pipe segment identifier, and the constraint state variable is used to characterize the current state representation quantity of the support point with applied constraints. To ensure a traceable correspondence between variables and observation data, the state estimation module binds and stores the above state variables with pipe segment identifiers and support point identifiers, so that subsequent observation inputs and residual calculations can be indexed by identifier.

[0051] Furthermore, regarding the establishment of state evolution relationships, the state estimation module establishes state evolution relationships for the set of coupled state variables. These relationships include coupled evolution terms that use the thermal and mechanical boundary quantities corresponding to the operating data as driving inputs, and equivalent thermal resistance parameters and support point constraint parameters as parameters to be calibrated and written into the coupled state model. Specifically, thermal boundary quantities can be obtained by combining operating data such as load, flue gas temperature, airflow and oxygen content, feedwater and steam parameters, and are used to drive the evolution of the pipe wall temperature distribution state variable; mechanical boundary quantities can be obtained by labeling operating events such as load change rate, start-up / shutdown stage indicators, and soot blowing status, or by combining operating data, and are used to drive the evolution of the pipe segment displacement state variable and constraint state variable.

[0052] The equivalent thermal resistance parameter is written into the parameter set associated with the pipe segment identifier to characterize the difference in heat transfer impedance under the same pipe segment identifier in the model; the support point constraint parameter is written into the parameter set associated with the support point identifier to characterize the change in constraint characteristics under the same support point identifier in the model. The state estimation module associates the above parameters with the state evolution relationship, enabling subsequent online calibration to update the parameters under the observation residual constraint and reflect the update results as state quantity output.

[0053] Furthermore, regarding the establishment of observation mapping relationships, the state estimation module establishes observation mapping relationships to map pipe wall temperature distribution data and displacement data into observation inputs for the coupled state model. The observation mapping relationships include at least two types of mappings: one is a temperature distribution mapping, used to merge temperature distribution data carrying spatial ranging markers according to pipe segment identifiers and map them into observation inputs isomorphic to the pipe wall temperature distribution state variables; the other is a displacement mapping, used to merge displacement data according to pipe segment identifiers and map them into observation inputs isomorphic to the pipe segment displacement state variables.

[0054] In some examples, to improve the robustness of the mapping, the observation mapping relationship can be further introduced with the data availability mask output by the edge processing module as an availability segment constraint, so that the observation input is constructed only from the observation data of the availability segment, thereby reducing the impact of bad data on model calibration. Through the observation mapping relationship, the coupled state model forms a corresponding structure of "state variables indexed by identifier - observation inputs indexed by identifier", laying a consistent data structure foundation for subsequent model prediction generation, observation residual calculation and anomaly mapping.

[0055] The modeling process of this embodiment is illustrated below with a specific example. Taking the flexible tube panel of the superheater as the object, the state estimation module reads the set of tube segment identifiers, the connection relationships between adjacent tube segments, and the corresponding support point identifiers of the tube panel, and constructs a structural state network. Among them, the connection relationship near the header is associated with several support point identifiers. The state estimation module configures the constraint relationship between the allowable thermal expansion direction and the limiting gap at these support points and writes the initial value of the support point constraint parameter. In the high-temperature flue gas scouring area in the middle section of the tube panel, the state estimation module writes the initial value of the equivalent thermal resistance parameter for the corresponding tube segment to reflect the heat transfer difference of different tube segments. During operation, when the edge processing module outputs the data object that can be fused, the state estimation module maps the temperature distribution data to the temperature observation input of each tube segment through the observation mapping relationship, maps the displacement data to the displacement observation input of each tube segment, and writes the thermal boundary quantity and mechanical boundary quantity formed by the operating condition data as driving inputs into the evolution term, thereby completing the construction of a coupled state model that can be called for online calibration.

[0056] Through the above embodiment, the coupled state model can use the structural state network as a carrier to explicitly parameterize the connectivity and support constraint relationships of the flexible tube assembly structure. Within a unified framework, it simultaneously carries the coupled state variables of tube wall temperature distribution, tube segment displacement, and constraint state. Combining the state evolution relationship driven by thermal and mechanical boundary quantities and the observation mapping relationship oriented towards temperature distribution and displacement data, it forms a model foundation that can be calibrated online and can calculate residual anomaly indications, thereby improving the structural consistency and traceability of subsequent state estimation and anomaly mapping.

[0057] In some embodiments, online calibration of the coupled state model is performed using fusionable data objects as observation inputs to output state variables associated with the flexible tube assembly structure, including: Based on the alignment index and data availability mask in the fusionable data object, the effective observation sequence for calibration is determined, and the effective observation sequence is aggregated into a pipe segment observation set according to the pipe segment identifier. Using the labeling of operating conditions as the calibration gating condition, when the gating is passed, the set of pipe segment observations within the event time window corresponding to the labeling of operating conditions is input into the coupled state model to trigger the incremental update of the parameters to be calibrated. Based on the state evolution relationship and observation mapping relationship of the coupled state model, the model prediction value and observation residual corresponding to the pipe segment observation set are calculated, and the equivalent thermal resistance parameter and support point constraint parameter are iteratively updated under multi-parameter joint constraints to obtain the parameter update amount associated with each pipe segment identifier. Based on the parameter update, thermal resistance degradation state quantity and constraint degradation state quantity are generated. Based on the pipe wall temperature distribution state variable and pipe segment displacement state variable, strain indication quantity associated with pipe segment identification is calculated, and strain risk state quantity is generated based on strain indication quantity and constraint degradation state quantity.

[0058] Specifically, calibration gating conditions refer to a set of rules used to control whether online calibration is triggered and which time window data is selected for calibration. Based on the labeling of operating events, they are used to pause calibration when a specific event occurs, trigger calibration only during stable operating conditions, or trigger calibration in segments before and after the event.

[0059] Multi-parameter joint constraints refer to the constraint relationships applied when updating equivalent thermal resistance parameters and support point constraint parameters simultaneously. They are used to limit the feasible domain, coupling consistency, or update step size relationship of parameter updates, and avoid conflicting update directions of different parameters within the same time window, thereby maintaining the consistency of model parameters in physical and structural logic.

[0060] First, regarding the determination of effective observation sequences, the state estimation module parses the alignment index and data availability mask in the fusionable data objects. The alignment index is used to extract data slices on the same aligned time axis from different data channels, and the data availability mask is used to identify the available segments of each channel's data within each time window. The state estimation module eliminates unavailable segments or reduces the weight of low-availability segments based on the data availability mask, and extracts effective observation sequences of pipe wall temperature distribution data and displacement data within the same time window in conjunction with the alignment index. Subsequently, the state estimation module aggregates the effective observation sequences according to pipe segment identifiers to form pipe segment observation sets, so that each pipe segment identifier corresponds to a set of synchronous temperature distribution observations and displacement observations, providing an identifier-organized data entry point for subsequent model predictions and observation residual calculations.

[0061] Furthermore, regarding calibration gating triggering, the state estimation module uses operating condition event labels as calibration gating conditions. Specifically, the state estimation module reads the event type, event time window, and confidence parameters from the operating condition event labels, and determines whether parameter updates are allowed within the event time window based on preset gating rules. For example, within the soot blowing event time window, structural acoustic and thermal disturbances are significant, and the gating rules can mark this time window as a shielded window that does not participate in parameter updates; within the load ramping event time window, the gating rules can divide it into a ramping phase and a plateau stabilization phase, and only select the pipe segment observation set corresponding to the plateau stabilization phase to trigger parameter updates; within the fuel switching event time window, the gating rules can trigger a switch in parameter update strategy, such as limiting the update step size of the equivalent thermal resistance parameter and requiring updates to be allowed only within a time window after the fuel switch when a preset stabilization period is met. When the gate passes, the state estimation module inputs the pipe section observation set within the event time window corresponding to the working condition event label into the coupled state model to trigger the incremental update of the parameters to be calibrated, thereby realizing online calibration rhythm control of "updating in the appropriate time window and freezing in the inappropriate time window".

[0062] Furthermore, regarding model prediction and residual-driven iteration, the state estimation module generates model predictions based on the state evolution relationship of the coupled state model and maps the pipe segment observation set to model observation inputs based on the observation mapping relationship. Within the same event time window, the state estimation module calculates the observation residual between the model prediction and the observation input for each pipe segment identifier and uses the observation residual as the driving force for parameter updates. Under multi-parameter joint constraints, the state estimation module performs iterative updates on the equivalent thermal resistance parameter and the support point constraint parameter to obtain the parameter update amount associated with each pipe segment identifier.

[0063] In some examples, iterative updates can employ an incremental approach: within each iteration cycle, a first update is generated for the equivalent thermal resistance parameter based on the temperature distribution residual, followed by a second update for the support point constraint parameter based on the displacement residual. Consistency checks and step size limits are then applied to both types of updates using multi-parameter joint constraints, resulting in the final parameter update value written into the model parameter set. In this method, temperature observations dominate the update of the equivalent thermal resistance parameter, displacement observations dominate the update of the support point constraint parameter, and joint constraints are used to ensure that the two types of parameter updates do not conflict within the same operating condition.

[0064] Furthermore, in terms of state quantity generation, the state estimation module generates thermal resistance degradation state quantities and constraint degradation state quantities based on parameter update quantities. Specifically, the state estimation module uses the parameter update quantity or parameter offset of the equivalent thermal resistance parameter as the representation of the thermal resistance degradation state quantity, and uses the parameter update quantity or parameter offset of the support point constraint parameter as the representation of the constraint degradation state quantity, and establishes an association record between the above state quantities and the corresponding pipe segment identifier.

[0065] The state estimation module calculates the strain indication quantity associated with the pipe segment identifier based on the pipe wall temperature distribution state variable and the pipe segment displacement state variable. For example, the strain indication quantity is obtained by combining the gradient characteristics of the temperature distribution along the spatial ranging mark direction with the displacement change amplitude characteristics, used to characterize the strain concentration trend of the pipe segment under thermal expansion and constraint. Subsequently, the state estimation module generates a strain risk state quantity based on the strain indication quantity and the constraint degradation state quantity, so that the strain risk state quantity simultaneously reflects the coupled effects of thermally induced deformation and constraint degradation. The state estimation module writes the thermal resistance degradation state quantity, the constraint degradation state quantity, and the strain risk state quantity into the state output record associated with the pipe segment identifier and outputs it to the anomaly identification module and the evidence package generation module, providing prior state support for the subsequent consistency determination of residual anomaly indications and event indications.

[0066] The online calibration process of this embodiment is illustrated below with a specific example. Taking the flexible tube panel of a biomass boiler superheater as the monitoring object, in the initial stage of operation, the equivalent thermal resistance parameter and the support point constraint parameter are both set to the initial values. After running for a period of time, the fusionable data object output by the edge processing module shows that the temperature distribution corresponding to a certain tube segment identifier shows a continuous rise in the local section of the spatial distance measurement mark, and the displacement data shows low-frequency drift within the corresponding time window; at the same time, the operating condition event label indicates that the time window is in the stable section of the load platform, and the gating condition judgment is passed.

[0067] Based on this, the state estimation module inputs the pipe segment observation set of the event time window into the coupled state model, generates model prediction values ​​and calculates observation residuals; during the iterative update process, the temperature distribution residual drives the equivalent thermal resistance parameter to be updated in the direction of higher impedance, and the displacement residual drives the support point constraint parameter to be updated in the direction of stronger constraint or smaller gap. The joint constraint limits the update step size of the two types of parameters and ensures that the updated parameters still meet the displacement continuity requirements of adjacent pipe segments under the action of the same support point.

[0068] After completing the incremental update, the state estimation module generates thermal resistance degradation state variables and constraint degradation state variables, and calculates strain indication variables by combining temperature gradient characteristics and displacement drift characteristics, thereby generating strain risk state variables. These state variables, output along with the pipe segment identifiers, provide a state background organized according to the pipe segment identifiers for the consistency determination of subsequent residual anomaly indications and structural acoustic event indications by the anomaly identification module.

[0069] Through the above embodiment, the state estimation module can stably select effective observation sequences under the constraints of alignment index and data availability mask, and use operating event annotation to form calibration gating conditions to control the timing of parameter update triggering. Driven by model prediction values ​​and observation residuals, it iteratively updates the equivalent thermal resistance parameters and support point constraint parameters, and outputs thermal resistance degradation state quantities, constraint degradation state quantities, and strain risk state quantities associated with pipe segment identifiers, thereby improving the continuity, traceability, and adaptability to operating condition disturbances of the state quantity output.

[0070] In some embodiments, the anomaly detection module is specifically used for: Based on fusionable data objects, the residual between the observation input corresponding to the pipe segment identifier and the model prediction value output by the coupled state model is calculated, and a set of candidate pipe segment identifiers and residual abnormal indicators corresponding to the set of candidate pipe segment identifiers are generated based on the temporal or spatial aggregation of the residuals. Event detection is performed on the structural acoustic data to generate event indications, and a sound source constraint domain is generated based on the location information of the structural acoustic array or a spatial representation equivalent to the location information. Based on the temporal consistency constraint between residual abnormality indication and event indication, and the spatial consistency constraint between the sound source constraint domain and the candidate pipe segment identifier set, the target pipe segment identifier is determined, and the mapping confidence parameter associated with the target pipe segment identifier is generated.

[0071] Specifically, the sound source constraint domain refers to the candidate spatial region of sound sources derived from the spatial representation of the structural acoustic array. It is used to limit the spatial range in which abnormal events may occur and can be mapped to a set of pipe segment identifiers that intersect or are adjacent to this spatial range.

[0072] Residual anomaly indication refers to the anomaly indication quantity formed by aggregating the observation residuals in the time or spatial dimensions. It is used to characterize the degree of residual anomaly of a certain pipe segment identifier or a set of adjacent pipe segments within a specific time window.

[0073] The mapping confidence parameter refers to the credibility characterization parameter of the mapping result of the target pipe segment identifier. It is generated by the degree of matching between residual abnormality indication and event indication under temporal and spatial consistency, and is output along with the target pipe segment identifier for evidence package generation and display by the upper-level system.

[0074] First, regarding residual calculation and candidate pipe segment generation, the anomaly identification module constructs observation inputs corresponding to each pipe segment identifier based on fusionable data objects, and obtains the model prediction values ​​output by the state estimation module on the same aligned time axis. The anomaly identification module calculates the residual between the observation input and the model prediction value for each pipe segment identifier. The residual can cover at least one of the following: temperature distribution residual and displacement residual, and forms a residual sequence within the same time window.

[0075] Subsequently, the anomaly detection module performs temporal or spatial aggregation on the residual sequence to generate residual anomaly indicators. Temporal aggregation is used to characterize the persistence characteristics of the residuals within a continuous time window, such as the length of the persistent segment where the residual exceeds a threshold, the cumulative value of the residual energy, or the number of residual mutations as the aggregation result. Spatial aggregation is used to characterize the connectivity characteristics of the residuals between adjacent pipe segment identifiers, such as the consistency of residual anomaly indicators between adjacent pipe segments or the range of connected regions as the aggregation result. Based on the aggregation results, the anomaly detection module generates a set of candidate pipe segment identifiers and records the residual anomaly indicators corresponding to the candidate pipe segment identifier set as the first type of evidence for subsequent consistency determination.

[0076] Furthermore, in terms of structural acoustic event detection and source constraint domain generation, the anomaly detection module performs event detection on the structural acoustic data in the fusionable data object to generate event indicators. Event detection can be obtained by judging features such as amplitude abrupt changes, energy surges, impact envelopes, or spectral changes in the structural acoustic data, so as to output event occurrence markers and event intensity representations within the event time window. After generating event indicators, the anomaly detection module generates source constraint domains based on the location information of the structural acoustic array or spatial representations equivalent to the location information.

[0077] Specifically, the location information of the structural acoustic array can characterize the installation position of each sensor in the boiler heating surface area or its proximity to the pipe segment marker. In one embodiment, the anomaly identification module determines that the sound source is more likely to be located in the vicinity of a certain sensor or in the spatial area covered by a certain sensor combination based on the relative intensity or arrival time difference of the event responses of each sensor, thereby forming a sound source constraint domain. In another embodiment, when it is difficult to reliably estimate the precise orientation, the anomaly identification module can define the sound source constraint domain as "the union of the coverage areas corresponding to the set of sensors that triggered the event" to obtain a higher-level and more robust spatial constraint.

[0078] Furthermore, regarding consistency constraint determination and target pipe segment identification, the anomaly identification module determines the target pipe segment identifier based on the temporal consistency constraint between residual anomaly indications and event indications, and the spatial consistency constraint between the sound source constraint domain and the candidate pipe segment identifier set. The temporal consistency constraint requires that the anomaly time window corresponding to the residual anomaly indication and the event time window corresponding to the event indication overlap or are adjacent, ensuring that both types of evidence point to the same anomalous process. The spatial consistency constraint requires that the candidate pipe segment identifier set intersects with, is adjacent to, or meets a preset distance threshold with the sound source constraint domain in space, thus ensuring that the candidate pipe segments selected by the model residuals are consistent with the spatial range defined by the acoustic event.

[0079] In some examples, when the consistency determination passes, the anomaly detection module selects the pipe segment identifier that satisfies the spatial consistency constraint and has the highest residual anomaly indication as the target pipe segment identifier, and generates a mapping confidence parameter associated with the target pipe segment identifier. The mapping confidence parameter can be obtained by combining the clustering intensity of the residual anomaly indication, the intensity characterization of the event indication, and the overlap ratio of the time window and the spatial intersection ratio, so that the output of the target pipe segment identifier not only includes the location result, but also includes a credibility metric that can be used for subsequent evidence solidification and visualization.

[0080] The positioning process of this embodiment is illustrated below with a specific example. Taking the flexible tube panel of a biomass boiler superheater as an example, during a single load platform operation, the anomaly identification module discovers from the fusionable data object that the residual between the displacement observation input and the model prediction value of a certain adjacent tube segment identifier set is consistently large over multiple consecutive time windows. After time-series aggregation, a high residual anomaly indication is formed, generating a candidate tube segment identifier set containing this adjacent tube segment identifier set. Almost simultaneously, a short-term impact event occurs in the structural acoustic data within the same time window, and event detection generates an event indication. Since the event response intensity of two structural acoustic sensors near a certain support point is significantly higher than that of other sensors, the anomaly identification module generates a sound source constraint domain covering the area adjacent to that support point.

[0081] Subsequently, the consistency constraint determination showed that the residual anomaly time window and the acoustic event time window highly overlapped, and the candidate pipe segment identifier set and the sound source constraint domain intersected spatially. The anomaly identification module ultimately identified the pipe segment identifier that was adjacent to the support point and had the highest residual anomaly indication as the target pipe segment identifier and output the mapping confidence parameters. This target pipe segment identifier was then used by the evidence package generation module to aggregate the corresponding time window data and model version information to form a verifiable chain of anomaly evidence.

[0082] Through the above embodiment, the anomaly identification module can utilize two complementary types of evidence, model residuals and structural acoustic events, in parallel under a unified and aligned data structure. By constraining temporal and spatial consistency, it can achieve a stable mapping from candidate pipe segments to target pipe segment identifiers and output mapping confidence parameters to support the solidification of subsequent evidence packages, thereby improving the reliability and traceability of anomaly localization and reducing the risk of misjudgment.

[0083] In some embodiments, the residual between the observed input corresponding to the pipe segment identifier and the model prediction value output by the coupled state model is calculated, and a set of candidate pipe segment identifiers and a residual anomaly indicator corresponding to the set of candidate pipe segment identifiers are generated based on the temporal or spatial aggregation of the residuals, including: Based on the alignment index and data availability mask in the fusionable data object, the effective observation sequence associated with the pipe segment identifier is selected, and the effective observation sequence is mapped to the observation input of the coupled state model; Obtain the model prediction values ​​output by the coupled state model for the pipe segment identification, and calculate the residual sequence between the observed input and the model prediction values; Perform temporal aggregation on the residual sequence to generate temporal anomaly scores, or perform spatial aggregation on the residual sequence according to adjacent pipe segment identifiers to generate spatial anomaly scores; The candidate pipe segment identifier set is determined based on the temporal anomaly score or the spatial anomaly score, and residual anomaly indication is generated for the candidate pipe segment identifier set.

[0084] Specifically, the temporal anomaly score is a quantitative indicator of anomalies obtained by aggregating the residual sequences of the same pipe segment identifier in the time dimension. It is used to characterize the persistence, accumulation, or abrupt change of residual anomalies within a time window. The spatial anomaly score is a quantitative indicator of anomalies obtained by aggregating the residual sequences of adjacent pipe segment identifiers in the spatial dimension. It is used to characterize the connectivity, consistency, or extent of expansion of residual anomalies within adjacent pipe segments.

[0085] First, regarding the selection of valid observation sequences and the mapping of observation inputs, the anomaly identification module parses the alignment index and data availability mask in the fusionable data objects. The alignment index is used to extract temperature distribution data frames, displacement sampling points, and their corresponding auxiliary fields on the same aligned time axis. The data availability mask is used to identify which sampling segments can be calculated within a specific time window. The anomaly identification module uses the pipe segment identifier as an index to select valid observation sequences associated with that pipe segment identifier and marked as usable by the data availability mask from the fusionable data objects, and organizes the valid observation sequences into observation inputs according to the observation mapping relationship of the coupled state model.

[0086] Specifically, when temperature distribution data is used as part of the observation input, the anomaly identification module can align the temperature distribution sequence to the spatial segment defined by the pipe segment identifier according to the spatial ranging mark; when displacement data is used as part of the observation input, the anomaly identification module can resample the displacement sequence to a time granularity that matches the temperature distribution frame according to the aligned time axis, thereby ensuring the temporal consistency of the observation input.

[0087] Furthermore, regarding model prediction acquisition and residual sequence calculation, the anomaly identification module acquires the model predictions output by the coupled state model for the corresponding pipe segment identifier. The model predictions correspond to the observed inputs on the same aligned time axis, and their generation process is influenced by the state evolution relationship and the parameters to be calibrated. The anomaly identification module calculates the difference between the observed inputs and the model predictions point-by-point within the same time window, forming a residual sequence. The residual sequence can contain at least one type of residual, either temperature distribution residual or displacement residual; when both types of residuals are used simultaneously, the anomaly identification module can organize the residual sequence into a multi-dimensional residual vector sequence to simultaneously characterize the coupling features of thermal and mechanical anomalies in subsequent aggregation.

[0088] Furthermore, regarding residual aggregation, the anomaly identification module performs temporal aggregation on the residual sequence to generate a temporal anomaly score, or performs spatial aggregation based on adjacent pipe segment identifiers to generate a spatial anomaly score. Temporal aggregation is used to emphasize the characteristic of "the same pipe segment continuously exhibiting anomalies over time." In one embodiment, temporal aggregation can accumulate residual amplitude or residual energy based on a sliding time window and use the accumulated value as the temporal anomaly score. In another embodiment, temporal aggregation can statistically analyze the number of consecutive segments with residuals exceeding a threshold and the distribution of consecutive segment lengths, and use the maximum length or weighted length of the consecutive segments as the temporal anomaly score, thereby suppressing misjudgments caused by short-term spikes.

[0089] Spatial clustering is used to emphasize the feature of "simultaneous anomalies or anomalous connections between adjacent pipe segments". The anomaly identification module can determine the identification relationship between adjacent pipe segments based on topological identifiers and perform consistent clustering on the residual sequence or temporal anomaly scores of adjacent pipe segments. For example, the connectivity range where the anomaly scores of adjacent pipe segments simultaneously exceed the threshold within the same time window can be used as the spatial anomaly score, or the weighted sum of the anomaly scores of adjacent pipe segments can be used as the spatial anomaly score, so that the candidate range can cover structural areas that are prone to coupling anomalies, such as the neighborhood of support points or the neighborhood of elbows.

[0090] Furthermore, regarding the determination of the candidate pipe segment identifier set and the generation of residual anomaly indicators, the anomaly identification module determines the candidate pipe segment identifier set based on temporal anomaly scores or spatial anomaly scores. Specifically, threshold judgment or sorting truncation methods can be adopted: when the temporal anomaly score exceeds a preset threshold, the corresponding pipe segment identifier is added to the candidate pipe segment identifier set; when the spatial anomaly score exceeds a preset threshold, the pipe segment identifiers within the corresponding connected domain are added to the candidate pipe segment identifier set; or when multiple pipe segments have anomaly scores simultaneously, the pipe segment identifiers with the highest anomaly scores are selected to form the candidate pipe segment identifier set.

[0091] Subsequently, the anomaly identification module generates residual anomaly indicators for the candidate pipe segment identifier set. The residual anomaly indicators include, but are not limited to, the candidate pipe segment identifier set, the anomaly score associated with each candidate pipe segment identifier, and the corresponding alignment time window index, so that the residual anomaly indicators can be subsequently judged for consistency with the structural acoustic event indicators and solidified by the evidence package generation module.

[0092] The candidate selection process of this embodiment is illustrated below with a specific example. Taking the flexible tube panel of a biomass boiler superheater as an example, the fusionable data object output by the edge processing module shows that there is a short-term packet loss in the displacement channel during a certain period. The data availability mask marks this segment as unavailable. The anomaly identification module selects the valid observation sequence and constructs the observation input based on the alignment index and the data availability mask. After eliminating the unavailable segment, the residual sequence is calculated. The results show that the displacement residual of a certain tube segment is consistently large in multiple consecutive time windows, and the temperature distribution residual also shows a stable deviation in the local segment corresponding to the spatial ranging mark; the temporal anomaly score obtained by temporal aggregation is significantly higher than that of other tube segments.

[0093] Meanwhile, two adjacent pipe segment identifiers also showed high anomaly scores, and the spatial anomaly scores within the connected domain exceeded the threshold. Based on this, the anomaly identification module identified the pipe segment identifiers within the connected domain as a set of candidate pipe segment identifiers and generated a residual anomaly indicator containing the anomaly score and alignment time window index. This indicator is then used for subsequent spatiotemporal consistency determination with structural acoustic event indicators and further location of the target pipe segment identifier.

[0094] Through the above embodiment, the anomaly identification module can stably construct the observation input and calculate the residual sequence under the constraints of alignment index and data availability mask. It forms quantifiable temporal anomaly scores and spatial anomaly scores through temporal and spatial aggregation, and generates a candidate pipe segment identifier set and residual anomaly indication accordingly. This enhances the robustness of candidate screening to noise spikes and bad data and improves the stability and verifiability of subsequent anomaly mapping.

[0095] In some embodiments, the evidence package generation module is specifically used for: Obtain the unified timestamp corresponding to the establishment of the exception mapping, and determine the event time window associated with the exception based on the unified timestamp; Based on the target pipe segment identification, pipe wall temperature distribution data, structural acoustic data and displacement data that fall within the event time window are extracted from the fusionable data objects to form time window data; Extract the labels of operating conditions that overlap with the event time window, and associate the operating condition event labels with the time window data; Read and generate the model version information corresponding to the status variables, residual abnormal indicators and event indicators, and write the model version information, target pipe segment identifier and unified time stamp together; The target pipe section identifier, time window data, operating condition event annotations, and model version information are encapsulated into an evidence package.

[0096] Specifically, model version information refers to the identification information used to uniquely identify the coupled state model configuration and parameter set version used when generating state variables, residual anomaly indicators and event indicators. It is used to describe at least the model structure version, parameter snapshot version and configuration version associated with gating rules or observation mapping rules, so that the same anomaly can reproduce the same calculation path according to the version at different times or under different environments.

[0097] First, regarding the determination of the event time window, the evidence package generation module first obtains the unified timestamp corresponding to the establishment of the anomaly mapping. This unified timestamp can be output synchronously by the anomaly identification module when outputting the target pipe segment identifier, or it can be read by the evidence package generation module from the mapping record output by the anomaly identification module. The evidence package generation module determines the event time window associated with the anomaly based on the unified timestamp.

[0098] In some examples, the event time window can be generated according to a preset window strategy, such as extending a preset pre-time period forward and a preset post-time period backward from a unified time mark to form a time window; or determining the start and end boundaries based on the aligned time window index recorded in the residual anomaly indication output by the anomaly identification module; when the structural acoustic event indication has more precise start and end times, the time window of the event indication can also be used first and the safety boundaries can be extended at both ends to ensure that the evidence package covers the key evolutionary segments before and after the anomaly occurs.

[0099] Furthermore, regarding the formation of time window data, the evidence package generation module extracts pipe wall temperature distribution data, structural acoustic data, and displacement data falling within the event time window from the fusionable data objects based on the target pipe segment identifier, thus forming time window data. To ensure the reproducibility of the time window data, the evidence package generation module retains the alignment index and data availability mask reference information associated with the time window data during extraction, enabling subsequent analysis to clearly define the correspondence between data slices and aligned time axes.

[0100] In some examples, for pipe wall temperature distribution data, the evidence package generation module extracts the temperature distribution sequence within the corresponding spatial ranging marker segment according to the target pipe segment identifier, enabling the evidence package to present the temperature distribution evolution of the abnormal pipe segment within the event time window. For displacement data, the evidence package generation module extracts the displacement sequence corresponding to the target pipe segment identifier and its sampling time. For structural acoustic data, the evidence package generation module extracts acoustic segments or acoustic feature sequences within the event time window and retains reference fields related to the spatial representation of the structural acoustic array to support subsequent verification of acoustic events. The resulting time window data is uniformly organized, enabling cross-source comparison around the same target pipe segment identifier and the same event time window.

[0101] Furthermore, regarding the association of operating condition event annotations, the evidence package generation module extracts operating condition event annotations that overlap with the event time window and associates these annotations with the time window data. Specifically, the evidence package generation module reads the encapsulated operating condition event annotations from the fusionable data object, filters out annotation records that are within or overlap with the event time window, and writes them into the evidence package as the operational context of the anomaly. To enhance the integrity of the evidence chain, the evidence package generation module can simultaneously write the event type, event time window, and confidence parameters of the operating condition event annotations, enabling reviewers to determine whether the anomaly occurred during periods of strong disturbance such as soot blowing, fuel switching, or load ramping, and the credibility of the identified operating condition event.

[0102] Furthermore, regarding the solidification of model version information, the evidence package generation module reads the model version information corresponding to the generated state variables, residual anomaly indicators, and event indicators, and associates and writes the model version information, target pipe segment identifier, and unified time stamp. Specifically, the evidence package generation module can read the model version information used to generate state variables from the state output record of the state estimation module, and read the version information used to generate residual anomaly indicators and event indicators from the mapping record of the anomaly identification module, and perform consistency verification on version information from multiple sources; when multiple version information exists, the evidence package generation module writes them into the version field of the evidence package according to their purpose to clarify the model configuration source on which different output items depend. By binding the model version information with the target pipe segment identifier and unified time stamp, the evidence package can accurately load the corresponding version of the model structure and parameter snapshot during subsequent reproduction.

[0103] Furthermore, regarding evidence package encapsulation, the evidence package generation module encapsulates the target pipe segment identifier, time window data, operating condition event annotations, and model version information into an evidence package. The encapsulated evidence package includes at least: a target pipe segment identifier and a unified time stamp for locating the anomaly object and time anchor point; an event time window for defining the range of evidence data; time window data for providing multi-source observation evidence; operating condition event annotations for providing the operational context; and model version information for providing the basis for reproducing the calculation path. In one embodiment, the evidence package generation module can also write the mapping confidence parameters and residual anomaly indication summary output by the anomaly identification module, so that the evidence package simultaneously contains the core indication information of "why the anomaly was determined to be valid," thereby reducing the secondary calculation overhead of the upper-level system.

[0104] The evidence package generation process of this embodiment is illustrated below with a specific example. Taking the flexible tube panel of a biomass boiler superheater as an example, the anomaly identification module maps a certain tube segment identifier to the target tube segment identifier within a certain platform operating condition, and outputs the unified time stamp and mapping confidence parameters corresponding to the establishment of the anomaly mapping. The evidence package generation module determines the event time window centered on the unified time stamp, and extracts the temperature distribution sequence, displacement sequence, and structural acoustic event fragments corresponding to the target pipe segment identifier within the time window from the fusionable data object to form time window data. Simultaneously, it extracts the operating condition event annotations overlapping with the time window, discovering that the time window contains a short-term soot blowing event annotation with a high confidence parameter. Therefore, this operating condition event annotation is bound to the time window data and written. Subsequently, the evidence package generation module reads the model version information used to generate thermal resistance degradation state quantities and constraint degradation state quantities, and reads the model version information used to generate residual anomaly indicators and event indicators. It associates the above version information with the target pipe segment identifier and the unified time stamp and writes it. Finally, it encapsulates the target pipe segment identifier, time window data, operating condition event annotations, and model version information into an evidence package and outputs it to the upper-level system, enabling the upper-level system to directly access the evidence package for review, display, and maintenance decision recording.

[0105] Through the above embodiment, the evidence package generation module can determine the event time window with a unified time stamp after the anomaly mapping is established, gather multi-source time window data around the target pipe segment identifier and associate it with the working condition event label, and solidify the model version information on which the generated state quantity and anomaly indication depend, forming a structured evidence package, thereby enhancing the verifiability and traceability of the anomaly conclusion, and reducing the dependence of subsequent reproduction analysis and operation and maintenance closed loop on the full backtracking of the original data.

[0106] In some embodiments, the multi-source sensing module further includes a non-destructive testing unit, which performs guided wave scanning on the flexible pipe group structure to output defect indication data. The edge processing module associates the defect indication data and the operating status data with a unified time stamp and writes them into a fusionable data object. The anomaly identification module further introduces defect indication data into the consistency constraints to correct the target pipe segment identification.

[0107] Specifically, guided wave scanning refers to a detection method that uses elastic guided waves propagating along the pipe body to detect internal and surface defects in pipes or pipe sections. It obtains echo characteristics related to defect reflection and scattering by exciting guided waves in the vicinity of the pipe section or header and receiving the echo response. Defect indication data refers to the data set output by guided wave scanning used to characterize the probability of defect presence and its location characteristics. It at least includes defect indication values ​​associated with pipe section identification and spatial location or section characterization corresponding to the defect indication values, and may further include defect type indications or confidence parameters.

[0108] Firstly, regarding guided wave scanning, the non-destructive testing (NDT) unit can be deployed in critical areas of flexible pipe structures, such as the connection between headers and pipe panels, the vicinity of elbows, or the vicinity of support points—locations with high stress concentration and corrosion thinning risk. When performing guided wave scanning, the NDT unit can scan according to a preset cycle or trigger a scan after the edge processing module detects a specific operating condition event marker or abnormal event indication. During the scanning process, the NDT unit excites guided waves to the target pipe segment or its coupled structural path and acquires the guided wave echo response at the receiving end. Based on the echo response, it extracts defect reflection-related features and outputs defect indication data. To facilitate alignment with subsequent pipe segment identification systems, the NDT unit can organize the defect indication data according to pipe segment identification, for example, outputting a record of "pipe segment identification—defect indication value—defect segment characterization," and writing a unified timestamp or an acquisition time field that can be converted to a unified timestamp into each record.

[0109] Furthermore, regarding the temporal correlation and writing of defect indication data, after receiving the defect indication data output by the non-destructive testing unit, the edge processing module correlates the defect indication data with the operating status data and writes it into the fusionable data object based on a unified time stamp. Specifically, when the time field output by the non-destructive testing unit and the unified time stamp are from the same source, the edge processing module can directly use the unified time stamp as the correlation key; when there is a deviation between the time field output by the non-destructive testing unit and the unified time base, the edge processing module can map it to the aligned time axis corresponding to the unified time stamp through a preset time alignment rule, and write the mapped unified time stamp into the defect indication data record. During writing, the edge processing module simultaneously retains the correspondence between the defect indication data and the pipe segment identifier, making the defect indication data a type of additional observational evidence in the fusionable data object, sharing the same time anchor point as temperature distribution, displacement, and acoustic events.

[0110] Furthermore, regarding the introduction of defect indication data to correct target pipe segment identifiers in the anomaly identification module, the anomaly identification module introduces a defect consistency constraint in addition to the existing time consistency constraint and spatial consistency constraint when performing consistency constraint determination. The defect consistency constraint is used to limit the target pipe segment identifier to be within the high indication range of the defect indication data, or to require that the candidate pipe segment identifier set and the defect segment indicated by the defect indication data satisfy a spatial intersection or proximity relationship.

[0111] Specifically, when the candidate pipe segment identifier set generated by residual anomaly indication contains multiple adjacent pipe segments, and the sound source constraint domain of structural acoustic event generation has a wide coverage, the anomaly identification module can preferentially select the pipe segment identifier with a higher defect indication value in the defect indication data as the target pipe segment identifier; when the target pipe segment identifier has been initially determined but there is a significant inconsistency with the defect indication data, the anomaly identification module can correct the target pipe segment identifier based on the defect consistency constraint, so that the final output target pipe segment identifier simultaneously satisfies the common constraints of residual evidence, acoustic event evidence and defect indication evidence, and update the mapping confidence parameters accordingly.

[0112] The following example illustrates the correction process of this embodiment. Taking a flexible tube panel of a biomass boiler superheater as an example, the anomaly identification module generates a candidate tube segment identifier set containing three adjacent tube segment identifiers based on residual temporal and spatial aggregation within a single platform operating condition. Structural acoustic event detection also generates an event indication and provides a sound source constraint domain covering the neighborhood of the support point within the same time window. However, due to the strong structural coupling near the support point, multiple feasible intersections still exist between the sound source constraint domain and the candidate tube segment identifier set, making it difficult for the target tube segment identifier to converge. At this point, the non-destructive testing unit triggers a guided wave scan after this time window, outputting defect indication data. This data shows that the defect indication value corresponding to one tube segment identifier is significantly higher than that of the other two tube segment identifiers, and the defect segment representation intersects with the spatial range of the candidate tube segment identifier set.

[0113] Furthermore, after the edge processing module writes the defect indication data into the fusionable data object based on a unified time stamp, the anomaly identification module introduces a defect consistency constraint into the consistency constraints. Ultimately, the pipe segment identifier with the highest defect indication value that simultaneously satisfies both temporal and spatial consistency is corrected to the target pipe segment identifier, and the mapping confidence parameters are updated. The evidence package generation module then solidifies the defect indication data from the guided wave scan, along with temperature distribution, displacement, and acoustic event fragments, into an evidence package, enabling the host system to verify the consistency of the three types of evidence—"state anomaly—event occurrence—defect indication"—within the same evidence chain.

[0114] Through the above embodiment, defect indication data formed by guided wave scanning is introduced on the basis of the original multi-source operation status monitoring, and the association with the unified time stamp is completed on the edge side. This enables the anomaly identification module to integrate defect evidence to correct and converge the target pipe segment identification within the consistency constraints, thereby improving the determinism and verifiability of the target pipe segment identification mapping and enhancing the evidence package's support capability for the maintenance closed loop.

[0115] The above embodiments have provided a detailed description of the composition, structure, and function of the operational status data monitoring system for the flexible tube assembly structure of the biomass boiler of this application. The implementation process of the operational status data monitoring method for the flexible tube assembly structure of the biomass boiler of this application will be described in detail below with reference to specific embodiments. Figure 2 This is a flowchart illustrating the method for monitoring the operational status data of a biomass boiler flexible tube assembly structure provided in this application embodiment, as shown below. Figure 2 As shown, the method may specifically include the following steps: S201, collects the operating status data of the flexible tube assembly structure. The operating status data includes tube wall temperature distribution data, structural acoustic data and displacement data, and writes a unified time stamp for the operating status data. S202, based on a unified time stamp, performs time alignment on the running status data, and performs data quality judgment and operational event labeling on the running status data to generate fusionable data objects; S203, based on the topological identifiers and constraint relationships of the flexible tube group structure, constructs a coupled state model, and performs online calibration of the coupled state model using fusionable data objects as observation inputs, outputting the state variables associated with the flexible tube group structure; S204, parallel computation of residual anomaly indicators of coupled state models and event indicators of structural acoustic data, and mapping anomalies to target pipe segment identifiers of flexible pipe group structures based on consistency constraints; S205, when the anomaly mapping is established, the corresponding time window data, working condition event annotations and model version information are aggregated based on the target pipe segment identifier to generate an evidence package associated with the anomaly; S206 outputs the target pipe segment identifier, status quantity, and evidence package to the upper system for use in monitoring the operational status data of the flexible pipe group structure.

[0116] It should be understood that the sequence number of each step in the above method embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0117] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although the technical solutions of this application have been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A monitoring system for the operational status of a flexible tube assembly structure in a biomass boiler, characterized in that, include: The multi-source sensing module is used to collect operational status data for the flexible tube assembly structure, including tube wall temperature distribution data, structural acoustic data, and displacement data, and write a unified time stamp for the operational status data. The edge processing module is used to perform time alignment on the running status data based on the unified time stamp, and to perform data quality judgment and working condition event labeling on the running status data to generate fusionable data objects. The state estimation module is used to construct a coupled state model based on the topological identifiers and constraint relationships of the flexible tube group structure, and to perform online calibration of the coupled state model with the fusionable data object as the observation input, so as to output the state variables associated with the flexible tube group structure. An anomaly identification module is used to calculate the residual anomaly indication of the coupled state model and the event indication of the structural acoustic data in parallel, and to map the anomaly to the target pipe segment identifier of the flexible pipe group structure based on consistency constraints. The evidence package generation module is used to generate an evidence package associated with the anomaly based on the time window data, working condition event annotations and model version information corresponding to the target pipe segment identifier when the anomaly mapping is established. The output interface module is used to output the target pipe segment identifier, status quantity and evidence package to the upper system for use in monitoring the operation status of the flexible pipe group structure.

2. The system according to claim 1, characterized in that, The multi-source sensing module is specifically used for: Obtain the set of pipe segment identifiers associated with the flexible pipe assembly structure and the corresponding measurement point mapping relationship of each pipe segment identifier. The measurement point mapping relationship is used to characterize the correspondence between the pipe wall temperature distribution acquisition channel, the structural acoustic acquisition channel and the displacement acquisition channel and the pipe segment identifier, respectively. Based on the measurement point mapping relationship, the pipe wall temperature distribution data, the structural acoustic data, and the displacement data are collected respectively, and a unified time stamp corresponding to the same time base is written for each collected data record. When collecting the pipe wall temperature distribution data, a spatial ranging marker corresponding to the pipe wall temperature distribution data is output using a distributed measurement method, and the spatial ranging marker is associated with the unified time marker and written into the operating status data.

3. The system according to claim 1, characterized in that, The edge processing module is specifically used for: Based on the unified time stamp, time alignment processing is performed on the pipe wall temperature distribution data, the structural acoustic data, and the displacement data, and an alignment index corresponding to each data channel is generated; For the aligned data of each data channel, extract quality parameters to characterize the validity of the data; Based on the quality parameters, a data availability mask is generated corresponding to each data channel. The data availability mask is used to indicate the available segment of the operating status data within the corresponding time window. Acquire operating condition data associated with the operating conditions of the biomass boiler, and align and associate the operating condition data with the operating status data based on the unified time stamp; Based on the joint features of the operating condition data and the operating status data, operating condition event identification is performed, operating condition event labels are generated, and event time windows and confidence parameters are written for each operating condition event label. The running status data, carrying the alignment index, data availability mask, and operating condition event annotations, is encapsulated to generate the fusionable data object.

4. The system according to claim 1, characterized in that, The construction of the coupled state model based on the topology identifiers and constraint relationships of the flexible tube group structure includes: Obtain the topology identifier of the flexible pipe assembly structure, the topology identifier including pipe segment identifier, connection relationship corresponding to the pipe segment identifier, and support point identifier associated with the connection relationship; Based on the topology identifier, a structural state network is constructed with pipe segment identifiers as nodes and connection relationships as edges, and constraint relationships are configured for each support point identifier in the structural state network; A set of coupled state variables is defined on the structural state network. The set of coupled state variables includes pipe wall temperature distribution state variables, pipe segment displacement state variables, and constraint state variables corresponding to the constraint relationship. A state evolution relationship is established for the set of coupled state variables. The state evolution relationship includes a coupled evolution term that takes the thermal boundary quantity and mechanical boundary quantity corresponding to the operating condition data as driving inputs, and writes the equivalent thermal resistance parameter and support point constraint parameter as parameters to be calibrated into the coupled state model. An observation mapping relationship is established, which is used to map the pipe wall temperature distribution data and displacement data into the observation inputs of the coupled state model.

5. The system according to claim 4, characterized in that, The step of performing online calibration of the coupled state model using the fusionable data object as observation input to output state variables associated with the flexible tube assembly structure includes: Based on the alignment index and data availability mask in the fusionable data object, the effective observation sequence for calibration is determined, and the effective observation sequence is aggregated into a pipe segment observation set according to the pipe segment identifier. The operating condition event label is used as a calibration gating condition. When the gating is passed, the pipe segment observation set within the event time window corresponding to the operating condition event label is input into the coupled state model to trigger an incremental update of the parameter to be calibrated. Based on the state evolution relationship and observation mapping relationship of the coupled state model, the model prediction value and observation residual corresponding to the pipe segment observation set are calculated, and the equivalent thermal resistance parameter and the support point constraint parameter are iteratively updated under the joint constraints of multiple parameters to obtain the parameter update amount associated with each pipe segment identifier. Based on the parameter update, thermal resistance degradation state quantity and constraint degradation state quantity are generated. Based on the pipe wall temperature distribution state variable and pipe segment displacement state variable, strain indication quantity associated with pipe segment identification is calculated, and strain risk state quantity is generated according to the strain indication quantity and constraint degradation state quantity.

6. The system according to claim 1, characterized in that, The anomaly detection module is specifically used for: Based on the fusionable data object, the residual between the observation input corresponding to the pipe segment identifier and the model prediction value output by the coupled state model is calculated, and a candidate pipe segment identifier set and a residual abnormality indicator corresponding to the candidate pipe segment identifier set are generated based on the temporal or spatial aggregation of the residuals. Event detection is performed on the structural acoustic data to generate event indications, and a sound source constraint domain is generated based on the location information of the structural acoustic array or a spatial representation equivalent to the location information. Based on the temporal consistency constraint between the residual abnormality indication and the event indication, and the spatial consistency constraint between the sound source constraint domain and the candidate pipe segment identifier set, the target pipe segment identifier is determined, and a mapping confidence parameter associated with the target pipe segment identifier is generated.

7. The system according to claim 6, characterized in that, The calculation of the residual between the observation input corresponding to the pipe segment identifier and the model prediction value output by the coupled state model, and the generation of a candidate pipe segment identifier set and a residual anomaly indicator corresponding to the candidate pipe segment identifier set based on the temporal or spatial aggregation of the residual, includes: Based on the alignment index and data availability mask in the fusionable data object, a valid observation sequence associated with the pipe segment identifier is selected, and the valid observation sequence is mapped to the observation input of the coupled state model; Obtain the model prediction value output by the coupled state model for the pipe segment identifier, and calculate the residual sequence between the observed input and the model prediction value; Perform temporal aggregation on the residual sequence to generate temporal anomaly scores, or perform spatial aggregation on the residual sequence according to adjacent pipe segment identifiers to generate spatial anomaly scores; The candidate pipe segment identifier set is determined based on the temporal anomaly score or the spatial anomaly score, and the residual anomaly indication is generated for the candidate pipe segment identifier set.

8. The system according to claim 1, characterized in that, The evidence package generation module is specifically used for: Obtain the unified time stamp corresponding to the establishment of the anomaly mapping, and determine the event time window associated with the anomaly based on the unified time stamp; Based on the target pipe segment identifier, pipe wall temperature distribution data, structural acoustic data and displacement data that fall within the event time window are extracted from the fusionable data object to form the time window data; Extract the operating condition event labels that overlap with the event time window, and associate the operating condition event labels with the time window data; Read and generate the model version information corresponding to the state variables, residual abnormal indicators and event indicators, and associate and write the model version information, the target pipe segment identifier and the unified time stamp; The target pipe segment identifier, time window data, operating condition event annotations, and model version information are encapsulated into the evidence package.

9. The system according to claim 1, characterized in that, The multi-source sensing module also includes a non-destructive testing unit, which performs guided wave scanning on the flexible pipe group structure to output defect indication data. The edge processing module associates the defect indication data with the operating status data based on a unified time stamp and writes it into the fusionable data object. The anomaly identification module further introduces the defect indication data into the consistency constraint to correct the target pipe segment identifier.

10. A method for monitoring the operational status data of a biomass boiler flexible tube assembly structure based on any one of the systems described in claims 1 to 9, characterized in that, include: The operation status data of the flexible tube assembly structure is collected. The operation status data includes tube wall temperature distribution data, structural acoustic data and displacement data, and a unified time stamp is written for the operation status data. Based on the unified time stamp, the operation status data is time aligned, and the operation status data is judged for data quality and labeled with operating conditions and events to generate a data object that can be fused. A coupled state model is constructed based on the topological identifiers and constraint relationships of the flexible tube group structure. The coupled state model is then calibrated online using the fusionable data object as the observation input, and the state variables associated with the flexible tube group structure are output. The residual anomaly indication of the coupled state model and the event indication of the structural acoustic data are calculated in parallel, and the anomalies are mapped to the target pipe segment identifier of the flexible pipe group structure based on consistency constraints. When the anomaly mapping is established, an evidence package associated with the anomaly is generated based on the time window data, operating condition event annotations, and model version information corresponding to the target pipe segment identifier. The target pipe segment identifier, the status quantity, and the evidence package are output to the upper-level system for use in monitoring the operational status data of the flexible pipe group structure.

Images