Multi-source collaborative water quality result traceability analysis method and system

Abstract

The invention relates to the technical field of substance analysis and water quality monitoring, in particular to a multi-source collaborative water quality result traceability analysis method and system. The method comprises the following steps: acquiring and standardizing field multi-source data, and generating an alignment index structure; performing quality control sample triggering and acceptance analysis to generate a drift diagnosis structure; constructing a watershed section topology, calculating an upstream and downstream relationship and a delay channel, screening abnormal candidates according to the upstream and downstream relationship and the delay channel, carrying out feature coding and joint inference, and generating an abnormal source candidate list; and finally, performing retest priority calculation, task work order arrangement, result write-back binding and structured arrangement, and generating a traceability report and an equipment calibration update structure. According to the method, physical topology, time-space lag and quality control data are deeply fused, so that the crossing from single-point quality control to drainage basin collaborative traceability is realized, an abnormal source can be automatically and accurately positioned, a complete closed loop of analysis-decision-action-calibration is formed, and the traceability efficiency and the automation level of water quality monitoring are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of material analysis and water quality monitoring technology, and in particular to a method and system for tracing the source of multi-source collaborative water quality results. Background Technology

[0002] In the field of material analysis and water quality monitoring technology, existing solutions for multi-point water quality monitoring and laboratory validation in watersheds typically consist of on-site multi-parameter raw data acquisition, sample metadata recording, and laboratory quality control processes. These solutions suffer from limitations such as inconsistent timeframes and sites for on-site multi-parameter raw data and sample metadata, fragmented zero-point calibration and range calibration, and disconnected triggering, rule-based acceptance, and batch determination for blank, spiked, and parallel samples. Existing methods largely rely on fixed operating procedures and discrete records, which, under resource constraints and cross-site collaboration scenarios, are prone to missing alignment index structures and insufficient basis for linkage analysis. This makes it difficult to stably generate alignment index structures, drift diagnosis structures, spatiotemporal relationship diagrams, anomaly source candidate list structures, and report and calibration update structures. For the joint processing of on-site multi-parameter raw data, sample metadata, and quality control plans, existing technologies generally suffer from discontinuities in links such as time alignment, site coding, linkage analysis, retest priority calculation, work order arrangement, write-back binding, and retest result linkage. This makes it difficult to form a unified process that runs through blank samples and spiked samples and parallel samples triggering, rule acceptance and batch judgment, cross-sectional topology construction and upstream and downstream relationship calculation and lag estimation, feature coding and joint inference in the application scenario of collection-alignment-judgment-control-recording. As a result, the structure of the anomaly source candidate list and the structure of the report and calibration update are lagging and inconsistent. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a multi-source collaborative water quality result tracing analysis method, comprising: Acquire raw data of multiple parameters and sample metadata and calibration update structure on site, perform timestamp standardization and site coding standardization, zero point calibration and range calibration processing, and generate an alignment index structure; Perform blank sample, spiked sample and parallel sample triggering processing, extract standard sample data from quality control sequence records, perform rule acceptance and batch judgment on parallel sample data and environmental records, and perform linkage analysis on quality control acceptance results to generate drift diagnostic structure; The process involves constructing a cross-sectional topology based on station indexes, geographic location fields, and cross-sectional level fields. Nodes are designated as stations, and edges represent water body connectivity paths. Upstream and downstream relationships are calculated to form a list of upstream and downstream relationships. Within the node attachment time index range, latency estimation is performed. Response peak sequences and background segments are searched on nodes with sufficient time coverage and stable channel segments to generate latency channels containing start and end times, reference nodes, affected nodes, and window boundary markers. Cross-sectional adjacency, flow direction constraints, and latency channels are extracted from the spatiotemporal relationship graph for anomaly candidate screening. Screening criteria include a sequential relationship and a difference between upstream and downstream fragment summaries within the same adjacency window, occurring within the latency channel's defined range. Feature encoding fields include adjacency level, direction markers, latency labels, candidate density labels, and diagnostic reference labels. Joint inference is performed on collaborative tracing features. Within the same connectivity path, segments are aggregated according to direction markers to form candidate links. Inference is then generated based on the combination of adjacency level, latency labels, and candidate density labels, resulting in an anomaly source candidate list structure. Perform retest priority calculation, work order arrangement, write-back binding, retest result linking, and structured organization to generate reports and calibration update structures.

[0004] Furthermore, the process of constructing a cross-sectional topology based on the site index, geographic location field, and cross-sectional level field, where nodes are sites and edges are water body connectivity paths, also includes: Based on the site index, geographic location field, and cross-section level field in the aligned index structure, cross-section topology is constructed for monitoring stations within the same watershed. Nodes are stations, and edges are possible water body connection paths. The path direction is determined based on the cross-section level and topographic flow direction records. When there is a conflict between historical connection records and recent maintenance registrations, two candidate edges are retained, and the source and effective range are marked in the edge annotation.

[0005] Furthermore, the process of calculating upstream and downstream relationships to form a list of upstream and downstream relationships and indexing them within the node attachment time range also includes: Establish parent-child relationships between node pairs according to the direction of the connected path, forming an upstream and downstream relationship list, and attach the time index range in the alignment index structure to each node for window limitation in subsequent time delay estimation; if a node is found to lack a connected path or has a broken segment, it is recorded as a broken segment placeholder and will not participate in the generation of time delay channels in the future.

[0006] Furthermore, the process of delay estimation, which searches for response peak sequences and background segments on node pairs with sufficient time coverage and stable channel segments to generate a delay channel containing start and end times, reference nodes, affected nodes, and window boundary markers, also includes: Lag estimation is performed on the upstream and downstream relationship list. The processing targets are node pairs with sufficient time coverage and channel stability segments. The time coverage of the node pairs comes from the time index of the alignment index structure, and the criteria for channel stability segments refer to the window type and status label in the drift diagnosis structure. Within the candidate time window, the response peak sequence and background segment of the upstream and downstream nodes are searched, the order and interval range of the responses of the two nodes are recorded, and a lag channel bound to the node pair is generated.

[0007] Furthermore, the process of screening abnormal candidates based on the condition that the upstream and downstream fragment summaries within the same adjacent window have a sequential or dissimilar relationship and occur within the time-lag channel limit also includes: For each time segment list, the three types of elements—upstream node response changes, downstream node response changes, and background changes—are compressed into fixed fields. Then, anomaly candidate screening is carried out. The screening criteria are that the upstream segment summary and the downstream segment summary within the same adjacent window have a sequential relationship and a difference relationship, and the difference relationship occurs within the time lag channel limit. Segments that meet the criteria are registered as anomaly candidates.

[0008] Furthermore, the process of jointly inferring collaborative source tracing features, segmenting and aggregating candidate links according to directional labels within the same connected path to form candidate links, and inferring based on the combination relationship of adjacency level, lag label, and candidate density label also includes: Within the same connected path, segments are aggregated according to direction labels to form several candidate links. Each candidate link consists of multiple candidate entries that satisfy the time delay label constraint in sequence. Joint inference is performed on each candidate link, based on the combination relationship between the adjacency level and the time delay label and the candidate density label.

[0009] Furthermore, the process of calculating retest priority and scheduling work orders also includes: First, read the link segment set and restricted markers in the candidate list structure of the anomaly source, aggregate candidate entries by site index, and construct the unit to be retested; then read the personnel scheduling, equipment status and safety requirements in the resource constraint configuration, generate executable time window and equipment binding constraints; perform retest priority calculation, the calculation is based on candidate density label, direction marker continuity, lag label, diagnostic reference set coverage, restricted marker strength and resource consumption intensity.

[0010] Furthermore, the work order arrangement process also includes: Within the executable time window, the site, channel, sample type, transportation route, and temperature control requirements are packaged into a site task. The site task includes task number, site index, time window, channel identifier, sample type set, pretreatment requirements, cold chain temperature range, return path, and receiving unit. At the same time, a parameter set is generated, which includes sampling volume, storage method, pretreatment device parameters, spiking source batch number, labeling rules, and barcode scanning caliber.

[0011] Furthermore, the process of generating reports and updating the calibration structure also includes: The calibration update structure is constructed as follows: Read the paragraph-level control entries and quality control reference summaries from the traceability analysis results, extract the zero-point offset trigger evidence, range mapping trigger evidence and temperature compensation trigger evidence from the control entries, and map them to the applicable channels and equipment list; generate a configuration fragment for each piece of evidence, which includes the correction category, scope of application, version number and validity period identifier.

[0012] Furthermore, a multi-source collaborative water quality result tracing and analysis system, applied to any of the methods mentioned above, includes: The metadata alignment and consistency calibration module is used to receive raw data of multiple parameters on site, sample metadata, and calibration update structure, perform timestamp standardization, site coding standardization, zero-point calibration, range calibration, and output alignment index structure. The quality control triggering and acceptance module is connected to the metadata alignment and consistency calibration module. It is used to read the alignment index structure and quality control plan, trigger blank samples, spiked samples, and parallel samples, complete rule acceptance and batch judgment, and output the quality control acceptance results. The linkage analysis and drift diagnosis module is connected to the quality control triggering and acceptance module. It is used to perform linkage analysis on the quality control acceptance results and generate a drift diagnosis structure. The spatiotemporal relationship construction module is connected to the metadata alignment and consistency calibration module and the linkage analysis and drift diagnosis module. It is used to perform cross-sectional topology construction, upstream and downstream relationship calculation, and lag estimation based on the alignment index structure and drift diagnosis structure, and outputs a spatiotemporal relationship graph. The collaborative source tracing and inference module is connected to the spatiotemporal relationship construction module. It is used to extract cross-sectional adjacency, flow direction constraints, and time-delay channels from the spatiotemporal relationship graph, perform anomaly candidate screening, feature encoding, joint inference, and output an anomaly source candidate list structure. The retest scheduling and work order orchestration module is connected to the collaborative tracing and inference module. It is used to read the candidate list structure of anomaly sources and resource constraint configuration, calculate the retest priority and complete the work order orchestration to form a sampling scheduling plan. The write-back binding and retesting module is connected to the retesting scheduling and work order orchestration module. It is used to extract the site task and parameter set from the sampling scheduling plan, perform write-back binding and retesting result hooking, generate source tracing analysis results, and send the reference identifier back to the collaborative source tracing inference module. The report generation and calibration update module, connected to the write-back binding and retesting module, is used to organize the traceability analysis results in a structured manner, output the report and calibration update structure, and send the calibration update structure back to the metadata alignment and consistency calibration module.

[0013] The key innovations of this invention include: (1) Set up write-back binding and retest result linking in the through link, connect the alignment index structure, the anomaly source candidate list structure and resource constraint configuration, and output report and calibration update structure.

[0014] (2) Combine cross-sectional topology construction, upstream and downstream relationship calculation and time delay estimation in a single process, and connect with feature encoding and joint inference processing to generate an anomaly source candidate list structure.

[0015] (3) The online quality control chain is formed by triggering blank samples, spiked samples, and parallel samples around the quality control plan, along with rule acceptance, batch judgment, and linked analysis and processing, generating a drift diagnostic structure. The following are its main beneficial effects.

[0016] The following are its main beneficial effects: (1) It is applied to the joint scenario of the alignment index structure and the candidate list structure of the abnormal source. By writing back and binding and linking the retest results, it connects the record link from the retest priority calculation and work order arrangement to the report and calibration update structure, forming a traceable source and version association, which is suitable for continuous processing under resource constraint configuration conditions.

[0017] (2) The spatiotemporal processing link acting on the alignment index structure and drift diagnosis structure constraints outputs an anomaly source candidate list structure through cross-section topology construction, upstream and downstream relationship calculation and time delay estimation, continuation feature encoding and joint inference processing. It provides candidate location in the connected path and direction constraint scenario and is applicable to multi-section and multi-path association scenarios.

[0018] (3) It acts on the sampling and acceptance link driven by the quality control plan. It triggers the acceptance and batch judgment of the succession rules through blank samples, spiked samples and parallel samples and enters the linkage analysis and processing output drift diagnosis structure to form a consistent expression of batch boundary and judgment basis. It is suitable for online quality control and traceability collaboration scenarios. Attached Figure Description

[0019] Figure 1 A flowchart illustrating a multi-source collaborative water quality result tracing analysis method provided in this application embodiment; Figure 2 This is a structural block diagram of a multi-source collaborative water quality result tracing analysis method provided in an embodiment of this application. Detailed Implementation

[0020] Example 1: Refer to Figure 1 This is a flowchart illustrating a multi-source collaborative water quality result tracing analysis method provided in an embodiment of the present invention. The process may include at least steps S100-S400: S100: Acquire raw data of multiple parameters and sample metadata and calibration update structure on site, perform timestamp standardization and site coding standardization, zero point calibration and range calibration processing, and generate alignment index structure; S200, performs blank sample, spiked sample and parallel sample triggering processing, extracts standard sample data from quality control sequence records, performs rule acceptance and batch judgment on parallel sample data and environmental records, and performs linkage analysis on quality control acceptance results to generate drift diagnostic structure; S300. Construct cross-sectional topology based on station index, geographic location field, and cross-sectional level field. Build a topology with nodes as stations and edges as water body connection paths. Calculate upstream and downstream relationships to form an upstream and downstream relationship list. Within the node attachment time index range, perform time lag estimation. Search for response peak sequences and background segments on nodes with sufficient time coverage and stable channel segments to generate time lag channels containing start and end times, reference nodes, affected nodes, and window boundary markers. Extract cross-sectional adjacency, flow direction constraints, and time lag channels from the spatiotemporal relationship graph. Perform anomaly candidate screening. The screening condition is that the upstream segment summary and downstream segment summary within the same adjacency window have a sequential relationship and a difference relationship, and occur within the time lag channel limit. Feature encoding. The encoding fields include adjacency level, direction marker, time lag label, candidate density label, and diagnostic reference label. Perform joint inference on collaborative tracing features. In the same connected path, segment and aggregate according to direction marker to form candidate links. Based on the combination relationship of adjacency level, time lag label, and candidate density label, infer and generate anomaly source candidate list structure. S400 performs retest priority calculation, work order arrangement, write-back binding, retest result linking, and structured organization to generate reports and calibration update structures.

[0021] S100: Acquire raw data of multiple parameters and sample metadata and calibration update structure on site, perform timestamp standardization and site coding standardization, zero point calibration and range calibration processing, and generate alignment index structure.

[0022] The steps take raw multi-parameter data from the field, sample metadata, and calibration update structure as inputs. The raw multi-parameter data from the field is an unprocessed data set collected by water quality sensors deployed at watershed sections and stations, including channel name, acquisition time, device number, channel range, storage medium location, and acquisition batch identifier. The sample metadata is a set of descriptive information related to manual sampling, transportation, and laboratory confirmation, including sample number, sampling container and preservation method, storage and transportation records, sampler information, field environment records, and reference sampling strategy. The calibration update structure is a set of configurations for the preceding closed-loop feedback, including zero-point correction items, range mapping items, temperature compensation configuration, applicable channel and device list, version number, and expiration date identifier. Specifically, the process begins with timestamp standardization, unifying the time fields of the original multi-parameter data and sample metadata to the same time scale. Regular mapping is then applied to day-night cycles, leap days, and regional time differences in cross-device records, creating a comparable time series. Next, site coding standardization is performed, standardizing the mapping of site names, administrative divisions, geographical locations, and cross-sectional levels based on the site master data dictionary. This resolves cases of identical names with different codes and vice versa, and the mapping relationships are recorded in the data annotation area. Furthermore, for duplicate sampling, missing fields, and abnormal time sequences in the collection records, fingerprints and sampling batch identifiers are used for merging, removal, and replenishment, retaining original shadow copies and processing markers for subsequent traceability. In essence, after the above processing, a candidate record set is formed, consisting of time, site, device, channel, batch, and sample number. The version number and applicable scope of the calibration update structure are then linked at the record level to guide subsequent calibration applications. The output field obtained after the above processing is named the alignment candidate set. The alignment candidate set is registered in the same data container and serves as the input for the next step to extract sampling records and instrument status from the alignment candidate set. It also serves as the time and site basis for the subsequent main step to construct the spatiotemporal relationship graph, which can be reused.

[0023] In the process of extracting sampling records and instrument status from the alignment candidate set, the sampling record refers to the subset of records aggregated from the candidate records by sample number and sampling batch, including sampling start time, sampling end time, volume or isochronous sampling settings, sampling method, and on-site environmental records; the instrument status refers to the set of status descriptions related to equipment operation, including equipment self-test logs, channel health markers, sensor unit replacement records, temperature and pressure compensation switch status, channel zero-point history, and range setting values. Specifically, the alignment candidate set is first grouped by sample number and site code, and the sampling time range, on-site environmental records, and sampling batch identifiers within each group are extracted to form sampling records that can be directly used for calibration; then, the instrument status within the same group is read by equipment number and channel name, and the status switching point and sensor unit replacement time point are aligned to the sampling time range to define the calibration window. Furthermore, based on the applicable channels and equipment list in the calibration update structure, zero-point calibration and range calibration are triggered: Zero-point calibration uses the blank medium response segment and the equipment static stability segment, selecting representative segments according to the rules of time continuity and noise stability, and calculating and registering the zero-point correction item; Range calibration uses the standard sample response segment and the field standard curve segment, performing segment mapping on the responses near the high and low points, and registering the range mapping item and applicable range; When an abnormal self-test flag is detected in the equipment status or the channel health flag is limited, the range calibration of that channel is paused, only the zero-point correction item is output, and the reason is recorded in the annotation area. When there is insufficient time overlap between the sampling record and the instrument status or the sample number is missing, the missing record registration is triggered, and an alarm flag is attached to the output of this step. The output field generated by the above processing is named Consistency Calibration Parameter. The content includes zero point correction item, range mapping item, temperature compensation configuration, applicable channel and version number and validity period identifier and notes area record. The application action of Consistency Calibration Parameter in this main step is to apply the Consistency Calibration Parameter, and it serves as the input for the application of Consistency Calibration Parameter in the next step, and as a reference configuration for quality control and drift diagnosis in subsequent main steps.

[0024] In the application of consistency calibration parameters, for each record in the alignment candidate set, the corresponding zero-point correction item, range mapping item, and temperature compensation configuration are matched according to the applicable channel and equipment list and version number of the consistency calibration parameters. The calibration window is divided according to the recording time and equipment status, and the channel response is mapped to the consistency metrological scale within each window. Specifically, for records with blank medium response segments, the zero-point correction item is applied first, and the values ​​before and after correction and the correction item version are recorded as record-level notes; for records with standard sample response segments or field standard curve segments, the range mapping item is applied, and responses exceeding the applicable range are marked with validity and suspended for review during the mapping process; for channels with temperature and pressure compensation switch status records, a compensation mark is written at the record level according to the temperature compensation configuration to characterize the processing method and source. When there are version conflicts across devices or channels, the effective version is selected according to the priority strategy in the consistency calibration parameters, and the ineffective version is recorded in the record-level notes for subsequent quality control acceptance and drift diagnosis review. Furthermore, to support the joint processing of subsequent multi-site data, a retrieval-oriented index is established at the record level. This index consists of a time index, site index, device index, channel index, sample number index, and version number index, and is organized by time period and site level. Understandably, after completing the above application and index construction, a unified entry point and access path are formed for subsequent quality control and traceability processing. The output field generated by the above processing is named the Alignment Index Structure. This Alignment Index Structure serves as the direct input for the next main step of obtaining the Alignment Index Structure and quality control plan, triggering the online quality control sequence for blank samples, spiked samples, and parallel samples. Simultaneously, its time index and site index will be directly reused in subsequent main steps when constructing cross-sectional topology and calculating upstream and downstream relationships and lag estimation, thus maintaining the same reference during data access across main steps.

[0025] In summary, the technical effects of this step are as follows: By linking the processing of raw multi-parameter data, sample metadata, and calibration update structure on-site, an aligned index structure that can be retrieved and reused is formed. Versioned traces of consistent calibration parameters are written at the record level, establishing an orderly link from data acquisition records and instrument status to unified metrological scales and queryable indexes. This facilitates subsequent online quality control, drift diagnosis, and collaborative traceability under the same data view.

[0026] S200, performs blank sample, spiked sample and parallel sample triggering processing, extracts standard sample data from quality control sequence records, performs rule acceptance and batch judgment on parallel sample data and environmental records, and performs linkage analysis on quality control acceptance results to generate drift diagnostic structure; Step S200 includes at least steps S210-S230: S210. Obtain the alignment index structure and quality control plan, perform blank sample, spiked sample and parallel sample triggering processing, and obtain the quality control sequence record; The steps involve taking the alignment index structure generated in the previous main step and the pre-set quality control plan as input. The alignment index structure is a retrieval-oriented index set, including time index, site index, device index, channel index, sample number index, and version number index. The quality control plan is an online quality control configuration set maintained in the method library, including sample type definition, trigger frequency, trigger timing, applicable channels, acceptance rule identifier, and batch caliber. Specifically, the aforementioned alignment index structure is used as the basis for time and site positioning. The monitoring section is located by the site index, and the sampling window is located by the time index. The set of channels participating in this batch's quality control is determined based on the device index and channel index. Simultaneously, the sample type definition in the quality control plan is read, where blank samples are media samples without the target component to be tested, spiked samples are samples with a known source and fixed value of the target component added to the sample to be tested, and parallel samples are samples repeatedly collected and pre-processed using the same method in the same sampling window. Furthermore, based on the triggering frequency and timing, sample types are mapped to specific windows: when the sampling window matches the planning window, a blank sample generation task is triggered; when the channel set contains target components requiring linear coverage, a spiked sample generation task is triggered; when the sampling window is identified as a critical section or a methodology switch, a parallel sample generation task is triggered. Understandably, each triggered task is bound to the site index, time index, and sample number index, and includes the sample type, methodology identifier, and acceptance rule identifier. During the triggering process, if a channel index is found to be marked as disabled or under maintenance, the triggering of that channel is suspended, and the reason for disabling is recorded. After triggering, the triggered task is added to the task queues of the Laboratory Information Management System (LIMS) and the field terminal. In the task feedback stage, tasks that are not fed back according to the window or whose feedback information is incomplete are registered as exceptions and associated with the version number index. After the above processing is completed, a structured output field named "Quality Control Sequence Record" is generated. This record is stored in the data container of this step and serves as direct input for the next step, extracting standard sample data, parallel sample data, and environmental records from the Quality Control Sequence Record. It can also be used to backtrack batch boundaries in subsequent main steps. In summary, the technical effect of this step is: by linking the alignment index structure with the quality control plan, a quality control sequence record strictly bound to time, site, and channel is formed, constructing a task list that can be directly invoked for subsequent acceptance judgment and linkage analysis.

[0027] S220. Extract standard sample data, parallel sample data and environmental records from the quality control sequence records, perform rule acceptance and batch judgment, and generate quality control acceptance results. The steps above use the quality control sequence records generated in the previous sub-step as input. Standard sample data consists of a set of measurement results from standard substances or certified reference materials; parallel sample data consists of a set of results from repeated measurements within the same sampling window; and environmental records are a set of condition records during sampling, pretreatment, and measurement, including temperature, humidity, sample storage and transportation status, pretreatment time period, and equipment operating status. Specifically, the sample type in the quality control sequence records is used as a search key to retrieve the corresponding measurement results from the data acquisition container and LIMS. For blank samples, the responses of the blank medium in each channel and related background information are read; for spiked samples, paired responses before and after spiking and methodological identifiers are read; and for parallel samples, grouped repeated responses and corresponding timestamps and processing batches are read. Furthermore, based on the acceptance rule identifiers carried in the quality control sequence records, a standard operating procedure (SOP) rule set matching the methodology is loaded from the rule base. The rule set and readback results are mapped along the sample number and channel dimensions. Background checks are performed on blank samples, recovery checks on spiked samples, and difference checks on parallel samples. When results are missing, or environmental records show abnormal shutdowns or preprocessing interruptions, the record is marked as pending review, and the reason and responsibility attribution are recorded. Subsequently, samples in the current period are aggregated by batch categorization using site and time indices. The corresponding judgment identifier for the rule set is calculated, and batch-level pass, warning, or rejection identifiers and corresponding explanations are generated. When multiple pending review records exist within the same batch, the batch identifier is set to pending review, and this status is written back to the status bit of the quality control sequence record. The output field generated by the above processing is named "Quality Control Acceptance Result". Within this step, the Quality Control Acceptance Result establishes a bidirectional reference with the sample number index and channel index, serving as input for the next step's linked analysis of the Quality Control Acceptance Result. It also provides batch boundary information in the subsequent main steps of cross-sectional topology construction, upstream and downstream relationship calculation, and lag estimation. In summary, the technical effect of this step is: through rule-based acceptance and batch determination of standard sample data, parallel sample data, and environmental records, it outputs information units with rule sources, methodological identifiers, and status bits, forming a Quality Control Acceptance Result that can be directly consumed by subsequent linked analysis.

[0028] S230. Perform a linkage analysis on the quality control acceptance results to generate a drift diagnosis structure; The quality control acceptance result output from the previous sub-step serves as input. This result includes sample number, channel identifier, rule source, judgment identifier, status bit, and batch caliber. It can be mapped to the time index, site index, and device index in the alignment index structure via the sample number index. Specifically, the batch sequence is first expanded on the time index. Judgment identifiers under the same site index and channel index are formed into a time series. Consecutive warning or rejection segments are identified as suspected drift windows. Records with a status bit indicating "pending review" are marked as not participating in threshold judgment, but their timestamps are retained for boundary reference. Further, within the suspected drift window, the device index and version number index in the alignment index structure are read back to check corresponding channel version switching, maintenance registration, and methodology switching records. When a window overlaps with a version switching or maintenance registration in time, it is marked as a device event-related window. When there is no overlap in device events but an abnormally concentrated distribution of environmental records, it is marked as an environmental event-related window. Subsequently, a linkage analysis was conducted based on window type and different calibers: within the equipment event-related window, the focus was on evaluating the relative change in response under the same methodology and the consistency of rule sources; within the environmental event-related window, the focus was on evaluating the background baseline and repeatability consistency under the same batch caliber; within unrelated windows, they were recorded as windows to be investigated and associated with the consistency calibration parameter version number of the previous main step for subsequent verification. Through linkage analysis, a structured diagnostic entry was generated for each suspected window. The entry included a drift type marker, start and end times, a list of impact channels, a reference rule identifier, batch caliber, and a placeholder for suggested actions, and a reverse reference was established between the entry and the original quality control acceptance results. After completing all window processing, the entries were aggregated to form an output field name drift diagnostic structure. This drift diagnostic structure served as input for the next main step to obtain the alignment index structure and drift diagnostic structure, used for cross-sectional topology construction, upstream and downstream relationship calculation, and lag estimation. Simultaneously, the placeholders for suggested actions in the entries were bound to specific write-back strategies in the report and calibration update structures of subsequent main steps. The technical effects of this step can be summarized as follows: By performing a linked analysis of the quality control and acceptance results across the dimensions of time index, site index, and device index, a drift diagnostic structure with type marking, boundary information, and reference relationships is generated, providing a directly accessible diagnostic basis for subsequent spatiotemporal constraint construction and collaborative tracing.

[0029] S300. Construct cross-sectional topology based on station index, geographic location field, and cross-sectional level field. Build a topology with nodes as stations and edges as water body connection paths. Calculate upstream and downstream relationships to form an upstream and downstream relationship list. Within the node attachment time index range, perform time lag estimation. Search for response peak sequences and background segments on nodes with sufficient time coverage and stable channel segments to generate time lag channels containing start and end times, reference nodes, affected nodes, and window boundary markers. Extract cross-sectional adjacency, flow direction constraints, and time lag channels from the spatiotemporal relationship graph. Perform anomaly candidate screening. The screening condition is that the upstream segment summary and downstream segment summary within the same adjacency window have a sequential relationship and a difference relationship, and occur within the time lag channel limit. Feature encoding. The encoding fields include adjacency level, direction marker, time lag label, candidate density label, and diagnostic reference label. Perform joint inference on collaborative tracing features. In the same connected path, segment and aggregate according to direction marker to form candidate links. Based on the combination relationship of adjacency level, time lag label, and candidate density label, infer and generate anomaly source candidate list structure. Step S300 includes at least steps S310-S330: S310. Obtain the alignment index structure and drift diagnosis structure, perform cross-sectional topology construction, upstream and downstream relationship calculation and time delay estimation to obtain the spatiotemporal relationship diagram. The steps utilize the aforementioned alignment index structure as a unified entry point for time series, stations, devices, and channels, and the aforementioned drift diagnostic structure as the diagnostic basis for channel status and window type. Specifically, firstly, based on the station index, geographic location field, and cross-sectional level field in the alignment index structure, cross-sectional topology is constructed for monitoring stations within the same watershed. Nodes represent stations, and edges represent possible water body connection paths. The path direction is determined based on the cross-sectional level and topographic flow direction records. When there is a conflict between historical connection records and recent maintenance registrations, two candidate edges are retained, and the source and effective range are marked in the edge annotations. Subsequently, upstream and downstream relationship calculations are performed, establishing parent-child relationships between node pairs according to the connection path direction to form an upstream and downstream relationship list. The time index range in the alignment index structure is attached to each node for subsequent window limitation for time lag estimation. If a node is found to lack a connection path or has a broken segment, it is recorded as a broken segment placeholder and will not participate in the subsequent time lag channel generation. Furthermore, time lag estimation processing is performed on the upstream and downstream relationship list. The processing targets are node pairs with sufficient time coverage and stable channel segments. The time coverage of the node pairs comes from the time index of the alignment index structure, and the criteria for the stable channel segment refer to the window type and status label in the drift diagnosis structure. Within the candidate time window, the response peak sequence and background segment of the upstream and downstream nodes are searched, and the order and interval range of the responses of the two nodes are recorded to generate time lag channels bound to the node pairs. The time lag channel includes the start and end time, reference node, affected node, and window boundary marker. When the window falls into the device event association window, only time lag channels marked as restricted are generated, and in subsequent steps, it is indicated that they are only used for auxiliary judgment in the screening of abnormal candidates. After completing the above processing, the cross-sectional topology, upstream and downstream relationship lists, and time-delay channels are integrated. Reverse references to the drift diagnosis structure are written into the node and edge structures, forming a structured output field name spatiotemporal relationship graph. This spatiotemporal relationship graph is registered in the data container of this main step, serving as the direct input for the next sub-step to extract cross-sectional adjacency and flow direction constraints and time-delay channels from the spatiotemporal relationship graph. Simultaneously, it serves as a shared foundation across main steps, accessing time and space constraints through adaptive sampling scheduling and report generation in subsequent main steps. In summary, the technical effect of this step is: through the joint processing of the alignment index structure and the drift diagnosis structure, a spatiotemporal relationship graph containing connectivity, direction, and time-delay elements is constructed, forming a unified constraint carrier for subsequent filtering, encoding, and inference.

[0030] S320. Extract cross-sectional adjacency, flow direction constraints, and time-delay channels from the spatiotemporal relationship diagram, perform anomaly candidate screening and feature encoding, and generate collaborative source tracing features. The process uses a spatiotemporal graph as a single input. Here, cross-sectional adjacency refers to pairs of nodes that are spatially directly in contact on the same connected path; flow direction constraints refer to the binding rules between path direction and parent-child relationships; and time lag channels refer to the set of response sequence and interval ranges recorded on upstream and downstream node pairs. Specifically, firstly, cross-sectional adjacency relationships are read from the node set of the spatiotemporal graph. Adjacency windows are constructed according to the edge direction markers and cross-sectional hierarchical numbers. Each adjacency window is then linked with the time index segments of the upstream and downstream nodes in the alignment index structure, resulting in a candidate set that can be used to compare time-series segments. Next, flow direction constraints are read from the edge set of the spatiotemporal graph. Candidate sets with unclear directions or those in placeholder positions of disconnected segments are eliminated, retaining adjacency windows with clear directions. Finally, the time lag channels bound to node pairs are read and overlapped with the candidate sets to limit the alignment segment range within each adjacency window, forming a list of time segments for anomaly retrieval. Furthermore, for each time segment list, without introducing new methodological rules, the three elements of upstream node response change, downstream node response change, and background change within the segment are compressed into fixed fields. These fields include segment identifier, upstream segment summary, downstream segment summary, and background summary. Simultaneously, the window type marker of the drift diagnostic structure is read back and a reference relationship is established with the segment identifier. If a segment falls into a restricted lag channel, only a restricted label is added to the field. Then, anomaly candidate screening is performed. The screening criteria are that the upstream and downstream segment summaries within the same adjacent window have a sequential relationship and a difference relationship, and the difference relationship occurs within the lag channel's defined range. Segments meeting these criteria are registered as anomaly candidates. During candidate registration, the station index, channel identifier, and segment start and end times are recorded to form a candidate item set. Feature encoding is performed on the candidate item set. The encoding target is structured elements that can support subsequent joint inference. The encoding fields include adjacency level, direction marker, lag label, candidate density label, and diagnostic reference label. If a break segment placeholder or an item with an unclear direction is encountered during encoding, it is encoded as an undeterminable item, and a retrospective reference is retained. After completing the above processing, all encoding results are summarized to form the output field name co-originating feature. The co-originating feature serves as the direct input for the next sub-step to perform joint inference on the co-originating feature, and as a reusable intermediate product for adaptive sampling scheduling and report generation in the cross-main step. The technical effect of this step can be summarized as follows: by extracting adjacency, direction, and lag elements from the spatiotemporal relationship graph, anomaly candidates that satisfy the time segment relationship are first screened out, and these are compressed into a uniformly readable co-originating feature, facilitating subsequent inference actions on the structured input.

[0031] S330. Perform joint inference on collaborative source tracing features to generate a candidate list structure of anomaly sources; The steps take the aforementioned collaborative tracing features as input. These features include adjacency hierarchy, direction markers, lag labels, candidate density labels, and diagnostic reference labels. They can also be traced back to the time index and site index of the alignment index structure through fragment identifiers. Specifically, firstly, within the same connected path, segments are aggregated according to the direction markers to form several candidate links. Each candidate link consists of multiple candidate entries that satisfy the lag label constraints in sequence, with the nodes at both ends of the link recorded as the start and end nodes. When an undecidable entry appears in the candidate link, a breakpoint is set on the link, and the entries at both ends of the breakpoint are recorded separately for subsequent integration with resource constraint configuration. Further, joint inference is performed on each candidate link, based on the combination relationship between adjacency hierarchy, lag labels, and candidate density labels. In segments with continuous adjacency hierarchy and consistent lag labels, aggregation is performed according to the candidate density labels to generate link segments. Simultaneously, diagnostic reference labels are written into the segment annotations. When a link segment overlaps with the device event association window in the drift diagnostic structure, the segment annotations are expanded to include device event references for subsequent handling suggestions. Subsequently, cross-screening is performed on the link set, comparing link segments pointing to adjacent areas on different connected paths. If there are conflicting direction markers or mutually exclusive start and end time ranges, they are registered as conflicting segments and all references are retained. If the directions are consistent and the start and end time ranges of the segments are continuous, they are registered as merged segments. After completing all joint inferences, structured candidate entries are formed around each connected path. Each entry includes a start node, an end node, a list of affected channels, a set of link segments, a set of diagnostic references, and a restricted marker. All entries are partitioned and saved according to the site index and time index, forming the anomaly source candidate list structure for the output field name. The anomaly source candidate list structure is passed to the next main step in the cross-main step to obtain the anomaly source candidate list structure and resource constraint configuration, driving the retest priority calculation and work order arrangement. At the same time, the diagnostic reference set and restricted markers registered in this structure are transformed into textual descriptions and parameter write-back strategies in subsequent reports and calibration update structures. In summary, the technical effect of this step is as follows: by aggregating collaborative tracing features and performing joint inference at the connectivity path level, a candidate list structure of anomaly sources with start and end nodes, paragraph structure, and reference set is formed, providing a directly readable candidate entry point for subsequent retesting scheduling and report compilation.

[0032] S400 performs retest priority calculation, work order arrangement, write-back binding, retest result linking, and structured organization to generate reports and calibration update structures; Step S400 includes at least steps S410-S430: S410. Obtain the candidate list structure of anomaly sources and resource constraint configuration, perform retest priority calculation and work order arrangement processing, and obtain the sampling scheduling plan; The steps use the aforementioned anomaly source candidate list structure as the trigger source and resource constraint configuration as the scheduling boundary. The anomaly source candidate list structure includes a starting node, an ending node, an impact channel list, a link segment set, a diagnostic reference set, and a restricted flag, which can be traced back to the aligned index structure through the site index and time index. The resource constraint configuration is a set of operating parameters oriented towards the field and laboratory, including personnel scheduling, available vehicle, vessel, and drone lists, sampler, pretreatment device, and cold chain container status, reagent and consumable inventory, operating time restrictions, water and chemical safety requirements, access permits, and communication guidelines. Specifically, the link segment set and restricted flags in the anomaly source candidate list structure are read first, and candidate entries are aggregated by site index to construct the unit to be retested; then, the personnel scheduling, equipment status, and safety requirements in the resource constraint configuration are read, and executable time window and equipment binding constraints are generated. For sites that do not meet safety requirements or lack access permits, deferred entries are generated and reasons are written. Subsequently, a retest priority calculation is performed, which revolves around candidate density labels, direction marker continuity, lag labels, diagnostic reference set coverage, restricted marker strength, and resource consumption intensity. A grade result and disposal suggestion placeholder are provided according to the rule base. When grade conflicts exist at the same site, manual authorization is triggered, and the authorizing person's identifier is recorded. Further, work order orchestration is performed: within the executable time window, sites, channels, sample types, transportation routes, and temperature control requirements are packaged into site tasks. Each site task includes a task number, site index, time window, channel identifier, sample type set (blank sample, spiked sample, parallel sample, and retest sample), pretreatment requirements, cold chain temperature range, return path, and receiving unit. Simultaneously, a parameter set is generated, including sampling volume, storage method, pretreatment device parameters, spiked source batch number, label rules, and barcode scanning caliber. After a work order is generated, the batch number is registered with the Laboratory Information Management System (LIMS), and the site task is pushed to the field terminal. If insufficient vehicles or personnel, unavailable cryogenic containers, or time window conflicts are detected, a rescheduling mark and alternative time window are written to the work order, and unmet items are recorded in the resource constraint notes area. Through the above processing, an output field name sampling scheduling plan is generated. This plan is registered in the data container of this step and serves as the input for the next sub-step to extract the site task and parameter set from the sampling scheduling plan. Simultaneously, the reserved time and site interval are registered in the aforementioned alignment index structure for locating task write-back actions across main steps. The technical effect of this step can be summarized as follows: through retest priority calculation and work order orchestration, a sampling scheduling plan that can directly drive collaboration between the field and the laboratory is obtained, and a binding relationship between tasks and parameters is formed within the resource boundaries.

[0033] S420. Extract the site tasks and parameter sets from the sampling scheduling plan, perform write-back binding and link with the retest results, and generate source analysis results; The process uses the sampling scheduling plan as a single input. This plan includes task number, site index, time window, channel identifier, sample type set, pretreatment requirements, cold chain temperature range, return path, and receiving unit. The parameter set lists sampling volume, storage method, pretreatment device parameters, source batch number for labeling, labeling rules, and barcode scanning caliber. Specifically, firstly, site tasks are distributed and registered, with task numbers and barcode information sent to the field terminals. After on-site barcode scanning and chain establishment, and sample packaging, a sample number is generated. During transportation, temperature trajectories and handover nodes are recorded according to the cold chain temperature range. Upon handover to the laboratory, the receiving unit confirms and sends back a receipt record. Simultaneously, the field terminals and laboratory terminals synchronously transmit process logs under the same task number. Subsequently, write-back binding is performed: the sample number, site index, and time window are written back to the aligned index structure according to the task number, completing the record-level binding; for blank samples, spiked samples, parallel samples, and retest samples in the sample type set, the corresponding channel identifier and preprocessing requirements are bound respectively, and the source batch number and barcode information of the spike are written into the parameter set notes; when there is non-compliant scanning caliber, duplicate sample number, or temperature trajectory gap, an anomaly entry is generated and written to the pending review status. After the laboratory completes the measurement and data review, the laboratory information management system returns the measurement results and batch number; the system performs retest result binding accordingly: the measurement results are matched according to the task number and sample number, and the response sequence, timestamp, and batch number are bound to the time index, site index, and channel index of the aligned index structure, while the binding record is used to establish a reverse reference to the link segment set in the anomaly source candidate list structure; when the binding object is a blank sample, spiked sample, or parallel sample, the online quality control and drift diagnosis review entry is triggered simultaneously and the reference identifier is recorded. Furthermore, link segment comparison is performed on the attached site tasks: the start and end nodes and direction markers in the anomaly source candidate list structure are read and matched with the background segments of adjacent sites in this batch of retest samples to form segment-level comparison entries; when the segment-level comparison shows a restricted marker overlapping with equipment maintenance records or a missing sample type, the entry is marked as pending supplementary sampling or pending verification. After the above processing, all entries are summarized to generate the source analysis results of the output field name. The source analysis results consist of segment-level comparison entries, attachment logs, anomaly entries, and reference relationships, which serve as the input for the next sub-step to structure the source analysis results and are used in the cross-main step for spatiotemporal constraint construction and collaborative source update statistical view. The technical effect of this step can be summarized as follows: by writing back and binding with the retest results, a traceable bidirectional link is established between the alignment index structure and the anomaly source candidate list structure, forming source analysis results oriented towards segment-level comparison.

[0034] S430. Organize the source tracing analysis results into a structured format, and generate a report and calibration update structure; The steps take the aforementioned source tracing analysis results as input. These results include paragraph-level comparison entries, connection logs, anomaly entries, and reference relationships, and can trace back to the alignment index structure and the anomaly source candidate list structure. Specifically, firstly, structured organization is performed: paragraph-level comparison entries are aggregated by site index and time index partitioning; key nodes and start / end time ranges are extracted from the link segment set to form a path segment set; task numbers, sample numbers, and batch numbers in the connection logs are mapped to the path segment set to generate a task connectivity view; the handling and transfer status and pending sampling location are given for anomaly entries, maintaining consistency with the original entry references. Subsequently, a report is generated: the main body of the report consists of the path segment set, the task connectivity view, and a quality control reference summary, including a site-level sampling window, a channel-level comparison summary, sample completion status, and an anomaly entry list; the report appendix records key items from the connection log and parameter set for easy reading by external systems. The calibration update structure is then constructed: Paragraph-level comparison entries and quality control reference summaries are read from the source tracing analysis results. Zero-point offset trigger evidence, range mapping trigger evidence, and temperature compensation trigger evidence are extracted from the comparison entries and mapped to the applicable channels and equipment list. A configuration fragment is generated for each piece of evidence, containing the correction category, scope of application, version number, and validity period identifier. The source entry and reference identifier are recorded on the configuration fragment. When different pieces of evidence exhibit contradictory trigger directions or scope of application conflicts on the same channel, a pending fragment is generated and an authorized placeholder is attached. After the structured processing is completed, the field name report and calibration update structure are output. In the report and calibration update structure, the report is available for download and archiving by the external management system, while the calibration update structure is fed back to the aforementioned initial processing link for obtaining on-site multi-parameter raw data and sample metadata and the calibration update structure in the cross-main step stage. In the next round S110, it is called as input to the calibration update structure, thus closing the loop of sampling—quality control—source tracing—writeback—recalibration. The technical effects of this step can be summarized as follows: By structuring the results of the source tracing analysis, on the one hand, a report that can be read externally is generated, and on the other hand, a calibration update structure that can be directly consumed by the preceding links is constructed, thus completing the connection from the retest closed loop to the parameter update closed loop.

[0035] Example 2: Figure 2 A structural block diagram of a multi-source collaborative water quality result tracing and analysis system according to an embodiment of the present invention is shown. Figure 2 As shown, the structure may include: The metadata alignment and consistency calibration module 01 is used to receive raw multi-parameter data from the field, sample metadata, and calibration update structure. It performs timestamp standardization, site coding standardization, zero-point calibration, and range calibration, and outputs an alignment index structure. Specifically, it receives raw multi-parameter data from the field acquisition end and sample metadata from the sample management end, loads the calibration update structure from the previous cycle, performs timestamp standardization according to time zone, leap day, and day / night switching rules, performs site coding standardization according to the site master data dictionary, performs zero-point calibration and range calibration based on channel health markers and applicable channel lists, records the reason for discontinuation of discontinued channels and attaches the version number, and generates indexes for time, site, equipment, channel, batch, and sample number at the record level. After the formed alignment index structure is registered in this module, it provides the alignment index structure as an entry field to the quality control triggering and acceptance module. At the same time, it retains version and effective interval records in this module for subsequent write-back binding and retesting module backtracking.

[0036] The Quality Control Triggering and Acceptance Module 02, connected to the Metadata Alignment and Consistency Calibration Module, is used to read the alignment index structure and quality control plan, trigger blank samples, spiked samples, and parallel samples, complete rule acceptance and batch judgment, and output quality control acceptance results. Specifically, it reads the site index, time index, and channel index in the alignment index structure, generates triggering tasks according to the sample type definition, trigger frequency, and trigger timing of the quality control plan, generates deferred entries for channels in a deactivated or maintenance state and registers the reasons, collects measurement records and process information returned from the field and laboratory, performs blank judgment, collection judgment, and difference judgment according to the methodology rule base, aggregates by site and time to form batch boundaries, and marks missing records as pending review. After completing the judgment, it generates quality control acceptance results, provides the quality control acceptance results as input to the linkage analysis and drift diagnosis module, and registers the corresponding time range and status flag in the message queue of this module for the spatiotemporal relationship construction module to read the batch caliber.

[0037] The Linkage Analysis and Drift Diagnosis Module 03, connected to the Quality Control Triggering and Acceptance Module, is used to conduct linkage analysis on the quality control acceptance results and generate a drift diagnosis structure. Specifically, after receiving the quality control acceptance results, it expands the time series according to the site index and channel identifier, locates continuous warning or rejection segments as suspected windows, reads back the corresponding device index and version number in the alignment index structure, queries channel version switching, maintenance registration, and methodology switching records, marks windows that overlap with device events as related to device events, marks windows with abnormal concentrations of environmental records as related to environmental events, and registers windows without overlapping records as pending investigation. Diagnostic entries are generated around each suspected window. Each entry contains a drift type marker, start and end time, a list of affected channels, a reference rule identifier, and batch caliber. The entry establishes a reverse reference with the original quality control acceptance results. All entries are aggregated to form a drift diagnosis structure, which is provided as input to the spatiotemporal relationship construction module, and the correspondence between entries and rule sources is recorded in the storage area of ​​this module.

[0038] The spatiotemporal relationship construction module 04, connected to the metadata alignment and consistency calibration module and the linkage analysis and drift diagnosis module, is used to perform cross-sectional topology construction, upstream and downstream relationship calculation, and time lag estimation based on the alignment index structure and drift diagnosis structure, and output a spatiotemporal relationship graph. Specifically, it reads the site level and geographic location fields in the alignment index structure to generate cross-sectional topology, establishes nodes and path directions, generates candidate paths and marks their sources when there are contradictions between historical connectivity records and maintenance registrations, then generates an upstream and downstream relationship list in the path direction, and uses the attachment time range of each node for window limitation; referring to the window type and status label of the drift diagnosis structure, it filters node pairs with effective coverage and stable segments, generates time lag channels according to the order and interval range of upstream and downstream responses, and marks restricted channels falling into the device event association window; after integrating the topology, upstream and downstream relationship list and time lag channels, it generates a spatiotemporal relationship graph, provides the spatiotemporal relationship graph as input to the collaborative tracing and inference module, and registers the reverse reference with the drift diagnosis structure in this module.

[0039] The Collaborative Source Tracing Inference Module 05, connected to the Spatiotemporal Relationship Construction Module, is used to extract cross-sectional adjacency, flow direction constraints, and time-delay channels from the spatiotemporal relationship graph. It performs anomaly candidate screening, feature encoding, and joint inference, and outputs an anomaly source candidate list structure. Specifically, it extracts adjacency windows and direction markers from the spatiotemporal relationship graph, eliminates candidates with unclear directions and broken segments, limits the time segment by overlapping time-delay channels with candidate windows, compresses upstream segment summaries, downstream segment summaries, and background summaries into a unified field, and generates candidate entries by combining window type markers. It encodes adjacency level, direction markers, time-delay labels, candidate density labels, and diagnostic reference labels on candidate entries. Within the connected path, it aggregates entries by direction markers to form link segments. It records event references for segments that overlap with device event associated windows and registers conflict entries for segments with conflicting directions or mutually exclusive start and end ranges. The resulting anomaly source candidate list structure is delivered to the retest scheduling and work order orchestration module as the entry point for retest priority calculation. At the same time, it archives the link segment set in this module for reverse reference by the write-back binding and retest attachment modules.

[0040] The retesting scheduling and work order orchestration module 06, connected to the collaborative tracing and inference module, is used to read the anomaly source candidate list structure and resource constraint configuration, calculate the retesting priority, complete work order orchestration, and form a sampling scheduling plan. Specifically, it reads the link segment set and restricted markers in the anomaly source candidate list structure, aggregates them by site to generate retesting units, loads the personnel schedule, transportation and sampler list, pre-processing device status, consumable inventory, work period restrictions, and access permits in the resource constraint configuration, generates executable time period windows and equipment binding constraints, and generates postponement entries for sites that do not meet safety requirements or lack permits. It calculates the retesting priority based on candidate density labels, directional continuity, lag labels, diagnostic reference coverage, restriction strength, and resource consumption intensity, forms a site task and parameter set, and orchestrates task numbers, time windows, channel identifiers, sample sets, pre-processing requirements, temperature control intervals, return paths, and receiving units. The output sampling scheduling plan is passed to the write-back binding and retesting connection module, and a rescheduling marker and alternative time window are registered in this module.

[0041] The write-back binding and retesting module 07, connected to the retesting scheduling and work order orchestration module, is used to extract site tasks and parameter sets from the sampling scheduling plan, execute write-back binding and retesting result linking, generate traceability analysis results, and send back reference identifiers to the collaborative traceability inference module. Specifically, it distributes site tasks from the sampling scheduling plan, sends task numbers and barcode information to field terminals, generates sample numbers after completing barcode scanning and sample packaging, records temperature trajectories and handover nodes during transportation, receives a receipt record from the receiving unit, and field terminals and laboratory terminals send back process logs under the same task number. It writes back sample numbers, site indexes, and time windows to the aligned index structure according to the task number, binds channel identifiers and preprocessing requirements to blank samples, spiked samples, parallel samples, and retesting samples, and writes the spiked source batch number and barcode information into the parameter set annotations. Compliance, duplicate sample numbers, or temperature trajectory gaps are registered as abnormal entries and placed in a pending review status. After the laboratory completes the measurement and review, it returns the measurement record and batch number. The system links the response sequence, timestamp, and batch number to the time index, site index, and channel index of the alignment index structure according to the task number and sample number, and establishes a reverse reference with the link segment set in the abnormal source candidate list structure. For blank samples, spiked samples, and parallel samples, the online quality control and drift diagnosis review entry is triggered simultaneously and the reference identifier is recorded. For the attached task, the start node, end node, and direction mark are read, and the batch retest sample record and adjacent site background segment are matched to form a segment-level comparison entry. Entries with restricted marks overlapping with maintenance records or missing sample types are marked as pending supplementary sampling or pending review. The entries and attached logs are summarized to form the traceability analysis results, and the reference identifier is returned to the collaborative traceability inference module.

[0042] The report generation and calibration update module 08 is connected to the write-back binding and retesting module. It is used to organize the traceability analysis results in a structured manner, output the report and calibration update structure, and send the calibration update structure back to the metadata alignment and consistency calibration module. Specifically, the process involves aggregating segment-level comparison entries from the source tracing analysis results by site index and time index, extracting key nodes and start / end ranges from the link segment set, and combining them to form a path segment set. The task number, sample number, and batch number in the connection log are mapped to the path segment set to generate a comprehensive view. For abnormal entries, the work flow status and locations awaiting supplementary sampling are recorded while maintaining consistent citations. Based on this, a report body and attachments are generated. The report body includes the path segment set, comprehensive view, and quality control citation summary, while the attachments record key items from the connection log and parameter set. Simultaneously, zero-point correction, range mapping, and temperature compensation evidence are extracted from the segment-level comparison entries and quality control citation summary, mapped to the applicable channels and equipment list, and configuration fragments of the calibration update structure are generated. These configuration fragments record the correction category, applicable scope, version number, and validity period identifier, as well as the source entry and citation identifier. Fragments with conflicting trigger directions or applicable scopes are registered as pending review. The output report and calibration update structure are delivered to an external management system for archiving. The calibration update structure is then fed back to the metadata alignment and consistency calibration module as input for subsequent rounds of calibration update structures, forming a data and configuration closed loop.

Claims

1. A method for source tracing analysis of multi-source synergistic water quality results, characterized in that, include: Acquire raw data of multiple parameters and sample metadata and calibration update structure on site, perform timestamp standardization and site coding standardization, zero point calibration and range calibration processing, and generate an alignment index structure; Perform blank sample, spiked sample and parallel sample triggering processing, extract standard sample data from quality control sequence records, perform rule acceptance and batch judgment on parallel sample data and environmental records, and perform linkage analysis on quality control acceptance results to generate drift diagnostic structure; The process involves constructing a cross-sectional topology based on station indexes, geographic location fields, and cross-sectional level fields. Nodes are designated as stations, and edges represent water body connectivity paths. Upstream and downstream relationships are calculated to form a list of upstream and downstream relationships. Within the node attachment time index range, latency estimation is performed. Response peak sequences and background segments are searched on nodes with sufficient time coverage and stable channel segments to generate latency channels containing start and end times, reference nodes, affected nodes, and window boundary markers. Cross-sectional adjacency, flow direction constraints, and latency channels are extracted from the spatiotemporal relationship graph for anomaly candidate screening. Screening criteria include a sequential relationship and a difference between upstream and downstream fragment summaries within the same adjacency window, occurring within the latency channel's defined range. Feature encoding fields include adjacency level, direction markers, latency labels, candidate density labels, and diagnostic reference labels. Joint inference is performed on collaborative tracing features. Within the same connectivity path, segments are aggregated according to direction markers to form candidate links. Inference is then generated based on the combination of adjacency level, latency labels, and candidate density labels, resulting in an anomaly source candidate list structure. Perform retest priority calculation, work order arrangement, write-back binding, retest result linking, and structured organization to generate reports and calibration update structures.

2. The method according to claim 1, characterized in that, The process of constructing a cross-sectional topology based on the site index, geographic location field, and cross-sectional level field, where nodes are sites and edges are water body connectivity paths, also includes: Based on the site index, geographic location field, and cross-section level field in the aligned index structure, cross-section topology is constructed for monitoring stations within the same watershed. Nodes are stations, and edges are possible water body connection paths. The path direction is determined based on the cross-section level and topographic flow direction records. When there is a conflict between historical connection records and recent maintenance registrations, two candidate edges are retained, and the source and effective range are marked in the edge annotation.

3. The method according to claim 1, characterized in that, The process of calculating upstream and downstream relationships to form a list of upstream and downstream relationships and indexing the nodes within the time range also includes: Establish parent-child relationships between node pairs according to the direction of the connected path, forming an upstream and downstream relationship list, and attach the time index range in the alignment index structure to each node for window limitation in subsequent time delay estimation; if a node is found to lack a connected path or has a broken segment, it is recorded as a broken segment placeholder and will not participate in the generation of time delay channels in the future.

4. The method according to claim 1, characterized in that, The process of delay estimation, which involves searching for response peak sequences and background segments on node pairs with sufficient time coverage and stable channel segments to generate a delay channel containing start and end times, reference nodes, affected nodes, and window boundary markers, also includes: Lag estimation is performed on the upstream and downstream relationship list. The processing targets are node pairs with sufficient time coverage and channel stability segments. The time coverage of the node pairs comes from the time index of the alignment index structure, and the criteria for channel stability segments refer to the window type and status label in the drift diagnosis structure. Within the candidate time window, the response peak sequence and background segment of the upstream and downstream nodes are searched, the order and interval range of the responses of the two nodes are recorded, and a lag channel bound to the node pair is generated.

5. The method according to claim 1, characterized in that, The criteria for abnormal candidate screening are that the upstream and downstream fragment summaries within the same adjacent window have a sequential or dissimilar relationship, and this occurs within the time-lag channel limit. Other processes also include: For each time segment list, the three types of elements—upstream node response changes, downstream node response changes, and background changes—are compressed into fixed fields. Then, anomaly candidate screening is carried out. The screening criteria are that the upstream segment summary and the downstream segment summary within the same adjacent window have a sequential relationship and a difference relationship, and the difference relationship occurs within the time lag channel limit. Segments that meet the criteria are registered as anomaly candidates.

6. The method according to claim 1, characterized in that, The process of jointly inferring collaborative source tracing features, segmenting and aggregating candidate links according to direction labels within the same connected path to form candidate links, and inferring based on the combination of adjacency level, lag label, and candidate density label also includes: Within the same connected path, segments are aggregated according to direction labels to form several candidate links. Each candidate link consists of multiple candidate entries that satisfy the time delay label constraint in sequence. Joint inference is performed on each candidate link, based on the combination relationship between the adjacency level and the time delay label and the candidate density label.

7. The method according to claim 1, characterized in that, The process of calculating retest priority and scheduling work orders also includes: First, read the link segment set and restricted markers in the candidate list structure of the anomaly source, aggregate candidate entries by site index, and construct the unit to be retested; then read the personnel scheduling, equipment status and safety requirements in the resource constraint configuration, generate executable time window and equipment binding constraints; perform retest priority calculation, the calculation is based on candidate density label, direction marker continuity, lag label, diagnostic reference set coverage, restricted marker strength and resource consumption intensity.

8. The method according to claim 7, characterized in that, The process of work order arrangement also includes: Within the executable time window, the site, channel, sample type, transportation route, and temperature control requirements are packaged into a site task. The site task includes task number, site index, time window, channel identifier, sample type set, pretreatment requirements, cold chain temperature range, return path, and receiving unit. At the same time, a parameter set is generated, which includes sampling volume, storage method, pretreatment device parameters, spiking source batch number, labeling rules, and barcode scanning caliber.

9. The method according to claim 1, characterized in that, The process of generating reports and updating calibration structures also includes: The calibration update structure is constructed as follows: Read the paragraph-level reference entries and quality control reference summaries from the traceability analysis results, extract the zero-point offset trigger evidence, range mapping trigger evidence and temperature compensation trigger evidence from the reference entries, and map them to the applicable channels and equipment list; generate a configuration fragment for each piece of evidence, which includes the correction category, scope of application, version number and validity period identifier.

10. A multi-source collaborative water quality result tracing and analysis system, applied to the method of any one of claims 1-9, characterized in that, include: The metadata alignment and consistency calibration module is used to receive raw data of multiple parameters on site, sample metadata, and calibration update structure, perform timestamp standardization, site coding standardization, zero-point calibration, range calibration, and output alignment index structure. The quality control triggering and acceptance module is connected to the metadata alignment and consistency calibration module. It is used to read the alignment index structure and quality control plan, trigger blank samples, spiked samples, and parallel samples, complete rule acceptance and batch judgment, and output the quality control acceptance results. The linkage analysis and drift diagnosis module is connected to the quality control triggering and acceptance module. It is used to perform linkage analysis on the quality control acceptance results and generate a drift diagnosis structure. The spatiotemporal relationship construction module is connected to the metadata alignment and consistency calibration module and the linkage analysis and drift diagnosis module. It is used to perform cross-sectional topology construction, upstream and downstream relationship calculation, and lag estimation based on the alignment index structure and drift diagnosis structure, and output the spatiotemporal relationship graph. The collaborative source tracing and inference module is connected to the spatiotemporal relationship construction module. It is used to extract cross-sectional adjacency, flow direction constraints, and time-delay channels from the spatiotemporal relationship graph, perform anomaly candidate screening, feature encoding, joint inference, and output an anomaly source candidate list structure. The retest scheduling and work order orchestration module is connected to the collaborative tracing and inference module. It is used to read the candidate list structure of anomaly sources and resource constraint configuration, calculate the retest priority and complete the work order orchestration to form a sampling scheduling plan. The write-back binding and retesting module is connected to the retesting scheduling and work order orchestration module. It is used to extract the site task and parameter set from the sampling scheduling plan, perform write-back binding and retesting result hooking, generate source tracing analysis results, and send the reference identifier back to the collaborative source tracing inference module. The report generation and calibration update module, connected to the write-back binding and retesting module, is used to organize the traceability analysis results in a structured manner, output the report and calibration update structure, and send the calibration update structure back to the metadata alignment and consistency calibration module.

Images