Compound transaction performance state migration method and system based on quality file analysis

Abstract

The application discloses a compound transaction performance state migration method and system based on quality file analysis, generates a project performance object according to a supplier response version confirmed by a purchaser, order data, project start data or performance start instructions; receives and analyzes a quality file, extracts a structure conformity measured index and determines a file type, generates a quality file association key and binds the quality file association key with the project performance object; compares the threshold value, judges whether the file type covers a demand file type set, and generates a quality conformity judgment result; triggers a preset performance state machine to perform state migration based on the judgment result, migrates to a quality file preliminary screening pass state, a purchaser confirmation state or a quality control review state when the judgment result is correct, and migrates to a file supplement state, a quality abnormality processing state or a deviation processing state when the judgment result is incorrect. The measured value extracted by file analysis is compared with the threshold value to drive state migration, and the problem of response lag and non-uniform standards caused by reliance on manual experience judgment is solved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for the migration of the fulfillment status of compound transactions based on quality document parsing. Background Technology

[0002] In chemical research procurement scenarios such as non-standard compounds, customized compounds, impurity reference standards, special standards, high-value intermediates, and project-based services, the transaction process does not end after the purchaser selects a supplier and generates an order. The supplier still needs to complete a series of fulfillment steps, including the preparation, testing, quality document generation, packaging, transportation, delivery, and receipt confirmation of the target compound. Unlike ordinary commodity transactions, the fulfillment management of compound transactions highly relies on the review and confirmation of quality documents. Common quality documents include Certificates of Analysis (COA), Material Safety Data Sheets (MSDS), Nuclear Magnetic Resonance (NMR) spectra, Mass Spectrometry (MS) spectra, High Performance Liquid Chromatography (HPLC) spectra, Liquid Chromatography-Mass Spectrometry (LCMS) spectra, third-party testing reports, batch records, packaging labels, and temperature control records. These documents contain key chemical information for determining whether the supplier's deliverables meet the agreed quality commitments. Therefore, the ability to process quality documents directly determines the reliability and efficiency of fulfillment management.

[0003] Currently, typical e-commerce systems generally only manage general transaction statuses such as order payment, logistics tracking, and receipt confirmation, lacking the capability for content-level processing of chemical quality documents. While general enterprise resource planning (ERP) systems or quality management systems can manage documents and processes, they lack mechanisms for automatically resolving, associating, and transitioning statuses between target compounds, batches, testing documents, spectral characteristics, purity commitments, hazardous properties, temperature control data, and fulfillment anomalies in compound transactions. In actual business processes, these quality documents typically exist in the system as email attachments, transmitted via instant messaging tools, or manually uploaded, leading to a series of technical problems.

[0004] First, there is a lack of automated, strong binding mechanism between quality documents and target compounds, orders, batches, suppliers, and their response versions. Document attribution relies on manual annotation, which is prone to errors or omissions, leading to breaks in the fulfillment traceability chain. Second, document content is not automatically parsed by the system. Key chemical quality data such as purity values, batch numbers, and testing dates in COAs, peak area percentages and retention times in HPLC, and spectral characteristics in NMR and MS cannot be automatically extracted and utilized, still requiring manual review and comparison. Third, document integrity checks and quality compliance judgments rely entirely on human experience, and status transitions are also triggered manually. The system cannot automatically advance the fulfillment process based on document parsing results and preset chemical quality thresholds, resulting in delayed and inconsistent responses to chemical quality issues such as "non-compliance with purity standards," "missing spectra," and "inconsistent batch numbers." Furthermore, when anomalies occur, such as missing documents, purity deviations from promised values, or exceeding transport temperature control limits, the existing system cannot automatically generate structured anomaly event records and correlate them with relevant batches, documents, and suppliers in multiple dimensions. Anomaly management relies on fragmented manual communication. Meanwhile, the quality data, anomaly records, and confirmation results accumulated during the historical performance process are difficult to form a structured data loop, which cannot effectively support subsequent repurchase decisions, supplier performance evaluation, and compliance audit traceability.

[0005] In summary, existing technologies suffer from technical problems such as missing quality document parsing during compound transaction fulfillment, inability to automatically compare chemical quality thresholds, inability to automatically migrate fulfillment status based on data-driven approaches, and inability to structurally identify and record chemical quality anomalies. There is an urgent need for a technical solution that can achieve in-depth parsing of quality documents, automatic chemical quality determination, and automatic migration of fulfillment status. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for transitioning the performance status of compound transactions based on quality document parsing. By automatically comparing the measured structural conformity indicators extracted from the parsed quality document content with the chemical quality threshold associated with the project performance object, a quality document association key is generated to bind the quality document to the project performance object. Based on the comparison result, a state machine is triggered to transition the performance status of the project performance object, which improves the automation level of the compound transaction performance process and solves the problems in the prior art where the performance status cannot be automatically transitioned by data, the reliance on manual experience judgment leads to delayed response, and the lack of unified standards.

[0007] To address the aforementioned technical problems, a first aspect of this invention provides a method for transitioning the fulfillment status of compound transactions based on quality document parsing, comprising the following steps: Step S1: Obtain the supplier's response version confirmed by the purchaser, order data, project initiation data or performance initiation instructions, and generate a project performance object based on the obtained data; wherein, the project performance object is associated with the target compound identifier, the chemical quality threshold extracted from the obtained data and the set of requirement document types; Step S2: Receive the quality document corresponding to the project performance object, parse the content of the quality document, extract the structural conformity test indicators, and determine the file type of the quality document; generate a quality document association key based on at least one of the following: file hash, target compound identifier, batch number, test type, supplier identifier, order identifier, response version identifier, and key quality features, and bind the quality document association key to the project performance object; Step S3: Compare the measured structural compliance index with the chemical quality threshold associated with the project performance object, and determine whether the file type of the received quality document covers the set of required file types, and generate a quality compliance judgment result. Step S4: Based on the quality compliance judgment result, trigger the preset performance state machine to perform state transition for the project performance object; wherein, when the quality compliance judgment result is compliant, the project performance object is transitioned to the quality document initial screening passed state, the state awaiting confirmation by the purchaser, or the state awaiting quality control review; when the quality compliance judgment result is non-compliant, the project performance object is transitioned to the document supplementation state, the quality anomaly handling state, or the deviation handling state.

[0008] Further, step S2 involves parsing the quality document to extract the structural compliance measurement indicators, including: Step S21: When the file type of the quality file is a spectrum file, perform feature peak extraction on the spectrum file to obtain a set of feature peak parameters; Step S22: Construct chemical feature descriptors based on the set of characteristic peak parameters to obtain chemical feature descriptors for the test sample; Step S23: Obtain the target chemical feature descriptor corresponding to the target compound identifier associated with the project performance object; perform graph structure matching between the test sample chemical feature descriptor and the target chemical feature descriptor; use the structure matching score as the measured index of structure conformity; and combine it with at least one of the measured purity value, main peak area percentage, molecular weight peak matching result, and impurity peak parameters to constitute the measured chemical quality information.

[0009] Further, the construction of chemical characteristic descriptors based on the set of characteristic peak parameters in step S22 to obtain chemical characteristic descriptors for the test sample includes: Step S221: Extract coupling constant information and / or nuclear Overhaus effect spectrum feature information from the set of characteristic peak parameters; Step S222: Based on the coupling constant information and / or the nuclear Overhouse effect spectrum characteristic information, determine the relative configurational properties of at least one chiral center in the test sample; Step S223: Encode the relative configuration attribute into a stereochemical feature vector, and fuse the stereochemical feature vector with the two-dimensional connectivity descriptor of the test sample to generate a three-dimensional chemical feature descriptor of the test sample, which serves as the chemical feature descriptor of the test sample.

[0010] Further, step S23, which involves performing graph structure matching between the chemical characteristic descriptor of the test sample and the target chemical characteristic descriptor, and using the structure matching score as the measured index of structure conformity, includes: Step S231: Perform the first round of graph structure matching to obtain the initial matching degree score and the correspondence between feature peaks and structural segments; Step S232: Based on the correspondence between the characteristic peak and the structural fragment, the set of characteristic peak parameters extracted in step S21 is subjected to attribution verification to identify at least one characteristic peak belonging to a specific structural fragment in the target chemical characteristic descriptor and at least one residual characteristic peak not belonging to the target chemical characteristic descriptor. Step S233: Re-extract the characteristic peak parameters for the residual characteristic peaks, and correct the chemical characteristic descriptor of the test sample according to the attribution verification result to obtain the corrected chemical characteristic descriptor of the test sample; Step S234: Perform a second round of graph structure matching between the modified chemical feature descriptor of the test sample and the target chemical feature descriptor, and use the structure matching score obtained from the second round of matching as the measured index of structure conformity.

[0011] Furthermore, after identifying at least one characteristic peak belonging to a specific structural fragment in the target chemical feature descriptor in step S232, the method further includes: Step S232a: Obtain the signal-to-noise ratio and peak shape symmetry parameter for each residual characteristic peak; Step S232b: When the signal-to-noise ratio value of any residual feature peak is lower than a preset signal-to-noise ratio threshold, or the peak shape symmetry parameter exceeds a preset symmetry tolerance range, the residual feature peak is marked as a resolution artifact. Step S232c: Remove the residual feature peaks marked as parsing artifacts from the residual feature peaks, and input only the verified residual feature peaks into step S233.

[0012] Furthermore, after step S4 describes migrating the project performance object to a quality anomaly handling state or deviation handling state when the quality compliance judgment result is non-compliant, it also includes: Step S41: Extract the chemical quality deviation characteristic parameters that cause non-compliance judgment from the quality compliance judgment results; Step S42: Match the chemical quality deviation characteristic parameters with the abnormal patterns in the preset preparation process abnormality knowledge base to obtain at least one candidate abnormal process step and its matching confidence level. Step S43: Associate the candidate abnormal process step and its corresponding matching confidence level with the project performance object that has been migrated to the quality abnormality handling state or deviation handling state, and generate an abnormality tracing record.

[0013] Further, the extraction of chemical quality deviation characteristic parameters that lead to non-compliance in step S41 includes: Step S411: Obtain at least one deviation index item in the quality compliance judgment result that leads to non-compliance judgment, and identify whether the deviation index item contains the superimposed contribution of multiple independent deviation sources. Step S412: When the deviation index is identified as a superimposed contribution containing multiple independent deviation sources, chemical semantic decoupling is performed on the deviation index item to decompose the composite deviation into at least two independent deviation components, wherein each of the independent deviation components corresponds to a chemical structural fragment or an impurity source type. Step S413: Extract the characteristic parameters of each independent deviation component to obtain a set of independent deviation characteristic parameters, which are used as the chemical quality deviation characteristic parameters.

[0014] Further, step S412, which involves chemically and semantically decoupling the deviation index term and decomposing the composite deviation into at least two independent deviation components, includes: Step S4121: Obtain the original spectral signal data corresponding to the deviation index item, input the original spectral signal data into the independent component analysis model, and obtain a set of independent signal components by maximizing the statistical independence of each output signal component; Step S4122: Perform feature peak identification on each of the independent signal components and extract the peak position, peak area and peak shape parameters corresponding to each of the independent signal components; Step S4123: Based on the extracted peak position, peak area and peak shape parameters, each of the independent signal components is mapped to an independent deviation component.

[0015] Furthermore, after decomposing the composite deviation into at least two independent deviation components as described in step S412, the method further includes: Step S412a: Obtain the characteristic peak parameters corresponding to each independent deviation component, and match the characteristic peak parameters with a preset impurity standard spectral library, wherein the impurity standard spectral library contains standard spectral characteristic parameters of known impurity compounds; Step S412b: When the characteristic peak parameter of any of the independent deviation components successfully matches a known impurity compound in the impurity standard spectral library, the chemical identity of the known impurity compound is obtained, and the chemical identity is assigned to the independent deviation component. Step S412c: Based on whether each independent deviation component has been assigned a chemical identity, each independent deviation component is marked, and the marking result is passed to step S413 so that the feature parameters of the independent deviation components that have been assigned a chemical identity are extracted first in step S413.

[0016] Further, step S42, which involves matching the chemical quality deviation characteristic parameters with anomaly patterns in a preset preparation process anomaly knowledge base to obtain at least one candidate anomalous process step and its matching confidence level, includes: Step S421: When there is no abnormal pattern record of the supplier for the target compound identifier in the preset preparation process abnormal knowledge base, obtain the chemical structure descriptor of the target compound identifier; Step S422: Calculate the structural similarity between the chemical structure descriptor and the compound structure descriptors associated with the existing abnormal pattern records in the preset preparation process abnormal knowledge base; Step S423: Select at least one similar compound whose structural similarity exceeds a preset similarity threshold, and extract the supplier's historical process anomaly pattern records on the similar compound; Step S424: Match the chemical quality deviation feature parameters with the extracted historical process anomaly pattern records to obtain the candidate abnormal process steps and their matching confidence levels.

[0017] Furthermore, before triggering the preset performance state machine to perform state transition for the project performance object in step S4, the following steps are also included: Step S401: After generating the project performance object, obtain the chemical structure attributes corresponding to the target compound identifier; Step S402: Select at least one specific verification state node from a preset state node library based on the chemical structure properties. Step S403: Inject the selected specific verification state node into the preset basic performance state machine topology to generate the performance state machine corresponding to the project performance object.

[0018] Further, step S402, which involves selecting at least one specific verification state node from a preset state node library based on the chemical structure properties, includes: Step S4021: Input the chemical structure attributes into a pre-trained chemical structure-risk association model to obtain the predicted risk impurity spectrum corresponding to the target compound, wherein the chemical structure-risk association model is trained based on the correspondence between the known compound structure and its typical impurities or degradation products; Step S4022: Based on the detection requirements of each impurity or degradation product in the predicted risk impurity spectrum, select the corresponding risk verification state node from the preset state node library. Step S4023: The risk verification state node is injected into the basic performance state machine topology of step S403 as the specific verification state node.

[0019] Further, step S403, which involves injecting the selected specific verification state node into a preset basic performance state machine topology to generate the performance state machine corresponding to the project performance object, includes: Step S4031: Obtain the timing constraints and / or resource constraints of at least two of the specific verification status nodes selected in step S402. Step S4032: Detect whether there are timing conflicts or resource conflicts between each of the specific verification state nodes and between the specific verification state nodes and existing nodes in the basic performance state machine topology. Step S4033: When a conflict is detected, a preset conflict resolution strategy library is invoked to merge, split, or rearrange the specific verification state nodes according to the conflict type, so as to obtain the resolved state node set. Step S4034: Inject the parsed set of state nodes into the basic performance state machine topology to generate the performance state machine.

[0020] Furthermore, the sources of the quality documents received in step S2 include: suppliers, third-party testing institutions, platform quality control systems, logistics monitoring systems, purchaser review systems, or experimental instrument systems.

[0021] Furthermore, the generation of quality compliance judgment results in step S3 also includes: When the judgment result is non-compliant, the non-compliance type is determined to be missing documents or substandard quality indicators; Step S4, which describes migrating the project performance object to a document supplementation state, a quality anomaly handling state, or a deviation handling state when the quality compliance judgment result is non-compliant, includes: When the non-compliance type is "file missing", the project fulfillment object will be migrated to the "file needs to be supplemented" state. When the non-compliance type is that the quality indicator fails to meet the standard, the project performance object will be moved to the quality anomaly handling status or deviation handling status.

[0022] Furthermore, in step S4, after migrating the project performance object to the initial screening status of quality documents upon receiving a satisfactory quality compliance judgment, the process further includes: Generate a review task and send it to the purchasing party or quality control role; Upon receiving confirmation instructions from the purchasing party or quality control role regarding the review task, the project performance object is moved from the initial screening status of the quality documents to the pending shipment status or the quality confirmation completion status.

[0023] Furthermore, after triggering the preset performance state machine to perform state transitions on the project performance object in step S4, the process further includes: Generate and store state transition audit records, which include file upload information, parsing results, threshold comparison results, file overwrite judgment results, state identifiers before and after state transition, trigger rule version, trigger time, and operation subject identifier.

[0024] Accordingly, a second aspect of the present invention provides a compound transaction performance status transition system based on quality document parsing, which processes project performance data based on the above-mentioned compound transaction performance status transition method based on quality document parsing, including: The performance object generation module is used to obtain the supplier response version confirmed by the purchaser, order data, project start data or performance start instructions, and generate project performance objects based on the obtained data; wherein, the project performance object is associated with the target compound identifier, the chemical quality threshold extracted from the obtained data and the set of requirement document types; The quality document parsing module is used to receive the quality document corresponding to the project performance object, parse the content of the quality document, extract the structural conformity test indicators, and determine the document type of the quality document; generate a quality document association key based on at least one of the following: document hash, target compound identifier, batch number, test type, supplier identifier, order identifier, response version identifier, and key quality features, and bind the quality document association key to the project performance object; The quality compliance judgment module is used to compare the measured structural compliance indicators with the chemical quality thresholds associated with the project performance object, and to determine whether the file type of the received quality documents covers the set of required file types, and to generate a quality compliance judgment result. The state machine transition execution module is used to trigger a preset performance state machine to perform state transitions on the project performance object based on the quality compliance judgment result. Specifically, when the quality compliance judgment result is compliant, the project performance object is transitioned to the quality document initial screening pass state, the state awaiting purchaser confirmation state, or the state awaiting quality control review state. When the quality compliance judgment result is non-compliant, the project performance object is transitioned to the document supplementation state, the quality anomaly handling state, or the deviation handling state.

[0025] Accordingly, a third aspect of the present invention provides an electronic device, including: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the above-described compound transaction fulfillment status transition method based on quality document parsing.

[0026] Accordingly, a fourth aspect of the present invention provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the above-described compound transaction fulfillment status transition method based on quality document parsing.

[0027] Accordingly, a fifth aspect of the present invention provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the above-described compound transaction fulfillment status transition method based on quality document parsing.

[0028] The above-described technical solutions of the embodiments of the present invention have the following beneficial technical effects: 1. By elevating the parsing of spectral quality files from traditional numerical extraction to graph structure matching at the chemical structure level, and further integrating stereochemical features such as coupling constants and nuclear Overhaus effect spectroscopy to construct three-dimensional descriptors, the consistency between the test sample and the target compound can be automatically verified from two dimensions: molecular skeleton connectivity and chiral configuration. This effectively overcomes the limitations of conventional methods that rely solely on purity values ​​or single peak area ratios for judgment, which may miss deep-seated quality defects such as isomer errors and unsatisfactory enantiomer excess rates. It significantly enhances the accuracy and chemical specificity of quality compliance determination in compound transactions, providing a more solid decision-making basis for the automatic migration of subsequent performance status. 2. When the quality compliance is determined to be non-compliant and the non-compliance type is a failure to meet quality indicators, the process no longer stops at simply marking the abnormal state. Instead, it decouples the deviation indicators chemically and semantically, uses independent component analysis to perform blind source separation on the original spectral signal, and decomposes the originally mixed and superimposed composite deviations into independent deviation components corresponding to specific chemical structural fragments or impurity source types. These components are then automatically matched with impurity standard spectral libraries to assign chemical identities. This achieves an automated leap from detecting an anomaly to diagnosing the cause of the anomaly, solving the technical problem of difficulty in clarifying deviation responsibility in multi-factor scenarios. This allows the anomaly tracing results to be directly linked to specific process steps or impurity sources, providing precise structured guidance for the subdivision and targeted handling of quality anomaly status. 3. In the process of tracing the source of manufacturing process anomalies, when the pre-set manufacturing process anomaly knowledge base lacks historical anomaly patterns of the supplier for the current target compound, the chemical structural similarity between the target compound and existing compounds in the knowledge base is calculated. The supplier's historical process anomaly patterns on similar compounds are then matched across compounds to obtain candidate anomaly process steps and their confidence levels. This breaks through the limitation of the knowledge base relying on the historical accumulation of the same compound, enabling new compounds, rare compounds, or compounds traded for the first time to obtain reasonable anomaly tracing inferences based on structural similarity. This significantly improves the coverage of process anomaly diagnosis and the usability in cold start scenarios, further improving the data-driven fulfillment status migration closed loop. Attached Figure Description

[0029] Figure 1 This is a flowchart of a compound transaction fulfillment status transition method based on quality document parsing provided in an embodiment of the present invention; Figure 2 This is a block diagram of a compound transaction fulfillment status migration system module provided in an embodiment of the present invention for quality document parsing.

[0030] Figure label: 1. Performance object generation module; 2. Quality document parsing module; 3. Quality compliance judgment module; 4. State machine transition execution module. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0032] In chemical research procurement scenarios involving non-standard compounds, customized compounds, impurity reference standards, and high-value intermediates, the transaction process does not end after the purchaser places an order. Instead, it extends into a complete fulfillment chain encompassing preparation, testing, delivery of quality documents, packaging and transportation, and receipt confirmation. In this chain, the supplier must submit various quality documents, such as analytical certificates, high-performance liquid chromatography (HPLC) chromatograms, nuclear magnetic resonance (NMR) spectra, mass spectra, and material safety data sheets, to demonstrate that the delivered compounds meet pre-committed quality indicators in terms of purity, structure, and batch consistency. However, the chemical information contained in these quality documents has long existed in an unstructured form. Interpretation, threshold comparison, and integrity verification all rely on manual review and experience-based judgment, resulting in slow progress in fulfillment, inconsistent judgment standards, and difficulty in timely identification of quality anomalies. To transform this manual process into a standardized workflow that can be automatically executed by computer programs, a technical solution is proposed that can parse the content of quality documents, extract chemical measured values, and automatically compare them with preset thresholds, thereby driving the automatic migration of the fulfillment status based on the comparison results.

[0033] Please refer to Figure 1 The first aspect of this invention provides a method for transitioning the fulfillment status of compound transactions based on quality document parsing, comprising the following steps: Step S1: Obtain the supplier's response version confirmed by the purchaser, order data, project initiation data or performance initiation instructions, and generate a project performance object based on the obtained data; wherein, the project performance object is associated with the target compound identifier, the chemical quality threshold extracted from the obtained data and the set of requirement document types.

[0034] In chemical research procurement scenarios involving non-standard compounds, customized compounds, or high-value intermediates, project execution initiation is not limited to a single confirmation path after multi-supplier comparison. It can also originate from various methods, such as automatic triggering of standard orders, direct access to historical repurchase data, structured import of offline projects, and proactive issuance of execution initiation instructions. The supplier response versions, order data, project initiation data, or execution initiation instructions confirmed by the procuring party not only contain commercial information such as transaction prices and delivery dates, but also explicitly or implicitly record chemically specific information such as purity commitments for the target compound, a list of quality documents to be delivered, and packaging and transportation requirements. Upon obtaining confirmation data from any of the above sources, a generic order is not simply generated. Instead, the structured commitments contained within are converted into a project execution object specifically for execution management. This object acts as a data container, uniformly associating the target compound identifier, chemical quality thresholds extracted from the acquired data, and a set of requirement document types. The target compound identifier can be a CAS number, a SMILES string, or a system-internal standardized code, used to uniquely identify the chemical identity of the compound to be delivered. Chemical quality thresholds are quantified quality indicators extracted from supplier commitments or historical performance records, such as purity not less than 98% and peak area percentage not less than 95%. The set of required document types clarifies the categories of quality documents the supplier must provide in this transaction, such as analytical certificates, high-performance liquid chromatography (HPLC) chromatograms, nuclear magnetic resonance (NMR) spectra, mass spectra, and material safety data sheets. Through this conversion process, the chemical commitments originally present in supplier response versions, order records, project initiation documents, or performance instructions are solidified into parameterized data that can be automatically read and compared by subsequent procedures. This establishes a complete data anchor point containing chemical quality benchmarks at the outset of performance, providing a unified reference for the automated processing and status determination of subsequent quality documents.

[0035] Step S2: Receive the quality documents provided by the supplier that correspond to the project performance object, parse the content of the quality documents, extract the measured indicators of structural conformity, and determine the document type of the quality documents.

[0036] After the project performance object is generated, the corresponding quality documents are submitted by the supplier or relevant systems through standardized data interfaces. The receipt of quality documents does not rely on manual classification; instead, computer programs retrieve electronic files via file transfer protocols, application programming interface calls, or message queues. Upon receipt, a content parsing process is performed. For different formats of quality documents, parsing methods include regular expression matching and keyword location for text files, row and column traversal and cell extraction for tabular files, text layer extraction and character recognition for portable document format files, and reading information such as creation time, author, and version from file metadata. Through these parsing operations, fields such as batch number, measured purity, and detection date can be extracted from analysis certificates; data such as main peak retention time and peak area percentage can be extracted from chromatographic reports; and parameters such as molecular weight peak signal intensity can be obtained from mass spectrometry reports. These actual detection values ​​extracted from the original document content are uniformly packaged into structural conformity test indicators. At the same time, the parsing process will automatically determine which type of detection the file belongs to based on the file content characteristics or file header information, such as whether it is an analysis certificate, high performance liquid chromatography, or nuclear magnetic resonance spectrum, so as to perform file type coverage comparison in subsequent steps.

[0037] Based on content parsing, the generation and binding of quality document association keys are performed simultaneously. Quality document association keys are a technical means of uniquely binding quality documents to project fulfillment objects, target compounds, batches, orders, suppliers, and response versions. The information used for their generation includes at least one of the following: document hash, target compound identifier, batch number, test type, supplier identifier, order identifier, response version identifier, and key quality features extracted from the document. The document hash is obtained by performing a hash operation on the quality document content and is used to uniquely identify the version and integrity of the document. The target compound identifier, batch number, test type, and supplier identifier are obtained from the parsing results and the association data of the fulfillment object, respectively, and are used to establish the correspondence between the document and the compound identity, production batch, test method, and responsible entity. Key quality features are core parameters with discriminative significance extracted from the document content, such as purity value, percentage of main peak area, or molecular weight peak position. Once the association key is generated, it is immediately bound to the current project performance object, so that the quality compliance judgment results, state transition events and audit records generated in subsequent steps can be accurately traced back to the corresponding quality documents and their source context through the association key, thus forming a complete data traceability link with the association key as the hub in the entire performance process.

[0038] Optionally, the sources of the quality documents received in step S2 have expanded from a single supplier to multiple participating nodes in the compound transaction fulfillment chain. Specifically, these may include the supplier itself, third-party testing institutions, platform quality control systems, logistics monitoring systems, purchaser review systems, or laboratory instrument systems. Suppliers provide core quality documents such as analytical certificates and spectral reports generated during their preparation and self-inspection processes; third-party testing institutions issue commissioned testing reports independent of the supplier, providing a neutral reference for quality assessment; platform quality control systems can automatically generate pre-inspection results for document integrity or historical fulfillment data comparison reports; logistics monitoring systems upload environmental monitoring data such as temperature curves and humidity records during cold chain transportation; purchaser review systems can provide feedback on the purchaser's internal re-inspection spectra or supplementary testing data for the test samples; and laboratory instrument systems directly output raw detection signals and instrument-acquired data through electronic lab logbooks or laboratory information management system interfaces. By unifying the above-mentioned multi-source quality documents and data into the receiving and parsing process, quality compliance judgment no longer relies solely on materials submitted by the supplier, but can instead form a more comprehensive judgment basis by integrating information from multiple sources.

[0039] Step S3: Compare the measured structural compliance indicators with the chemical quality thresholds associated with the project performance objects, and determine whether the file types of the received quality documents cover the set of required file types, and generate a quality compliance judgment result.

[0040] After obtaining the structural compliance measured indicators and document type information, the quality compliance judgment stage begins. This judgment process involves two dimensions of verification. The first dimension is numerical threshold comparison, where the structural compliance measured indicators extracted in step S2 are compared one by one with the chemical quality thresholds extracted from the supplier's response version and associated with the project performance object in step S1. For example, the measured purity value in the analysis certificate is compared with the supplier's promised 98% threshold, and the percentage of the main peak area extracted from the chromatogram is compared with the preset minimum area ratio requirement. The second dimension is document integrity verification, which determines whether the set of uploaded quality document types identified in step S2 completely covers the set of required document types recorded in the project performance object; that is, it checks whether any agreed-upon quality document is missing. When both dimensions meet the conditions—that is, the measured value reaches or exceeds the threshold and all required document types have been uploaded and identified—a compliant quality compliance judgment result is generated. If any measured value fails to reach the corresponding threshold, or if there are missing items in the required document type set, a non-compliant quality compliance judgment result is generated. This stage transforms raw file data into structured judgment conclusions, converting the original process of relying on manual reading of files and making judgments based on experience into computer-executable numerical and set operations, ensuring the consistency of judgment criteria and execution speed.

[0041] Step S4: Based on the quality compliance judgment result, trigger the preset performance state machine to perform state transitions for the project performance object. Specifically, when the quality compliance judgment result is compliant, the project performance object is transitioned to the status of initial quality document screening passed, pending confirmation from the purchaser, or pending quality control review. When the quality compliance judgment result is non-compliant, the project performance object is transitioned to the status of document supplementation required, quality anomaly handling, or deviation handling.

[0042] After the quality compliance assessment result is generated, it serves as a trigger signal input to the preset performance state machine. The performance state machine defines the various states that the project performance object may be in during the performance cycle, as well as the trigger conditions required for transitions between states. Specifically, the state machine includes at least several key nodes such as the initial screening pass status for quality documents, the pending confirmation status for the purchaser, the pending quality control review status, the document supplementation status, the quality anomaly handling status, and the deviation handling status. When the quality compliance assessment result is compliant, the state machine transitions the project performance object from its current state to the initial screening pass status for quality documents, the pending confirmation status for the purchaser, or the pending quality control review status, indicating that the batch of chemicals has passed the system's initial screening and will enter different review and confirmation stages according to business configuration. When the quality compliance assessment result is non-compliant, the state machine transitions the project performance object to the document supplementation status, the quality anomaly handling status, or the deviation handling status, indicating that there are incomplete quality documents or non-compliance with quality indicators in the current performance process, and will enter different processing paths such as supplementary submission, anomaly investigation, or deviation assessment depending on the specific type of non-compliance. The entire state transition process is automatically driven by the judgment results, requiring no manual intervention. The state machine itself can be pre-configured according to different transaction types, compound characteristics, or compliance requirements, but its core mechanism is to execute deterministic state switching actions based on the logical values ​​of input signals. Through this data-driven event response approach, the pace of the fulfillment process is tightly coupled with the parsing results of quality documents, avoiding delays caused by manual judgment and errors in information transmission, thus enabling batch-level fulfillment management to achieve real-time performance and traceability.

[0043] In step S4, when the quality compliance judgment result is compliant, the project performance object is moved to the following states: initial screening of quality documents passed, pending confirmation from the purchaser, or pending quality control review. None of these three states represent final quality acceptance; rather, they indicate that the automated review of the uploaded quality documents has been completed and a preliminary approval conclusion has been reached. Depending on the specific business scenario and risk control configuration, different review levels are entered. The initial screening of quality documents passed indicates that the document integrity verification, chemical quality threshold comparison, and document type coverage judgment performed by the program have all passed, but manual review has not yet been conducted. The pending confirmation from the purchaser indicates that after initial screening, the review task has been pushed to the technical personnel or purchase manager designated by the purchaser, awaiting their manual verification of the analysis results against the original documents and issuance of confirmation instructions. The pending quality control review state is applicable when the purchaser has a dedicated quality control department or has commissioned a third-party quality auditing agency. The review task is pushed to the quality control role, who independently reviews the spectral data, purity indicators, and batch consistency of the test samples based on internal quality standards. The three states described above are intermediate rather than final states in the quality verification stage. Their common characteristic is that the results of automatic analysis and comparison require confirmation by personnel with chemical expertise before proceeding to the subsequent shipping or receiving stages. This approach clarifies the boundary between automatic analysis and manual confirmation, avoiding equating preliminary procedural judgments directly with final quality conclusions, thus preserving necessary professional review processes while ensuring automation efficiency.

[0044] When the quality compliance assessment result is non-compliance, the project performance object is moved to three states: Document Supplementation Required, Quality Anomaly Handling, or Deviation Handling. These three states correspond to different anomaly types and handling strategies. The Document Supplementation Required state applies when the quality document types do not cover the required document types. In this case, the supplier needs to upload the specified quality documents according to the missing document list. Once the documents are complete, the parsing and comparison process can be retried; this is a repairable process-related anomaly. The Quality Anomaly Handling state applies when the measured chemical quality indicators do not reach the preset chemical quality threshold, indicating that the chemical quality of the current batch of products has been determined to be non-compliant with the committed standards. This requires in-depth handling processes such as deviation investigation, supplier rectification, or re-inspection application. The Deviation Handling state applies when the quality indicators do not meet the standards but the deviation is within an acceptable range, or when the supplier has submitted an explanation of the deviation and requested concessionary acceptance. In this case, the deviation assessment and risk release decision process begins. By refining non-compliance situations into three different handling states and configuring corresponding subsequent migration paths for each, the state machine can perform differentiated process control for different types of performance anomalies, avoiding the delays and unclear responsibility caused by uniformly handling all non-compliance situations.

[0045] Through the coordinated operation of the above four steps, the chemical commitment parameters in the supplier's response version are stored as associated attributes of the project performance object. The system automatically parses quality documents to extract chemical measurement data, performs dual checks on numerical thresholds and document coverage to generate a compliance judgment, and triggers the performance state machine to complete automatic state transitions based on the judgment results. This achieves end-to-end automation from chemical commitment extraction, quality data parsing, standard comparison and judgment to performance process advancement. This solution overcomes the response lag and inconsistency in standards caused by existing technologies that rely on manual review of quality documents, experience-based quality judgment, and manual operation for state transitions. It endows compound transaction performance management with data-driven, rule-unified, and traceable technical characteristics.

[0046] Specifically, in step S2, the quality document is parsed to extract the measured structural conformity indicators, including: Step S21: When the quality file is a spectrum file, perform feature peak extraction on the spectrum file to obtain a set of feature peak parameters.

[0047] After determining the file type of the quality file, for those identified as spectral files, such as high-performance liquid chromatography (HPLC), nuclear magnetic resonance (NMR), mass spectrometry (MS), or liquid chromatography-mass spectrometry (LC-MS) spectrometry (LC-MS), further characteristic peak extraction is required. Spectral files differ from text-based quality files in that their core quality information is not presented directly as text fields, but rather as peak curves formed by changes in signal intensity with retention time, chemical shift, or mass-to-charge ratio. The characteristic peak extraction process first requires baseline correction of the spectrum to eliminate baseline shifts caused by instrument drift and background noise. Then, a peak-finding algorithm is used to locate the position of each peak in the spectral data point sequence where the response signal exceeds a preset detection threshold. The retention time, chemical shift value, mass-to-charge ratio, peak height, peak area integral value, and peak width parameters of each peak are recorded. For high-performance liquid chromatography (HPLC), the extracted characteristic peak parameter set includes the retention time and peak area percentage of each chromatographic peak; for nuclear magnetic resonance (NMR) spectra, the characteristic peak parameter set includes the chemical shift values ​​and coupling splitting information of each peak; for mass spectra, the characteristic peak parameter set includes the signal intensity and isotopic distribution of each mass-to-charge ratio peak. These parameters are summarized into a characteristic peak parameter set, which serves as the raw data input for subsequent chemical characterization descriptor construction. Since the number, position, and intensity distribution of peaks in spectral files directly reflect the chemical composition and structural characteristics of the sample, the accuracy of characteristic peak extraction is crucial for subsequent steps.

[0048] Step S22: Construct chemical characteristic descriptors based on the characteristic peak parameter set to obtain the chemical characteristic descriptors of the test sample.

[0049] The set of characteristic peak parameters is still a set of scattered numerical information and cannot be directly compared with the chemical structure of the target compound. It needs to be further transformed into a chemical descriptor that can characterize the molecular structure. A chemical descriptor is a mathematical or graph-theoretic expression of molecular structure information, which can include multi-dimensional features such as atomic connections, functional group composition, ring system structure, and chiral information. For the set of characteristic peak parameters extracted from spectral files, the process of constructing a chemical descriptor includes mapping the chemical shift or retention time information of each characteristic peak to possible chemical structural fragments, converting the coupling splitting mode into the connection relationship between adjacent atoms, and converting the integral area ratio into the ratio of hydrogen or carbon atoms in different chemical environments. For example, in 1H NMR spectroscopy, a set of signal peaks located in a specific chemical shift range and exhibiting triple splitting can be mapped to a methyl structural fragment adjacent to the methylene group; in HPLC, the main peak in a specific retention time range corresponds to the standard retention time of the target compound, and its area percentage reflects purity information. Through the summarization and recombination of this kind of chemical information, a chemical descriptor that can represent the chemical structure characteristics of the test sample is finally generated. The descriptor is organized in the form of a graph structure, where nodes represent atomic or functional group segments and edges represent chemical bonds or spatial proximity relationships, thereby converting the spectral signal into a structured expression that can be directly compared with the structure of the target compound.

[0050] Step S23: Obtain the target chemical feature descriptor corresponding to the target compound identifier associated with the project performance object, perform graph structure matching between the chemical feature descriptor of the test sample and the target chemical feature descriptor, use the structure matching score as the measured index of structure conformity, and combine it with at least one of the measured purity value, main peak area percentage, molecular weight peak matching result, and impurity peak parameters to constitute the measured chemical quality information.

[0051] After generating the chemical descriptor of the test sample, it needs to be compared with the standard chemical descriptor of the target compound to determine whether the test sample is consistent with the target compound at the chemical structure level. The target chemical descriptor is a reference descriptor pre-constructed and stored based on the chemical structural formula or standard spectrum of the target compound, and it corresponds one-to-one with the target compound identifier associated with the project implementation object. The graph structure matching process takes the test sample descriptor and the target descriptor as two molecular graph inputs, and calculates the degree of consistency between the two at the node and edge levels using a subgraph isomorphism algorithm or a maximum common subgraph algorithm. During the matching process, the algorithm traverses each node and edge in the test sample descriptor, determines whether there is a corresponding element in the target descriptor, and accumulates the number of successfully matched node pairs and edge pairs. The final output structure matching score is a normalized value. For example, when all characteristic peaks of the test sample and the target compound can be correctly assigned and the connection relationship is completely consistent, the matching score can be close to or equal to 100%; when there are additional peaks that cannot be assigned or peaks corresponding to key structural fragments are missing, the matching score will decrease accordingly. This score is used as a structural conformity test indicator, rather than as the entirety of the chemical quality test values. This structural conformity test indicator, along with at least one of the following: purity test values ​​extracted from the analysis certificate text, main peak area percentage extracted from the chromatographic report, molecular weight peak matching results extracted from the mass spectrometry report, and impurity peak parameters, constitutes the chemical quality test information. This information set serves as the complete data foundation for quality conformity judgment in step S3. By incorporating structural conformity into the constituent dimensions of chemical quality test information as a structural conformity test indicator, quality judgment is no longer limited to simple comparisons of discrete values. Instead, it verifies the correctness of the chemical identity of the test sample at the molecular skeleton level. Together with multi-dimensional test data such as purity, chromatography, and mass spectrometry, it forms a comprehensive characterization of the test sample's quality, thus providing a more comprehensive chemical basis for subsequent state transition decisions.

[0052] Further, in step S22, a chemical characteristic descriptor is constructed based on the characteristic peak parameter set to obtain the chemical characteristic descriptor of the test sample, including: Step S221: Extract coupling constant information and / or nuclear Overhaus effect spectrum characteristic information from the characteristic peak parameter set.

[0053] After obtaining the set of characteristic peak parameters in step S21, coupling constant information and nuclear Overhaus effect spectral characteristic information can be further extracted from the characteristic peak data derived from the NMR spectrum. The coupling constant is a measure of the spin coupling strength between adjacent hydrogen atoms in a proton NMR spectrum, usually expressed in Hertz (Hz), and its value has a quantitative relationship with the dihedral angle between the coupled atoms. When two hydrogen atoms are located on adjacent carbon atoms in an anti-coplanar conformation, their coupling constant is usually large, reaching twelve to eighteen Hz; while in an adjacent cross-conformation, the coupling constant decreases significantly, typically in the range of two to six Hz. The nuclear Overhaus effect spectroscopy reflects the distance between atomic nuclei in three-dimensional space by detecting the magnetization transfer signal between spatially close pairs of hydrogen atoms, even if there is no direct chemical bond between them. In proton NMR data, the coupling constant is obtained by measuring the frequency difference between split peaks, while the nuclear Overhaus effect spectral characteristic information is determined by analyzing the presence and integral intensity of cross-peaks. Unlike chemical shift values, which only reflect the characteristics of the functional group's chemical environment, these information directly encode the spatial arrangement of atoms, providing a crucial data source for inferring the stereo configuration of chiral centers.

[0054] Step S222: Based on coupling constant information and / or nuclear Overhouse effect spectrum characteristic information, determine the relative configurational properties of at least one chiral center in the test sample.

[0055] Based on the coupling constant information or nuclear Overhaus effect spectral characteristics extracted in step S221, the relative configurational properties of the chiral center in the test sample can be further determined. A chiral center is a carbon atom in a molecule that connects four different substituents. Its two possible spatial arrangements constitute mirror-image enantiomers. While their chemical properties are almost identical in achiral environments, their physiological activities and pharmacological properties may differ significantly. Therefore, controlling the enantiomer excess rate is a crucial aspect of quality verification in compound trading. When inferring the relative configuration using coupling constant information, the empirical relationship between the coupling constant described by the Karplacian equation and the dihedral angle is utilized. The measured coupling constant values ​​of adjacent hydrogen atoms are mapped to the range of dihedral angles between the carbon-hydrogen bonds containing the two hydrogen atoms, thereby inferring the relative spatial orientation of each substituent on the chiral center. When inferring relative configuration using nuclear Overhausian effect spectral information, the principle that spatially proximal hydrogen atom pairs will exhibit cross-peak signals in the spectrum is used to determine which hydrogen atoms are located on the same side of the molecule. For example, in a six-membered ring compound, if a nuclear Overhausian effect-related signal is observed between two hydrogen atoms located on the same side of the ring, then these two hydrogen atoms can be determined to be in a cis configuration. By performing the above configuration inference on multiple chiral centers in the molecule one by one, the relative configurational properties of the sample at each chiral position can be obtained.

[0056] Step S223: Encode the relative configuration attributes into stereochemical feature vectors, and fuse the stereochemical feature vectors with the two-dimensional connectivity descriptors of the test sample to generate a three-dimensional chemical feature descriptor of the test sample, which serves as the chemical feature descriptor of the test sample.

[0057] After obtaining the relative configurational attributes of the chiral centers, they need to be encoded into computer-processable stereochemical feature vectors and fused with the constructed two-dimensional connectivity descriptors of the test sample to generate a three-dimensional chemical feature descriptor for the test sample. The two-dimensional connectivity descriptor represents the chemical bond connections between atoms in a molecule using graph theory; its nodes represent atom types, and edges represent bond types. It can completely describe the planar structural skeleton of the molecule, but it cannot distinguish stereoisomers that differ only in spatial arrangement. The encoding method for the stereochemical feature vectors can be to assign a directional label to each chiral center, for example, using positive and negative signs to represent two enantiomeric configurations, or using cis and trans labels to mark the relative spatial relationships of substituents in a cyclic structure. The specific method for fusing the stereochemical feature vectors with the two-dimensional connectivity descriptors can be to add a stereochemical attribute field to each chiral center node in the graph structure, or to add a stereochemical feature dimension to the entire molecular graph descriptor. The resulting 3D chemical descriptor fully describes the molecular skeleton connections and spatial configuration of key chiral positions of the sample. It can perform graph structure matching, including the stereochemical dimension, with the target chemical descriptor, which also carries stereochemical information. This allows for more accurate structural consistency assessment of compounds containing chiral centers during the matching process in step S23. This extension enables quality compliance assessment not only to detect compounds with incorrect molecular skeleton connections but also to identify deeper quality issues that cannot be detected by 2D structure matching alone, such as insufficient enantiomer excess or incorrect isomer types.

[0058] Further, in step S23, graph structure matching is performed between the chemical characteristic descriptor of the test sample and the target chemical characteristic descriptor, and the structure matching score is used as the measured index of structure conformity, including: Step S231: Perform the first round of graph structure matching to obtain the initial matching score and the correspondence between feature peaks and structural segments.

[0059] During the initial graph structure matching, the chemical descriptor of the test sample constructed in step S22 and the target chemical descriptor corresponding to the target compound identifier are used as input. A first round of comparison is performed using a subgraph isomorphism or maximum common subgraph matching algorithm, outputting two results: an initial matching score and the correspondence between characteristic peaks and structural fragments. The initial matching score reflects the proportion of nodes and edges in the test sample descriptor that can find matching terms in the target descriptor. However, due to the potential influence of inaccurate characteristic peak attribution or the presence of interfering peaks in the spectrum during the first round of matching, this score cannot be used as the final judgment criterion. More importantly, the characteristic peak-structural fragment correspondence generated in this round of matching records which structural fragment in the target descriptor each characteristic peak node in the test sample descriptor matches. For example, a set of peaks with a chemical shift of 7.26 ppm is matched to aromatic hydrogen atoms on the benzene ring, and a chromatographic peak with a retention time of 5.3 minutes is matched to the main product peak of the target compound. Establishing this correspondence provides a peak-by-peak, traceable chain of matching evidence for subsequent attribution verification.

[0060] Step S232: Based on the correspondence between characteristic peaks and structural fragments, the set of characteristic peak parameters extracted in step S21 is subjected to attribution verification to identify at least one characteristic peak belonging to a specific structural fragment in the target chemical characteristic descriptor and at least one residual characteristic peak not belonging to the target chemical characteristic descriptor.

[0061] Based on the peak-structural fragment correspondence obtained in the first round of matching, the attribution verification is performed on the original peak parameter set extracted in step S21. The purpose of attribution verification is to examine whether each peak can indeed be reasonably attributed to a known structural fragment in the target chemical descriptor. For peaks that can be clearly assigned—that is, whose peak position, peak area ratio, and coupling splitting mode match the expected values ​​of a specific structural fragment in the target descriptor—they are marked as assigned peaks. For peaks that cannot establish a reasonable correspondence with any structural fragment in the target descriptor, such as extra chromatographic peaks appearing at unexpected retention time positions in HPLC and not corresponding to any known impurities, or unknown peaks appearing in chemical shift intervals where no signal should exist in NMR spectra, they are marked as residual peaks. These residual peaks may originate from actual impurities or degradation products in the sample, or from instrument noise, baseline fluctuations, or analytical artifacts generated during integral spectral analysis. The result of attribution verification divides the original peak parameter set into two parts: an assigned peak subset and a residual peak subset, providing clear processing targets for subsequent correction operations.

[0062] Step S233: Re-extract the characteristic peak parameters for the residual characteristic peaks, and correct the chemical characteristic descriptor of the test sample according to the attribution verification results to obtain the corrected chemical characteristic descriptor of the test sample.

[0063] After identifying residual characteristic peaks, they are not simply added directly to the final judgment as impurity signals. Instead, the residual characteristic peaks undergo a new peak parameter extraction process to obtain more accurate peak parameter information. This re-extraction process may include local baseline correction of the spectral region containing the residual characteristic peak, deconvolution processing of overlapping peaks, and fine delineation of peak boundaries. These operations are more detailed and targeted than the global rapid extraction in step S21, helping to eliminate integration errors introduced by improper global parameter settings. After obtaining the accurate parameters of the residual characteristic peaks, the chemical characteristic descriptor of the test sample is corrected based on the attribution verification results: characteristic peaks confirmed to belong to the target structural fragment are retained and confirmed in the descriptor; residual characteristic peaks confirmed to exist after re-extraction are added to the descriptor as new impurity characteristic nodes; and residual characteristic peaks whose peak area is found to be below the detection limit or whose signal-to-noise ratio is insufficient to support a reliable judgment are removed from the descriptor. Through the above correction operations, a corrected chemical characteristic descriptor of the test sample that more accurately reflects the actual chemical composition of the test sample is obtained. This descriptor eliminates some of the interference from analytical noise while retaining the information of the actual unknown components.

[0064] Step S234: Perform a second round of graph structure matching between the corrected chemical characteristic descriptor of the test sample and the target chemical characteristic descriptor, and use the structure matching score obtained from the second round of matching as the measured index of structure conformity.

[0065] The corrected chemical characteristic descriptor of the test sample generated in step S233 is used for a second round of graph structure matching with the target chemical characteristic descriptor. Since the corrected descriptor has undergone assignment confirmation and impurity node addition / removal, the composition of its node and edge sets has changed compared to the first round of matching. Therefore, the results of the second round of matching can more accurately reflect the true structural differences between the test sample and the target compound. In the second round of matching, the assigned structural fragment nodes can be accurately located in the target descriptor, while newly added impurity characteristic nodes cannot find matching terms in the target descriptor. This difference is quantified in the matching score. Finally, the structure matching score obtained from the second round of matching is output as the measured index of structural conformity. This score eliminates the bias caused by inaccurate peak assignment in the first round of matching and can more reliably reflect the degree of consistency between the test sample and the target compound at the chemical structure level, providing corrected and verified structure matching data for the quality conformity judgment in step S3.

[0066] Furthermore, after identifying at least one characteristic peak belonging to a specific structural fragment in the target chemical characterization descriptor in step S232, the method further includes: Step S232a: Obtain the signal-to-noise ratio and peak shape symmetry parameters for each residual characteristic peak.

[0067] After identifying the characteristic peaks belonging to the target chemical characteristic descriptor and the unassigned residual characteristic peaks in step S232, not all residual characteristic peaks are directly sent to step S233 for subsequent correction processing. Instead, the reliability of each residual characteristic peak is first quantitatively evaluated. The evaluation dimensions include two key parameters: signal-to-noise ratio (SNR) and peak shape symmetry parameter. The SNR is the ratio between the signal intensity of a characteristic peak and the noise level of the spectral baseline, used to measure whether the peak has a sufficient signal response to distinguish it from random instrument background fluctuations. In chromatographic analysis, a SNR of three times is usually used as a reference standard for the detection limit; signals below this value may simply be random fluctuations caused by noise rather than a true chemical component response. The peak shape symmetry parameter measures the degree to which the shape of the characteristic peak deviates from an ideal Gaussian distribution, and is usually expressed as an asymmetry factor or tailing factor in chromatographic analysis. A normally eluted chromatographic peak should exhibit a basically symmetrical Gaussian curve shape. However, when the column is overloaded, the column bed collapses, or there is strong interference from retained components in the sample, the peak shape may exhibit asymmetry phenomena such as forward extension or tailing. These distorted peaks are prone to area calculation errors during integration, and treating them as true impurity signals may lead to misjudgment. By calculating these two parameters for each residual characteristic peak individually, objective quantitative evidence can be provided for subsequent artifact identification.

[0068] Step S232b: When the signal-to-noise ratio of any residual characteristic peak is lower than the preset signal-to-noise ratio threshold, or the peak shape symmetry parameter exceeds the preset symmetry tolerance range, the residual characteristic peak is marked as a resolution artifact.

[0069] After obtaining the signal-to-noise ratio (SNR) and peak shape symmetry parameters, these two parameters are compared with preset judgment thresholds to distinguish between genuine impurity signals and analytical artifacts introduced by the instrument or algorithm. The preset SNR threshold is the minimum acceptable signal level set according to the sensitivity requirements of the detection method and the performance characteristics of the instrument. For example, the SNR threshold can be set to 3:1 or 10:1. Residual characteristic peaks below this threshold are considered insufficient in signal intensity to reliably confirm the presence of genuine chemical components. The preset symmetry tolerance range defines the upper and lower limits of acceptable peak shape distortion. Taking the asymmetry factor as an example, the range of 0.8 to 1.5 is usually considered the acceptable symmetry interval. Peaks exceeding this range indicate that their peak shape has seriously deviated from the theoretical shape of normal elution peaks, and the reliability of their integral area value is questionable. When the SNR value of a residual characteristic peak is lower than the preset SNR threshold, or its peak shape symmetry parameter exceeds the preset symmetry tolerance range, the residual characteristic peak is marked as an analytical artifact. Analysis artifacts are not actual chemical components present in the sample, but rather spurious signals generated during spectrum acquisition, data processing, or integration algorithm execution due to electronic noise, baseline drift, improper integration parameters, or algorithm limitations.

[0070] Step S232c: Remove the residual feature peaks marked as parsing artifacts from the residual feature peaks, and input only the verified residual feature peaks into step S233.

[0071] After artifact labeling and judgment of all residual feature peaks, a removal operation is performed to remove residual feature peaks marked as analytical artifacts from the set of residual feature peaks to be processed. Only residual feature peaks whose signal-to-noise ratio and peak shape symmetry have been verified are retained and passed to step S233 for subsequent re-extraction and descriptor correction. The significance of this filtering operation is to avoid incorrectly including instrument noise or algorithmic artifacts as real impurity signals in the correction process of the chemical feature descriptor of the test sample, thereby preventing the erroneous addition of non-existent impurity feature nodes to the descriptor in step S233. If analytical artifacts are not removed and enter the subsequent process, the corrected chemical feature descriptor of the test sample may contain false structural difference information, which may lead to an inappropriate reduction of the matching score in the second round of structure matching in step S234, ultimately causing step S3 to misjudge the originally qualified test sample as non-compliant. Through the above artifact identification and filtering mechanism, only real residual signals verified by signal reliability are used for descriptor correction, so that subsequent structure matching and state transition decisions are based on more accurate chemical data.

[0072] Specifically, in step S4, after migrating the project performance object to the quality anomaly handling state or deviation handling state when the quality compliance judgment result is non-compliance, the following steps are also included: Step S41: Extract the chemical quality deviation characteristic parameters that cause non-compliance judgment from the quality compliance judgment results.

[0073] When step S4 results in a non-compliance judgment, and the non-compliance type is a failure to meet quality indicators, thus migrating the project performance object to a quality anomaly handling state or deviation handling state, the anomaly tracing mechanism is activated. This requires extracting the chemical quality deviation characteristic parameters that led to the non-compliance judgment from the quality compliance judgment result generated in step S3. The quality compliance judgment result not only contains a conclusive label of compliance or non-compliance but also associates with specific deviation information that triggered the non-compliance judgment. This only applies to cases where quality indicators fail to meet standards and does not include cases where documents are missing. The extraction of deviation characteristic parameters varies depending on the type of quality indicator deviation in the non-compliance judgment: when non-compliance is triggered due to insufficient purity, the extracted parameters include the promised purity threshold, the measured purity value, and the deviation magnitude between the two; when non-compliance is triggered due to the percentage of main peak area being lower than the threshold, the extracted parameters include the measured main peak area ratio, the threshold requirement, and the difference; when non-compliance is triggered due to inconsistent batch numbers, the extracted parameters include the difference information between the batch number recorded in the quality documents and the batch number associated with the project performance object. For non-compliance due to missing documents, the project fulfillment entity is in a document supplementation status, and the extraction of chemical quality deviation feature parameters in this step is not performed, nor is it included in the subsequent process anomaly knowledge base matching process. The aforementioned chemical quality deviation feature parameters constitute a quantitative description of the current quality anomaly situation, providing input feature vectors for subsequent process anomaly pattern matching, enabling source tracing analysis to be conducted based on specific numerical deviations, rather than relying solely on fuzzy anomaly classification labels.

[0074] Step S42: Match the chemical quality deviation characteristic parameters with the abnormal patterns in the preset preparation process abnormality knowledge base to obtain at least one candidate abnormal process step and its matching confidence level.

[0075] After extracting the chemical quality deviation characteristic parameters, these parameters are matched with the anomaly patterns in a pre-defined preparation process anomaly knowledge base to identify candidate abnormal process steps that may lead to the current quality deviation and their matching confidence levels. The preparation process anomaly knowledge base is a pre-constructed set of relational data that stores the correspondence between the types of anomalies that may occur in each step of the preparation process and the typical quality deviations caused by those anomalies. For example, the knowledge base may record: insufficient recrystallization times in the purification process usually lead to low product purity and residual peaks of specific impurities; deviations in reaction time control may lead to an increased proportion of by-reaction products and the appearance of additional impurity peaks with characteristic retention times in the chromatogram; insufficient drying may lead to excessive moisture content and the appearance of water peak signals in the product's NMR spectrum. During the matching process, the deviation characteristic parameters extracted in step S41, such as the deviation amplitude value, the retention time range or chemical shift value of specific impurity peaks, and the missing file type, are compared item by item with the condition description fields recorded in each anomaly pattern in the knowledge base. The degree of matching between each candidate abnormal process step and the current deviation characteristic is calculated and output in the form of matching confidence. For example, when a characteristic retention time peak corresponding to a known impurity is detected in the chromatogram and the purity deviation is within a specific range, a candidate step in the recrystallization process abnormality may be matched with high confidence; when only a purity deviation is detected but no identifiable impurity peak is found, multiple candidate steps may be matched with moderate confidence. Through this matching process, the quality deviation is transformed from a phenomenological description into a traceable diagnostic result associated with the preparation process.

[0076] Step S43: Associate the candidate abnormal process step and its corresponding matching confidence level with the project performance object that has been migrated to the quality abnormality handling state or deviation handling state, and generate an abnormality tracing record.

[0077] After obtaining at least one candidate abnormal process step and its corresponding matching confidence level, a structured data association is established between this traceability information and the project performance object that has been moved to the quality anomaly handling or deviation handling state, and an anomaly traceability record is generated accordingly. Specifically, the association is established by writing the identifier of the candidate abnormal process step and the matching confidence level value into the anomaly information field of the project performance object's data structure, while simultaneously associating the current timestamp and the unique identifier of the quality compliance judgment result to ensure the traceability of the traceability information. As a component of the performance file, the anomaly traceability record contains the following key information: the unique identifier of the abnormal event, the deviation type that triggered the anomaly, the extracted chemical quality deviation characteristic parameters, the list of matched candidate abnormal process steps and their respective confidence levels, the associated supplier identifier and target compound identifier, and the time information of record generation. This record is directly bound to the project performance object, enabling a complete retrospective of the entire decision-making chain from deviation detection and process inference to responsibility determination during subsequent supplier performance evaluation, quality dispute handling, or compliance audits. By persistently linking the traceability results with the fulfilling entities in the quality anomaly handling or deviation handling process, the originally scattered quality anomaly events are integrated into structured process diagnostic data assets, providing a queryable and comparable historical reference for subsequent procurement decisions and supplier management of similar compounds.

[0078] Further, in step S41, the chemical quality deviation characteristic parameters that lead to non-compliance are extracted, including: Step S411: Obtain at least one deviation index item in the quality compliance judgment result that leads to non-compliance judgment, and identify whether the deviation index item contains the superimposed contribution of multiple independent deviation sources.

[0079] When extracting chemical quality deviation characteristic parameters in step S41, the deviation index items leading to the non-compliance judgment are first obtained from the quality compliance judgment results. Deviation index items are specific quantitative indicators that trigger non-compliance judgments, such as the percentage deviation between the measured purity value and the threshold, the difference in the percentage of the main peak area, and the excess amount of a specific impurity peak's peak area. In practical scenarios of compound preparation and detection, the quality problem reflected by a single deviation index item may not be caused by a single factor, but rather by the combined contribution of multiple independent deviation sources. For example, a low total purity observed in the chromatogram may simultaneously include the contribution of the main product peak area decreasing due to incomplete reaction conversion, and the contribution of one or more impurity peaks remaining due to insufficient purification; similarly, redundant signal peaks appearing in a specific chemical shift range in the NMR spectrum may include both residual signals from unreacted raw materials and characteristic signals from reaction byproducts. Step S411 identifies whether there are signs of such multi-source superposition through structured analysis of the deviation index items. The identification method can be to check whether there are multiple distinguishable independent signal peaks in the spectral region associated with the deviation index, or to determine whether the value corresponding to the deviation index is significantly different from the typical deviation amplitude of a single known abnormal mode, so as to preliminarily determine whether the deviation index is a single-source deviation or a multi-source composite deviation.

[0080] Step S412: When the deviation index is identified as a superimposed contribution containing multiple independent deviation sources, chemical semantic decoupling is performed on the deviation index term, and the composite deviation is decomposed into at least two independent deviation components, wherein each independent deviation component corresponds to a chemical structural fragment or an impurity source type.

[0081] When step S411 identifies that the deviation index item contains the superimposed contribution of multiple independent deviation sources, a chemical semantic decoupling operation needs to be performed on the deviation index item to decompose the originally mixed composite deviation into at least two independent deviation components. The core idea of ​​chemical semantic decoupling is that although the composite deviation appears as a whole deviation quantity at the numerical level, at the chemical level, this deviation quantity is formed by the superposition or joint contribution of independent signals generated by different chemical structural fragments or different impurity sources on the spectrum. The decoupling process is to separate these signals from the superposition state. Taking the total purity deviation in a high-performance liquid chromatography (HPLC) chromatogram as an example, if the purity deviation is caused by two different impurity peaks, the decoupling operation needs to allocate the total purity deviation quantity according to the peak area ratio of each impurity peak, so that each impurity peak corresponds to an independent deviation component, and each deviation component is characterized by the relative retention time and peak area percentage of the impurity. Taking redundant signals in NMR spectra as an example, if multiple additional peaks from different sources exist within a certain chemical shift range, the decoupling operation needs to assign them to different independent deviation components based on the chemical shift value and coupling splitting mode of each peak. Each component corresponds to a possible impurity structural fragment. Through this chemical semantic decoupling, the originally general "purity not up to standard" or "spectral anomaly" is refined into multiple independent deviation components with clear chemical meanings. Each component can be independently traced to a chemical structural fragment or a type of impurity source, providing a finer input granularity for subsequent process anomaly matching.

[0082] Step S413: Extract the characteristic parameters of each independent deviation component to obtain a set of independent deviation characteristic parameters, which are used as chemical quality deviation characteristic parameters.

[0083] After the composite deviation is decomposed into at least two independent deviation components, the feature parameters of each independent deviation component need to be extracted separately. This results in a set of independent deviation feature parameters, which replace the original mixed-state deviation feature parameters and serve as input for process anomaly knowledge base matching in step S42. The feature parameter extraction for each independent deviation component is closely related to the deviation type and chemical source corresponding to that component. For an independent deviation component corresponding to a specific impurity peak, the extracted feature parameters include the relative retention time, peak area percentage, UV absorption wavelength characteristics, and peak shape characteristics in the spectrum. For an independent deviation component corresponding to a specific chemical shift redundant signal, the extracted feature parameters include the chemical shift value, coupling splitting mode, integral area ratio, and signal multiplicity. For an independent deviation component corresponding to a missing file type, the extracted feature parameters include the missing file type, the detection method number corresponding to that file type, and the degree of impact of the missing file on the integrity of the quality verification. Each set of independent deviation feature parameters carries the chemical source identifier of its respective deviation component. This allows for precise one-to-one comparison of specific impurity chemical characteristics with known anomalies in specific process steps within the knowledge base during subsequent matching, rather than fuzzy matching of mixed deviations with multiple anomaly patterns. This feature extraction method, moving from composite to independent features, significantly improves the precision and accuracy of anomaly tracing analysis.

[0084] Further, in step S412, the deviation index term is chemically and semantically decoupled, and the composite deviation is decomposed into at least two independent deviation components, including: Step S4121: Obtain the original spectral signal data corresponding to the deviation index item, input the original spectral signal data into the independent component analysis model, and obtain a set of independent signal components by maximizing the statistical independence of each output signal component.

[0085] After step S411 identifies that the deviation index contains the superimposed contributions of multiple independent deviation sources, the mixed deviation signals need to be separated mathematically. To this end, the raw spectral signal data corresponding to the deviation index is acquired. This raw spectral signal data may be a chromatographic data sequence containing multiple overlapping chromatographic peaks, or it may be a segment of NMR spectrum data where multiple signals overlap within a certain chemical shift interval. In the raw spectra, deviation signals from different sources superimpose, presenting as complex peaks or signal envelopes that are difficult to distinguish directly. Conventional integration or deconvolution methods are often insufficient to accurately separate the components. Therefore, an independent component analysis model is used to process the raw spectral signal data.

[0086] The model's input data consists of a two-dimensional observation signal matrix formed by expanding the original spectral signal according to the detection channel or sampling time. Each row of the matrix corresponds to a chromatographic response sequence at a detection wavelength or a signal intensity vector at a sampling time point, while each column corresponds to the response value of a sampling point or a detection channel at different times or under different conditions. Logically, the independent component analysis model first performs mean zeroing and whitening preprocessing on the observation signal matrix. Principal component analysis is then used to remove second-order correlations between data and normalize the variance. Subsequently, based on the negative entropy maximization criterion, a fixed-point iterative algorithm is used to solve the unmixing matrix. In each iteration, the non-Gaussianity measure of each row of the unmixing matrix is ​​calculated. The statistical independence of the current separation results is evaluated using an approximate function of negative entropy, and the row vectors of the unmixing matrix are updated in the gradient descent direction until the mutual information between the output signal components converges to below a preset threshold. The model ultimately outputs a set of statistically independent signal components, each represented as a data sequence of the same length as the original spectral signal, corresponding to the contributions from different chemical species in the chromatographic or spectroscopic signal. These independent signal components have been mathematically separated, each representing a relatively independent signal source in the spectrum, providing a pure signal basis free from mutual interference for subsequent chemical feature identification.

[0087] Step S4122: Identify the characteristic peaks of each independent signal component and extract the peak position, peak area and peak shape parameters corresponding to each independent signal component.

[0088] After obtaining a set of independent signal components, characteristic peak identification is performed on each independent signal component. Since independent component analysis has separated the superimposed signals in the original spectrum, the peak signals contained in each independent signal component are usually relatively pure, with complete peak shapes and clear baselines, which is conducive to more accurate extraction of characteristic peak parameters. The characteristic peak identification process includes local baseline correction for each independent signal component, determining the start and end boundaries of the peak, calculating the peak area through numerical integration, and recording the positional parameters corresponding to the peak apex. For independent signal components originating from chromatographic signals, the extracted peak position is the relative retention time, the peak area is the integrated response value of the independent component on the detector, and the peak shape parameters include the half-width at half-maximum (HWHM) and the asymmetry factor. For independent signal components originating from nuclear magnetic resonance (NMR) signals, the extracted peak position is the chemical shift value, the peak area is the relative integrated intensity of the signal, and the peak shape parameters include the HWHM and the splitting mode characteristics. Through this process, each independent signal source that was originally difficult to distinguish in the mixed spectrum is transformed into a set of specific and quantified peak parameter descriptions, enabling subsequent steps to chemically assign and quantify the bias of each independent component based on these precise values.

[0089] Step S4123: Based on the extracted peak position, peak area and peak shape parameters, each independent signal component is mapped to an independent deviation component.

[0090] After extracting the peak position, peak area, and peak shape parameters of each independent signal component, these physical signal parameters need to be mapped into independent deviation components with chemical meaning. The mapping process is based on the chemical information reflected by the peak parameters: peak position parameters, such as relative retention time or chemical shift values, are used to preliminarily infer the type of chemical species corresponding to the independent signal. For example, by comparing the standard retention time or standard chemical shift table of known impurities, it can be determined whether the signal belongs to an impurity with a certain structural feature or an unreacted raw material; peak area parameters are used to quantify the contribution of the independent deviation component. For example, the percentage of the integral area of ​​the peak to the total peak area represents the proportion of the deviation source in the total deviation; peak shape parameters help verify whether the signal is a real chemical component signal rather than a baseline perturbation or an instrument spurious peak. After mapping, each independent deviation component carries its corresponding chemical source type identifier and quantitative deviation contribution value. Multiple independent deviation components together constitute the complete decoupling result of the deviation index item and are output as a set of structured independent deviation feature parameters for use in step S413 and subsequent step S42 for process anomaly pattern matching. Through a chemical semantic decoupling process aided by independent component analysis, complex deviations that were originally mixed and superimposed and whose source of responsibility was difficult to distinguish are decomposed into multiple independently traceable chemical deviation units, providing a technical foundation for the refinement and automation of process anomaly tracing.

[0091] Furthermore, after decomposing the composite deviation into at least two independent deviation components in step S412, the method further includes: Step S412a: Obtain the characteristic peak parameters corresponding to each independent deviation component, and match the characteristic peak parameters with a preset impurity standard spectral library, wherein the impurity standard spectral library contains standard spectral characteristic parameters of known impurity compounds.

[0092] After successfully decomposing the composite bias into at least two independent bias components in step S412, each independent bias component currently only carries physical signal features such as peak position, peak area, and peak shape parameters extracted from its corresponding independent signal component, lacking clear chemical identification information. Step S412a assigns chemical identification criteria to these bias components by matching the characteristic peak parameters corresponding to each independent bias component with a pre-defined impurity standard spectral library. The impurity standard spectral library is a pre-constructed reference database that includes standard spectral feature parameters of various known impurity compounds. These known impurities include process impurities that may be generated by the target compound in common synthetic routes, degradation products that may be formed during storage, and by-reaction products that may be generated under specific reaction conditions. Each record in the spectral library contains at least the chemical identification of the known impurity compound and its standard spectral feature parameters under specific detection conditions, such as the standard relative retention time and UV absorption wavelength ratio in high-performance liquid chromatography, the standard mass-to-charge ratio and isotope distribution pattern in mass spectrometry, and the standard chemical shift value and coupling splitting characteristics in nuclear magnetic resonance spectroscopy. The matching process calculates the similarity between the characteristic peak parameters of the independent deviation component to be identified and the standard parameters of each known impurity in the spectral library. It judges the degree of agreement between the two in terms of peak position, peak area ratio, and peak shape characteristics, and uses this as the basis for determining whether the independent deviation component corresponds to the known impurity.

[0093] Step S412b: When the characteristic peak parameter of any independent deviation component successfully matches a known impurity compound in the impurity standard spectral library, the chemical identity of the known impurity compound is obtained and the chemical identity is assigned to the independent deviation component.

[0094] When spectral library matching calculations reveal that the characteristic peak parameters of a certain independent deviation component match the standard spectral characteristic parameters of a known impurity compound recorded in the impurity standard library, reaching a preset matching criterion, the independent deviation component is identified as corresponding to the known impurity compound. At this point, the chemical identifier of the known impurity compound is obtained from the impurity standard library. This identifier can be the impurity's chemical name, CAS number, molecular formula, or a unique internal code, clearly indicating the impurity's chemical composition and structural type. This chemical identifier is then assigned to the corresponding independent deviation component, transforming it from an anonymous deviation signal with only physical signal characteristics into an identified deviation component carrying explicit chemical identity information. For example, when the retention time and UV absorption characteristics of an independent deviation component successfully match the standard parameters of a dechlorination byproduct recorded in the library, the independent deviation component is assigned the chemical identifier "dechlorination byproduct," indicating that the deviation originates from a dechlorination side reaction occurring during synthesis. Independent deviation components that do not match are temporarily left unassigned a chemical identifier, remaining in an unknown state. This identification process provides crucial information for subsequent process anomaly pattern matching, because impurities with known chemical identities often have a clear correspondence with specific process anomaly steps, which can significantly improve the accuracy and confidence of matching candidate anomaly process steps in step S42.

[0095] Step S412c: Based on whether each independent deviation component has been assigned a chemical identity, each independent deviation component is marked, and the marking result is passed to step S413 so that the characteristic parameters of the independent deviation components that have been assigned a chemical identity are extracted first in step S413.

[0096] After completing the spectral library matching and identity assignment operations for all independent deviation components, all components are discriminatively labeled based on whether each independent deviation component has been successfully assigned a chemical identity. Independent deviation components that have been assigned a chemical identity are labeled as identified components, while those that have not are labeled as unidentified components. After labeling, all independent deviation components, along with their respective labeling results, are passed to step S413 for feature parameter extraction. In this process, the labeling results are used to guide the feature parameter extraction order and priority in step S413: feature parameter extraction is prioritized for identified components, as these components already possess clear chemical identity information, and their feature parameters can be matched more effectively with known anomaly patterns in the process anomaly knowledge base, helping to quickly identify candidate anomaly process steps; for unidentified components, feature parameter extraction is still performed according to the conventional process, but their weight and processing strategy in process anomaly matching can be different from that of identified components, for example, they may be labeled as objects requiring subsequent manual review or supplementary analysis. Through this priority passing mechanism based on chemical identity markers, step S413 can make more efficient use of the obtained chemical identification information when extracting deviation feature parameters, which significantly improves the processing efficiency of the entire deviation feature parameter extraction process and the targeting of subsequent anomaly tracing.

[0097] Further, in step S42, the chemical quality deviation characteristic parameters are matched with the abnormal patterns in the preset preparation process abnormality knowledge base to obtain at least one candidate abnormal process step and its matching confidence level, including: Step S421: When there is no record of an abnormal pattern for the target compound identifier by the supplier in the preset preparation process anomaly knowledge base, obtain the chemical structure descriptor of the target compound identifier.

[0098] When matching chemical quality deviation characteristic parameters with a knowledge base of preparation process anomalies, the first step is to determine whether the pre-defined knowledge base contains any anomaly pattern records for the current supplier's identification of the target compound. In actual compound transaction fulfillment scenarios, for compounds traded for the first time or for newly registered suppliers, the knowledge base may not yet contain historical anomaly data for that specific compound. In such cases, direct matching would fail to yield effective results due to a lack of reference. When it is determined that no anomaly pattern record for that specific relationship exists in the knowledge base, the process does not immediately abandon tracing or output a null value. Instead, it retrieves the chemical structure descriptor corresponding to the target compound's identification. A chemical structure descriptor is a computer-readable vectorized or graph-structured representation of a compound's molecular structure, which can take the form of molecular fingerprints, molecular diagrams, or topological descriptors based on atom types and bond relationships. This descriptor encodes the core chemical structural information of the target compound, such as its molecular skeleton features, functional group composition, and ring system structure, providing a data foundation for subsequently using structural similarity to find comparable similar compounds. Through this step of judgment and acquisition, the tracing process can automatically switch to a migration matching path based on structural similarity when the knowledge base coverage is insufficient, rather than being interrupted due to missing data.

[0099] Step S422: Calculate the structural similarity between the chemical structure descriptor and the compound structure descriptors associated with the existing abnormal pattern records in the preset preparation process anomaly knowledge base.

[0100] After obtaining the chemical structure descriptor of the target compound, the structural similarity between this descriptor and the structural descriptors of each compound associated with existing anomalous pattern records in the preparation process anomaly knowledge base is calculated. The purpose of calculating structural similarity is to select compounds from the group of compounds with accumulated anomalous pattern records in the knowledge base that are most similar to the target compound in terms of chemical structure. Structural similarity can be calculated using various mature similarity measurement methods in cheminformatics, such as the Tanimoto coefficient calculation based on molecular fingerprints, which unfolds the molecular fingerprints of two compounds into binary bit vectors and quantifies the degree of similarity by calculating the ratio of the number of shared bits to the total number of bits; alternatively, it can be based on the matching results of the largest common substructure, calculating the proportion of the largest continuous molecular skeleton shared by the two compounds in the entire molecule. After calculation, a structural similarity value between the target compound and each existing recorded compound in the knowledge base is obtained, ranging from 0 to 1. The closer the value is to 1, the more similar the two compounds are structurally. This step transforms the comparison of chemical structures into a quantified similarity score, providing an objective and repeatable numerical basis for judging the degree of structural similarity.

[0101] Step S423: Select at least one similar compound whose structural similarity exceeds a preset similarity threshold, and extract the supplier's historical process anomaly pattern records for the similar compound.

[0102] After obtaining the structural similarity calculation results between the target compound and each existing recorded compound, each similarity value is compared with a preset similarity threshold. At least one compound with a similarity exceeding the threshold is selected as a similar compound. The preset similarity threshold is a configurable parameter, and its value needs to balance the trade-off between the coverage and accuracy of the migration inference. It can usually be set according to the requirements of different compound types or application scenarios. The selected similar compounds have a high degree of structural consistency with the target compound in terms of molecular skeleton, functional group composition, or reactive sites. Therefore, the types of process anomalies they may experience during preparation and the quality deviations caused by these anomalies are of reference value for inferring the causes of the current anomalies of the target compound. Subsequently, the supplier's historical process anomaly pattern records for these similar compounds are extracted. These records are the actual quality anomaly events that the supplier has experienced in the past performance of contracts for similar compounds and their corresponding process step diagnostic conclusions. Through this step, the anomaly tracing, which was originally impossible due to the lack of direct records of the target compound in the knowledge base, is supplemented with transferable historical experience data from structurally similar compounds.

[0103] Step S424: Match the chemical quality deviation characteristic parameters with the extracted historical process anomaly pattern records to obtain candidate abnormal process steps and their matching confidence levels.

[0104] After extracting historical process anomaly patterns from the supplier on similar compounds, the current chemical quality deviation characteristic parameters obtained in step S41 are matched with these historical anomaly patterns. The matching process is logically consistent with the main flow of step S42, comparing the degree of agreement between the current deviation characteristic parameters and the conditional description fields of each anomaly pattern in the historical records. The difference is that the matching object here is not the original record for the current target compound in the knowledge base, but rather historical records migrated from similar compounds. Since the source compound and the target compound are structurally similar but not identical, a confidence adjustment mechanism based on structural similarity can be introduced during the matching process. This ensures that the calculation of the matching confidence reflects both the degree of agreement between the deviation characteristic parameters and the anomaly patterns, and also comprehensively considers the structural similarity between the source compound and the target compound. Finally, at least one candidate abnormal process step and its corresponding matching confidence are output. This cross-compound migration matching method breaks through the limitation that traditional knowledge bases can only work effectively when sufficient historical data of the same compound is accumulated. It enables the first quality anomaly of new or rare compounds to obtain process traceability inference results with a certain degree of confidence, significantly expanding the applicable boundaries of automatic anomaly traceability.

[0105] Furthermore, before triggering the preset performance state machine to perform state transitions on the project performance object in step S4, the following steps are also included: Step S401: After generating the project fulfillment object, obtain the chemical structure attributes corresponding to the target compound identifier.

[0106] After generating the project fulfillment object in step S1 and before triggering the state transition of the fulfillment state machine in step S4, the construction of the fulfillment state machine itself needs to be completed. The fulfillment state machine does not use a uniform fixed topology for all compound transactions; instead, it needs to be customized based on the chemical specificity of the target compound in each transaction. Therefore, step S401 first obtains the chemical structural attributes corresponding to the target compound identifier. Chemical structural attributes are a series of characteristic information extracted or derived from the chemical structural formula of the target compound. They can be obtained by querying a pre-built compound attribute database through the target compound identifier, or by real-time calculation based on the target compound's structural formula file. Specific contents of the chemical structural attributes may include: whether the compound molecule contains one or more chiral centers, thus requiring additional verification at the enantiomeric purity level; whether the compound belongs to a hazardous chemical classification, thus requiring additional safety data file integrity review and verification of hazardous materials transportation conditions; whether the compound has specific photosensitive or thermosensitive functional groups, thus requiring review of whether appropriate light-protection or temperature control measures are configured during transportation; and whether the compound belongs to a controlled substance on a specific regulatory list, thus requiring the insertion of a compliance review node in the fulfillment process. These chemical structural attributes are not generalized product attribute labels, but technical parameters directly derived from molecular structural characteristics and directly related to compound quality verification and performance risk control. They provide a chemical basis for the selection of subsequent state nodes.

[0107] Step S402: Select at least one specific verification state node from the preset state node library based on the chemical structure properties.

[0108] After obtaining the chemical structural properties of the target compound, step S402 selects at least one specific verification state node from a pre-defined state node library based on these properties. The state node library is a predefined set of state nodes that stores various types of verifiable state nodes that can be invoked. Each node type corresponds to a quality or compliance verification operation in the compliance process. For example, the state node library may include a chiral purity verification node to trigger a review process for chiral-related indicators such as enantiomeric excess rate or specific rotation; it may include a dangerous goods transportation compliance verification node to trigger a verification process for items such as dangerous goods transportation packaging grade, transportation temperature range, and leakage emergency response documents; it may include a temperature control record verification node to trigger a process for reading temperature recorder data and comparing thresholds during cold chain transportation; and it may also include a controlled substance compliance review node to trigger an automatic comparison of specific regulatory lists and verification of permit documents. There is a direct mapping relationship between the rules for selecting specific verification state nodes and chemical structural attributes: when the target compound contains a chiral center, a chiral purity verification node is selected from the state node library; when the target compound is a hazardous chemical, a hazardous materials transportation compliance verification node is selected; when the target compound requires cryogenic transportation, a temperature control record verification node is selected. Through this node selection mechanism based on chemical structural attributes, the performance state machine of each compound transaction includes a verification step that matches the chemical characteristics of the compound, rather than an indiscriminate traversal of all possible verification nodes.

[0109] Step S403: Inject the selected specific verification state node into the preset basic performance state machine topology to generate a performance state machine corresponding to the project performance object.

[0110] After selecting specific verification status nodes, step S403 injects these nodes into the preset basic performance state machine topology, generating a performance state machine corresponding to the current project performance object. The basic performance state machine topology is a state machine template containing the general process skeleton of compound transaction performance, which predefines the status nodes of each general link in the performance process and the default transition paths between nodes, such as the sequential connection relationship of conventional states from "to be prepared" to "in preparation" to "to be tested" to "file to be uploaded" to "quality document initial screening passed". The injection operation embeds the specific verification status nodes selected in step S402 into the appropriate positions of the basic topology. For example, the chiral purity verification node is inserted after the conventional purity verification node and before the quality document initial screening passed state; the dangerous goods transportation compliance verification node is inserted between the pending shipment state and the transportation state; and the temperature control record verification node is inserted in parallel with or before the receipt confirmation state. The injection process requires handling the connection relationships between specific verification nodes and existing nodes in the basic topology. This includes determining the preconditions and post-migration targets for the new nodes to ensure that the injected state machine topology still constitutes a logically coherent, deadlock-free, and reachable state transition diagram. The final generated performance state machine is a customized state transition model bound to a specific project performance object, containing a general performance process and compound-specific verification steps. When triggered in step S4, state transitions are executed according to this customized topology. This mechanism makes the state machine itself no longer a static structure universally applicable to all transactions, but a dynamically generated processing flow driven by the chemical properties of the target compound. This avoids omissions or redundancies in specific chemical risk verification steps caused by using a fixed state machine topology.

[0111] Further, in step S402, at least one specific verification state node is selected from a preset state node library based on the chemical structure properties, including: Step S4021: Input the chemical structure attributes into the pre-trained chemical structure-risk association model to obtain the predicted risk impurity spectrum corresponding to the target compound. The chemical structure-risk association model is trained based on the correspondence between the known compound structure and its typical impurities or degradation products.

[0112] When it is necessary to select a specific verification state node from the state node library based on the chemical structure properties in step S402, not only can it be directly mapped and selected based on known and deterministic chemical properties, but it can also perform deeper risk prediction through a pre-trained chemical structure-risk association model, thereby discovering potential quality risks that are not obvious but actually exist.

[0113] This chemical structure-risk association model is a supervised learning model. Its input features are molecular structure descriptors of the target compound, which can include extended connectivity fingerprints, molecular access system fingerprints, or topological descriptors based on molecular graphs. These descriptors encode the compound's molecular skeleton, functional group composition, active site distribution, and interatomic connections into fixed-dimensional numerical vectors or binary bit strings. The model's training data comes from the correspondence between the structural information of a large number of known compounds and typical impurities or degradation products recorded in literature, databases, or historical performance records. Each training sample contains a molecular structure descriptor of a known compound as input features, and a list of known impurity types or degradation product types corresponding to that compound as labels. The scale of the training data can be gradually expanded by continuously aggregating historical quality data from multiple suppliers and publicly available chemical stability data. During training, the model uses the molecular structure descriptor as input and the corresponding impurity or degradation product type as the prediction target. By minimizing the multi-label classification loss function between the predicted output and the labels, the model's internal weight parameters are iteratively adjusted until the model's prediction accuracy on the validation set reaches the preset requirements. After the chemical structural attributes of the target compound obtained in step S401 are converted into molecular structure descriptors and input into the trained model, the model calculates the degree of matching between the descriptor and the learned structure-risk association patterns, and outputs the predicted risk impurity spectrum corresponding to the target compound. This predicted risk impurity spectrum is a structured dataset that lists the types, chemical structural characteristics, predicted confidence scores, and possible detection scenarios of various potential impurities or degradation products that may be generated during the synthesis, purification, or storage of the target compound. Unlike direct mapping based on a rule base, the predicted risk impurity spectrum can identify complex risks caused by the combined effects of multiple structural factors that are outside the coverage of conventional rules, providing a more forward-looking decision-making basis for the selection of subsequent state nodes.

[0114] Step S4022: Based on the detection requirements of each impurity or degradation product in the predicted risk impurity spectrum, select the corresponding risk verification state node from the preset state node library.

[0115] After obtaining the predicted risk impurity spectrum of the target compound, step S4022 selects the corresponding risk verification state nodes from the preset state node library according to the detection requirements of various potential impurities or degradation products listed in the impurity spectrum. Each potential impurity or degradation product in the predicted risk impurity spectrum needs to be effectively identified and quantified in the actual delivered test sample through a specific detection method. For example, for the predicted chiral isomer impurities, a chiral detection node capable of verifying enantiomeric excess or specific rotation needs to be selected; for the predicted thermal degradation products, a method node capable of chromatographic separation and detection of the degradation products with sufficiently high sensitivity at a specific detection wavelength needs to be selected; for the predicted oxidation byproducts, a mass spectrometry detection node may need to be selected to verify whether a molecular weight peak corresponding to the oxidation product appears. The logic for selecting risk verification state nodes is to transform the predicted chemical risks into corresponding detection verification steps: each risk item output in the model is associated with one or more recommended detection method types, and these detection method types have a corresponding relationship with the verification state nodes in the state node library. By querying this correspondence, the selection from risk items to state nodes can be completed. This step extends the selection of state nodes beyond known deterministic chemical properties to include potential risk domains predicted by statistical learning and pattern recognition, thereby injecting proactive risk prevention capabilities into the performance state machine.

[0116] Step S4023: The risk verification state node is injected as a specific verification state node into the basic performance state machine topology of step S403.

[0117] After selecting the corresponding risk verification state nodes based on the predicted risk impurity spectrum, step S4023 injects these risk verification state nodes as components of the specific verification state nodes selected in step S402 into the basic performance state machine topology of step S403. The risk verification state nodes and the conventional specific verification state nodes selected based on deterministic chemical properties employ the same processing mechanism in terms of injection method and topology arrangement; that is, the insertion position and migration path in the basic performance state machine are determined according to the verification content of the node. The difference lies in that the risk verification state nodes correspond to potential risk detection items that have not yet been explicitly committed to by the supplier or required by the purchaser. Their triggering conditions and anomaly judgment thresholds in the state machine can be set to flexible configuration methods different from mandatory quality indicators. For example, they can be configured as information prompt-level verification nodes, generating early warning information instead of directly blocking the state transition process when the corresponding risk signal is detected. By injecting the predicted risk verification nodes into the performance state machine, the resulting customized state machine not only includes the verification of the quality indicators promised by the supplier, but also pre-sets verification steps to actively explore potential risks that may exist at the chemical structure level of the target compound. This improves the quality control of the entire performance process from passively responding to the data provided by the supplier to proactively discovering potential chemical risks that may affect the quality of the test sample.

[0118] Further, in step S403, the selected specific verification state node is injected into the preset basic performance state machine topology to generate a performance state machine corresponding to the project performance object, including: Step S4031: Obtain the timing constraints and / or resource constraints of at least two specific verification status nodes selected in step S402.

[0119] After selecting at least two specific verification state nodes from the preset state node library in step S402, these nodes need to be injected into the basic fulfillment state machine topology, but this cannot be done simply by inserting them one by one. Different types of verification state nodes have their own execution condition requirements and resource requirements in the fulfillment process, and these constraints need to be explicitly obtained before being injected into the topology. Step S4031 obtains the timing constraints and resource constraints of these specific verification state nodes. Timing constraints refer to the timing restrictions on the execution of a certain verification state node in the fulfillment process, including after which the node must start execution, before which state it must complete execution, and whether there are sequential dependencies between the node and other nodes. For example, the chiral purity verification node must be executed after the chemical purity verification is completed, because the determination of chiral purity usually requires the premise that the main component of the test sample has been confirmed; the temperature control recording verification node must be executed before the end of the transportation state and the start of the receipt confirmation state, because temperature data can only be fully obtained after transportation is completed. Resource constraints refer to whether a specific verification node requires exclusive access to the data or detection resources it needs during execution. For example, a mass spectrometry node and an NMR node might need to use the same sample, but the sample quantity might be insufficient to support both detections simultaneously, creating a resource competition between them. Another example is that the execution of certain verification nodes requires the intervention of authorized reviewers, whose processing capacity has an upper limit at any given time. This constraint information can be read from the attribute fields of each node in the status node library, or it can be dynamically derived based on the node type and compound characteristics.

[0120] Step S4032: Detect whether there are timing conflicts or resource conflicts between specific verification state nodes and between specific verification state nodes and existing nodes in the basic performance state machine topology.

[0121] After obtaining the constraints of each specific verification state node, step S4032 performs conflict detection to determine whether there are timing or resource conflicts between the specific verification state nodes, and between these specific verification state nodes and existing nodes in the basic fulfillment state machine topology. A timing conflict refers to a situation where the constraints of two or more verification state nodes contradict each other in the time dimension, making it impossible to schedule these nodes simultaneously in a legal execution order. For example, the chiral purity verification node requires execution after the detection state, while the temperature control recording verification node requires execution before the receipt confirmation state. If multiple other state nodes are inserted between the detection state and the receipt confirmation state in the basic topology, the two verification nodes themselves do not directly conflict, but their respective timing constraints may indirectly contradict the positional relationships of other nodes in the basic topology. A more typical timing conflict manifests as two nodes interlocking: node A requires execution after node B, and node B requires execution after node A, forming a circular dependency that makes it impossible to arrange a legal order. A resource conflict refers to a situation where two or more verification state nodes need to compete for the same restricted resource during execution, and this resource can only be exclusively used by one node at a time. For example, mass spectrometry detection nodes and high-performance liquid chromatography-mass spectrometry (HPLC-MS) detection nodes may both require the use of the same mass spectrometer, which can only perform one detection task per analysis cycle. Another example is that multiple calibration nodes require manual verification by the purchasing party's quality control personnel, whose work queues have processing capacity limits. The purpose of conflict detection is to identify potential execution obstacles before nodes are injected into the topology, preventing logical or resource deadlocks during the actual execution of the generated fulfillment state machine.

[0122] Step S4033: When a conflict is detected, the preset conflict resolution strategy library is invoked to merge, split or rearrange the specific verification state nodes according to the conflict type, so as to obtain the resolved state node set.

[0123] When step S4032 detects a timing conflict or resource conflict, step S4033 calls the preset conflict resolution strategy library to perform corresponding resolution operations on the specific verification state nodes involved according to the detected conflict type, in order to eliminate the conflict. The conflict resolution strategy library is a pre-built set of rules that defines standardized processing strategies corresponding to different types of conflicts. For timing conflicts, the conflict resolution strategy may include a sequence rearrangement operation, that is, adjusting the relative execution order of nodes in the topology without violating the core constraints of each node, such as rearranging nodes without strong dependencies according to their respective allowed time windows to eliminate interlocking; it may also include a node merging operation, where two verification state nodes have overlapping functions and can be completed by the same detection method, and are merged into a composite verification node, such as merging a mass spectrometry molecular weight verification node and a mass spectrometry fragment resolution node into a unified mass spectrometry analysis node, thereby eliminating resource competition between the two; it may also include a node splitting operation, where a verification node cannot be fully executed due to resource conflicts, and is split into two sequentially executed sub-nodes, which use competing resources in different time periods. For resource conflicts, conflict resolution strategies can include resource scheduling priority ranking, allocating different resource usage priorities based on the criticality of each verification node in the quality compliance judgment, with low-priority nodes being postponed or downgraded to optional verification; alternative resource selection can also be included, when a resource is contested by multiple nodes and cannot be resolved by ranking, a functionally equivalent alternative verification node using different resources is selected from the state node library to replace it. Through the above resolution operations, the original set of specific verification state nodes with conflicts is reorganized into a logically consistent and reasonably resource-allocated set of resolved state nodes.

[0124] Step S4034: Inject the parsed set of state nodes into the basic performance state machine topology to generate the performance state machine.

[0125] After conflict resolution is completed, step S4034 injects the resolved set of state nodes into the basic performance state machine topology, ultimately generating a performance state machine corresponding to the current project performance object. This injection operation is consistent with the regular injection in step S403 in terms of process, the difference being that the injected object has changed from the original specific verification state nodes initially selected in step S402 to the set of state nodes after conflict detection and resolution. During the injection process, based on the temporal attributes and transition relationships of each node in the resolved node set, the precise insertion position of each node in the basic topology needs to be determined, the transition conditions between the node and its predecessor and successor states need to be established, and the node's internal parameters and triggering logic need to be configured. After completing the insertion and connection configuration of all nodes, a complete performance state machine definition that can be called by the state machine engine is generated. The fulfillment state machine integrates the general framework of the basic fulfillment process with specific verification steps customized for the chemical properties of the target compound in terms of topology. The node arrangement has been verified for timing legality and resource feasibility. When triggered in step S4, it can smoothly execute the verification tasks of each state node according to the predetermined migration path, avoiding state machine execution interruption or deadlock caused by node conflicts, and ensuring the reliability and stability of the customized fulfillment process in actual operation.

[0126] Furthermore, the generation of quality compliance judgment results in step S3 also includes: when the judgment result is non-compliance, determining the non-compliance type as missing documents or substandard quality indicators. In generating quality compliance judgment results, when the judgment conclusion is non-compliance, step S3 no longer simply outputs a single "non-compliance" label, but further determines the specific type of non-compliance as missing documents or substandard quality indicators. The technical significance of this distinction lies in the fundamental difference in the handling logic and subsequent processing path of the two types of non-compliance situations in performance management. Missing documents refer to the supplier's uploaded quality documents failing to cover the set of required document types recorded in the project performance object in terms of type or quantity. For example, the promised high-performance liquid chromatography or nuclear magnetic resonance spectrum is missing. In this case, the quality indicators themselves have not yet entered a judgmentable state; the problem lies in incomplete information rather than unmet quality requirements. Substandard quality indicators refer to the fact that the document types are complete, but the extracted structural compliance measured indicators fail to reach the preset chemical quality threshold. For example, the measured purity value in the analysis certificate is lower than the promised value, or the percentage of the main peak area in the chromatogram does not meet the minimum requirement. In this case, the problem lies in the chemical quality of the deliverables not meeting the agreed standards.

[0127] Accordingly, step S4, after the quality compliance judgment result is non-compliant and the project performance object is migrated to the document supplementation state, quality anomaly handling state, or deviation handling state, includes: when the non-compliance type is missing documents, migrating the project performance object to the document supplementation state; when the non-compliance type is substandard quality indicators, migrating the project performance object to the quality anomaly handling state. The state migration mechanism in step S4 executes differentiated state routing based on the distinction of non-compliance type. When the non-compliance type is missing documents, the project performance object is migrated to the document supplementation state. This state is used to prompt the supplier to upload the specified type of quality documents according to the missing list. After the documents are completed, the parsing and comparison process can be retried, which is a repairable process anomaly. When the non-compliance type is substandard quality indicators, the project performance object is migrated to the quality anomaly handling state or deviation handling state. The "Quality Anomaly Handling Status" applies when there is a clear and significant deviation between the measured chemical quality indicators and the preset chemical quality thresholds. This indicates that the chemical quality of the current batch of products has been determined to be non-compliant with the committed standards, requiring a more in-depth anomaly handling process, such as deviation investigation, supplier rectification, re-inspection application, or return. This constitutes a substantive quality anomaly. The "Deviation Handling Status" applies when the quality indicators are not up to standard but the deviation is within the preset acceptable boundary, or when the supplier has submitted an explanation of the deviation and requested concessionary acceptance. In this case, the deviation assessment and risk release decision process is initiated, where the purchasing party or quality control personnel decide whether to accept the batch of products based on the severity of the deviation and its impact on subsequent use. This approach, which further subdivides non-compliance types into missing documents and non-compliance with quality indicators, and further classifies non-compliance with quality indicators into two subtypes—those requiring the Quality Anomaly Handling Process and those eligible for the Deviation Assessment Process—enables the performance status machine to perform precise process control for different types of anomalies. This avoids delays and unclear responsibility caused by conflating incomplete information, severe quality non-compliance, and boundary deviations.

[0128] Optionally, in step S4, after the quality compliance judgment result is compliant and the project performance object is moved to the quality document initial screening pass status, the method further includes: generating a review task and sending it to the purchaser or quality control role; receiving the confirmation instruction from the purchaser or quality control role for the review task, and moving the project performance object from the quality document initial screening pass status to the pending shipment status or the quality confirmation completion status.

[0129] In step S4, after the quality compliance assessment result is compliant and the project performance object has been moved to the initial screening pass status of quality documents, the performance process does not directly enter the delivery stage. Instead, it triggers an intermediate mechanism for generating and confirming a review task. Specifically, the system automatically generates a review task, which includes at least a summary of the quality indicators judged to be compliant, a list of parsed documents, and key information such as structural matching scores. This task is then pushed to the technical personnel, procurement manager, or quality control role designated by the purchaser. Upon receiving the task, the reviewer manually verifies the original quality documents against the system's parsing results. After confirming that all data interpretations are accurate, document attribution is correct, and there are no system misjudgments, a confirmation instruction is issued. Upon receiving the confirmation instruction, the system moves the project performance object from the current initial screening pass status of quality documents to the pending delivery status or the quality confirmation completion status, and the performance process then enters the logistics and delivery stage. This mechanism sets up a necessary professional review barrier between the automatic analysis results and the final quality assessment, so that the efficient processing capability of the computer program and the experience judgment of chemical professionals can complement each other effectively, avoiding automatic misjudgments caused by spectral analysis deviations, improper threshold settings or file association errors from directly entering subsequent business processes.

[0130] In a specific embodiment of the present invention, after triggering the preset performance state machine to perform state transition on the project performance object in step S4, the method further includes: generating and storing a state transition audit record, wherein the audit record includes file upload information, parsing results, threshold comparison results, file overwrite judgment results, state identifiers before and after state transition, trigger rule version, trigger time, and operation subject identifier.

[0131] After step S4 triggers the preset performance state machine to perform state transitions on the project performance object, a structured state transition audit record is generated and stored. This audit record is not a simple operation log, but a full-element data snapshot of a complete state transition event, containing at least the following information dimensions: file upload information for tracing the source and version of the quality file that triggered the transition; parsing results and threshold comparison results record the comparison details of the measured structural compliance indicators and their corresponding thresholds; file coverage judgment results record the coverage comparison conclusion between the set of required file types and the actual received file types; state identifiers before and after the transition clearly record from which state the project performance object transitioned to which state; trigger rule version for identifying the version number of the comparison rule and state machine configuration on which this transition was based; trigger time precisely records the moment the transition occurred; and operation subject identifier distinguishes whether the transition was automatically triggered by the program or triggered by manual review and confirmation. All of the above dimensions of information together constitute an immutable and traceable electronic evidence chain, ensuring that every state change during the performance process is verifiable, providing a structured data foundation for subsequent quality dispute resolution, supplier performance capability evaluation, and compliance audits.

[0132] The following is a complete embodiment to illustrate the present technical solution. The specific values ​​and compound information in the embodiment are exemplary and do not constitute a limitation on the scope of protection.

[0133] A company's purchasing department needs to procure a batch of key intermediate compounds. The target compounds are heterocyclic carboxylic acid derivatives with a single chiral center, requiring a purity of no less than 98% and an enantiomeric excess of no less than 99%. The purchasing party sent quotations to three candidate suppliers through a price inquiry platform. Each supplier returned a response version, which included a price quote, delivery date, purity commitment, and a list of quality documents to be delivered. After comparison, the purchasing party confirmed the second version of the response returned by Supplier A as the target supplier's response version. This response version explicitly committed to a purity of no less than 98% for the target compound, an enantiomeric excess of no less than 99%, and provided five types of quality documents, including an analytical certificate, high-performance liquid chromatography (HPLC) chromatogram, proton NMR spectroscopy, mass spectrometry, and a material safety data sheet.

[0134] Based on the supplier's response version confirmed by the purchaser, a project fulfillment object was generated. The target compound associated with this fulfillment object is identified by its CAS number and internal standardized code within the system. Chemical quality thresholds extracted from the response version include a purity threshold of 98% and a main peak area percentage threshold of 98%. The set of required document types includes five items: analytical certificate, high-performance liquid chromatography (HPLC), proton NMR spectroscopy, mass spectrometry, and material safety data sheet. Simultaneously, because the target compound contains a chiral center, its chemical structural properties were acquired during the fulfillment state machine generation stage, identifying it as containing one chiral carbon atom and a carboxyl functional group, making it sensitive to temperature and humidity. Based on these chemical structural properties, a chiral purity verification node and a temperature control record verification node were selected from a pre-defined state node library and injected into the basic fulfillment state machine topology. The chiral purity verification node is scheduled after the routine purity verification is passed and before the initial quality document screening is passed; the temperature control record verification node is scheduled after the "in transit" state and before the "receipt confirmation" state. This generates a customized fulfillment state machine corresponding to the fulfillment tasks of this batch of intermediates.

[0135] Supplier A completed the preparation and testing of the target compound within the agreed delivery period and uploaded five quality documents via a file transfer interface. Upon receiving the documents, the content of each document was first parsed. The analysis certificate, a text file, had its batch number B20240501, measured purity of 98.5%, and testing date of May 10, 2024, extracted through text parsing. The high-performance liquid chromatography (HPLC) chromatogram, proton nuclear magnetic resonance (NMR) spectrum, and mass spectrum were spectral files, and characteristic peak extraction was performed on each. Characteristic peak extraction from the HPLC chromatogram yielded a set of characteristic peak parameters, including a main peak with a retention time of 8.2 minutes and an area percentage of 98.2%, and a trace impurity peak with a retention time of 12.5 minutes and an area percentage of 0.6%. Characteristic peak extraction from the proton NMR spectrum yielded the integrated area and coupling splitting information of the signal peaks corresponding to each chemical shift value. Characteristic peak extraction from the mass spectrum yielded the mass-to-charge ratio of the molecular ion peak and its isotopic distribution pattern.

[0136] After obtaining the characteristic peak parameter set of the spectral files, coupling constant information was extracted from the characteristic peak data of the proton NMR spectrum. By analyzing the coupling constant value between two hydrogen atoms on adjacent carbon atoms of the chiral center, which is 9.8 Hz, and based on the correspondence between the coupling constant and dihedral angle described by the Karplacian equation, it was inferred that these two hydrogen atoms are in an anti-coplanar conformation, thus determining the relative configuration of the chiral center to be R configuration. This relative configuration attribute was encoded into a stereochemical feature vector, which was then fused with the two-dimensional connectivity descriptor of the sample to generate a three-dimensional chemical feature descriptor of the sample containing chiral configuration information.

[0137] Subsequently, the three-dimensional chemical characterization descriptors of the test sample were matched with the target chemical characterization descriptors corresponding to the target compound identifiers in the first round of structural matching, resulting in an initial matching score of 92%. The correspondence between each characteristic peak and the target structural fragment was also obtained. Based on this correspondence, attribution verification was performed. The main signal peaks in the NMR spectrum were successfully assigned to the various structural fragments of the target compound. The main peak with a retention time of 8.2 minutes in the HPLC spectrum was assigned to the main product peak of the target compound. However, the trace impurity peak with a retention time of 12.5 minutes could not be assigned to any known structural fragment in the target chemical characterization descriptors and was marked as a residual characteristic peak.

[0138] Before sending the residual characteristic peak into the subsequent correction process, its signal-to-noise ratio (SNR) and peak shape symmetry parameters were evaluated. The SNR of the residual characteristic peak was 15:1, higher than the preset SNR threshold of 10:1, and its peak shape asymmetry factor was 1.1, falling within the preset symmetry tolerance range of 0.8 to 1.5. Therefore, the residual characteristic peak passed the reliability verification, was not marked as a resolution artifact, and was retained for subsequent processing. A more refined peak area integration was performed on the verified residual characteristic peak, confirming its peak area percentage of 0.6%. Based on the attribution verification results, the chemical characteristic descriptor of the test sample was corrected, the assigned characteristic peaks were confirmed, and the residual characteristic peak was added to the descriptor as a new impurity characteristic node. A second round of graph structure matching was performed between the corrected chemical characteristic descriptor of the test sample and the target chemical characteristic descriptor. In this round of matching, all structural fragments of the target compound were correctly matched, but the newly added impurity characteristic node could not find a corresponding item in the target descriptor. The final output structure matching score was 96%.

[0139] During the quality compliance assessment phase, the measured purity value of 98.5% extracted from the analysis certificate was compared with the chemical quality threshold of 98%, and the measured value met the threshold requirement. The percentage of the main peak area in the high-performance liquid chromatography (HPLC) chromatogram (98.2%) also met the threshold requirement. The analytical results of the nuclear magnetic resonance (NMR) spectrum and mass spectrum were consistent with the structure of the target compound. The type set of received documents covered all five items of the required document type set, and the document completeness met the requirements. Simultaneously, the structure matching score of 96%, as a comprehensive structural compliance indicator, reflected that the test sample was highly consistent with the target compound at the chemical structure level. Based on the above assessments, a compliant quality compliance assessment result was generated. Based on this compliance result, the preset performance status machine was triggered, and the project performance object transitioned from the current state to the initial screening of quality documents. After receiving the confirmation notification, the purchaser's quality control personnel reviewed the analytical results and original documents, confirmed their accuracy, and the performance object was further transitioned to the pending shipment state.

[0140] Supplier A subsequently delivered the products via cold chain transportation as agreed, with a temperature recorder continuously collecting temperature data during transport. Upon completion of transportation, the temperature control recording verification node was activated, reading the entire temperature curve data and comparing it with the preset temperature control range of 2 to 8 degrees Celsius. The temperature remained within the specified range throughout, and the verification passed. After receiving the goods, the purchaser conducted a visual inspection and internal re-inspection, confirming the product's quality and completing the receipt confirmation. At this point, the fulfillment process for this batch of intermediates was completed, and the project fulfillment object was moved to the completed status.

[0141] During this contract fulfillment process, a non-compliance assessment was found in the quality documents uploaded by Supplier B for another procurement order targeting the same compound. This case illustrates the working process of the anomaly tracing mechanism. Supplier B uploaded a high-performance liquid chromatography (HPLC) chromatogram for another procurement order for the same heterocyclic carboxylic acid derivative. Analysis revealed that the main peak area percentage was only 94.5%, below the 98% threshold requirement. Furthermore, an impurity peak with an area percentage of 2.5% appeared at retention time 12.5 minutes, and another impurity peak with an area percentage of 0.8% appeared at retention time 15.8 minutes. The quality compliance assessment result was non-compliant, and the project was moved to a quality anomaly handling status.

[0142] After initiating anomaly tracing, the chemical quality deviation characteristic parameters leading to the non-compliance judgment were extracted from the quality compliance assessment results. These included the total purity deviation of 3.5 percentage points, and the retention time and peak area percentage of each of the two impurity peaks. During the identification of deviation indicators, it was found that the total purity deviation was not caused by a single impurity source, but rather by the combined contribution of two impurity peaks with different retention times, indicating a situation involving multiple independent deviation sources. Chemical semantic decoupling was performed on the deviation indicators, decomposing the composite deviation into two independent deviation components: the first component corresponds to the impurity with a retention time of 12.5 minutes, contributing 2.5% to the deviation; the second component corresponds to the impurity with a retention time of 15.8 minutes, contributing 0.8% to the deviation.

[0143] After obtaining the characteristic peak parameters of each independent deviation component, they were matched with a pre-defined impurity standard library. The retention time and UV absorption characteristics of the first component successfully matched the standard parameters of a dechlorination byproduct recorded in the library. This dechlorination byproduct is a known impurity generated during the dechlorination reaction of a chlorination intermediate in the synthesis of the target compound. The first component was assigned the chemical identity of "dechlorination byproduct". No matching entry was found for the second component in the library, so it was not assigned a chemical identity for the time being. Based on the labeling results, the identified first component was prioritized when extracting characteristic parameters in the subsequent process.

[0144] The characteristic parameters of the identified first deviation component were matched with a pre-defined knowledge base of abnormal preparation processes. The knowledge base contained records of abnormal patterns for this heterocyclic carboxylic acid derivative by Supplier B. Among these, the candidate abnormal process steps corresponding to excessive residue of dechlorination byproducts were "chlorination reaction time control deviation" and "insufficient purification and recrystallization times," with confidence levels of 0.85 and 0.62, respectively. The candidate abnormal process steps and their confidence levels were associated with project fulfillment objects already under quality abnormality or deviation handling status, generating an abnormality traceability record. Supplier B checked the preparation records for this batch based on the traceability results and confirmed that the chlorination step's reaction time was shortened by approximately 2 hours compared to the standard process, resulting in incomplete conversion of the chlorinated intermediate. The unreacted intermediate underwent dechlorination in subsequent steps and remained in the final product. Supplier B adjusted the reaction time parameters accordingly and re-prepared the product. The purchaser, based on the complete traceability record, assessed and recorded this supplier's performance abnormality.

[0145] It should be noted that the reception of quality documents in this invention is not limited to a unified shopping cart page or a single file upload entry. In actual deployment, quality documents can be accessed from various business interfaces or system interfaces, including but not limited to: the file upload area in the project details page, the performance document submission entry in the order management page, the quality control review page dedicated to the quality control department, the task response interface in the supplier's workbench, open application programming interfaces (APIs), electronic lab notebook system interfaces, laboratory information management system interfaces, data output interfaces of analytical instrument workstations, and data feedback interfaces of logistics monitoring systems. All quality documents or quality data submitted from the above-mentioned access sources undergo format standardization processing through a unified data receiving layer before entering the content parsing and association key generation process in step S2. Regardless of the entry point from which the quality document is submitted, it does not affect the extraction of structural compliance measurement indicators, the generation of quality compliance judgment results, or the triggering of the performance state machine migration in subsequent steps. This approach allows the method of this invention to adapt to the differences in the information infrastructure of purchasers and suppliers of different sizes, avoiding the circumvention of the scope of protection of this invention by competitors dispersing the file submission function across multiple business pages or independent systems.

[0146] Accordingly, please refer to Figure 2 A second aspect of this invention provides a compound transaction performance status transition system based on quality document parsing, which processes project performance data based on the aforementioned compound transaction performance status transition method based on quality document parsing, including: The performance object generation module 1 is used to obtain the supplier response version confirmed by the purchaser, order data, project start data or performance start instructions, and generate project performance objects based on the obtained data; wherein, the project performance object is associated with the target compound identifier, the chemical quality threshold extracted from the obtained data and the set of requirement document types; The quality document parsing module 2 is used to receive the quality document corresponding to the project performance object, parse the content of the quality document, extract the structural conformity test indicators, and determine the document type of the quality document; generate a quality document association key based on at least one of the following: document hash, target compound identifier, batch number, test type, supplier identifier, order identifier, response version identifier, and key quality features, and bind the quality document association key to the project performance object; The quality compliance judgment module 3 is used to compare the measured structural compliance indicators with the chemical quality threshold associated with the project performance object, and to determine whether the file type of the received quality document covers the set of required file types, and to generate a quality compliance judgment result. The state machine transition execution module 4 is used to trigger a preset performance state machine to perform state transitions on the project performance object based on the quality compliance judgment result. Specifically, when the quality compliance judgment result is compliant, the project performance object is transitioned to the quality document initial screening pass state, the state awaiting purchaser confirmation state, or the state awaiting quality control review state. When the quality compliance judgment result is non-compliant, the project performance object is transitioned to the document supplementation state, the quality anomaly handling state, or the deviation handling state.

[0147] Accordingly, a third aspect of the present invention provides an electronic device, including: at least one processor and a memory connected to the at least one processor. The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, cause the at least one processor to perform the aforementioned compound transaction fulfillment status transition method based on quality document parsing.

[0148] Accordingly, a fourth aspect of the present invention provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the above-described compound transaction fulfillment status transition method based on quality document parsing.

[0149] Accordingly, a fifth aspect of the present invention provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the above-described compound transaction fulfillment status transition method based on quality document parsing.

[0150] The embodiments of the present invention aim to protect a method and system for transitioning the fulfillment status of compound transactions based on quality document parsing, which has the following effects: 1. By elevating the parsing of spectral quality files from traditional numerical extraction to graph structure matching at the chemical structure level, and further integrating stereochemical features such as coupling constants and nuclear Overhaus effect spectroscopy to construct three-dimensional descriptors, the consistency between the test sample and the target compound can be automatically verified from two dimensions: molecular skeleton connectivity and chiral configuration. This effectively overcomes the limitations of conventional methods that rely solely on purity values ​​or single peak area ratios for judgment, which may miss deep-seated quality defects such as isomer errors and unsatisfactory enantiomer excess rates. It significantly enhances the accuracy and chemical specificity of quality compliance determination in compound transactions, providing a more solid decision-making basis for the automatic migration of subsequent performance status. 2. When the quality compliance is determined to be non-compliant and the non-compliance type is a failure to meet quality indicators, the process no longer stops at simply marking the abnormal state. Instead, it decouples the deviation indicators chemically and semantically, uses independent component analysis to perform blind source separation on the original spectral signal, and decomposes the originally mixed and superimposed composite deviations into independent deviation components corresponding to specific chemical structural fragments or impurity source types. These components are then automatically matched with impurity standard spectral libraries to assign chemical identities. This achieves an automated leap from detecting an anomaly to diagnosing the cause of the anomaly, solving the technical problem of difficulty in clarifying deviation responsibility in multi-factor scenarios. This allows the anomaly tracing results to be directly linked to specific process steps or impurity sources, providing precise structured guidance for the subdivision and targeted handling of quality anomaly status. 3. In the process of tracing the source of manufacturing process anomalies, when the pre-set manufacturing process anomaly knowledge base lacks historical anomaly patterns of the supplier for the current target compound, the chemical structural similarity between the target compound and existing compounds in the knowledge base is calculated. The supplier's historical process anomaly patterns on similar compounds are then matched across compounds to obtain candidate anomaly process steps and their confidence levels. This breaks through the limitation of the knowledge base relying on the historical accumulation of the same compound, enabling new compounds, rare compounds, or compounds traded for the first time to obtain reasonable anomaly tracing inferences based on structural similarity. This significantly improves the coverage of process anomaly diagnosis and the usability in cold start scenarios, further improving the data-driven fulfillment status migration closed loop.

[0151] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0152] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0153] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0154] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for transitioning the fulfillment status of compound transactions based on quality document parsing, characterized in that, Includes the following steps: Step S1: Obtain the supplier's response version confirmed by the purchaser, order data, project initiation data or performance initiation instructions, and generate a project performance object based on the obtained data; wherein, the project performance object is associated with the target compound identifier, the chemical quality threshold extracted from the obtained data and the set of requirement document types; Step S2: Receive the quality document corresponding to the project performance object, parse the content of the quality document, extract the structural conformity test indicators, and determine the file type of the quality document; generate a quality document association key based on at least one of the following: file hash, target compound identifier, batch number, test type, supplier identifier, order identifier, response version identifier, and key quality features, and bind the quality document association key to the project performance object; Step S3: Compare the measured structural compliance index with the chemical quality threshold associated with the project performance object, and determine whether the file type of the received quality document covers the set of required file types, and generate a quality compliance judgment result. Step S4: Based on the quality compliance judgment result, trigger the preset performance state machine to perform state transition for the project performance object; wherein, when the quality compliance judgment result is compliant, the project performance object is transitioned to the quality document initial screening passed state, the state awaiting confirmation by the purchaser, or the state awaiting quality control review; when the quality compliance judgment result is non-compliant, the project performance object is transitioned to the document supplementation state, the quality anomaly handling state, or the deviation handling state.

2. The compound transaction fulfillment status transition method based on quality document parsing according to claim 1, characterized in that, Step S2, which involves parsing the quality document to extract the structural conformity measurement indicators, includes: Step S21: When the file type of the quality file is a spectrum file, perform feature peak extraction on the spectrum file to obtain a set of feature peak parameters; Step S22: Construct chemical feature descriptors based on the set of characteristic peak parameters to obtain chemical feature descriptors for the test sample; Step S23: Obtain the target chemical feature descriptor corresponding to the target compound identifier associated with the project performance object; perform graph structure matching between the test sample chemical feature descriptor and the target chemical feature descriptor; use the structure matching score as the measured index of structure conformity; and combine it with at least one of the measured purity value, main peak area percentage, molecular weight peak matching result, and impurity peak parameters to constitute the measured chemical quality information.

3. The compound transaction fulfillment status transition method based on quality document parsing according to claim 2, characterized in that, Step S22, which involves constructing a chemical characteristic descriptor based on the set of characteristic peak parameters to obtain a chemical characteristic descriptor for the test sample, includes: Step S221: Extract coupling constant information and / or nuclear Overhaus effect spectrum feature information from the set of characteristic peak parameters; Step S222: Based on the coupling constant information and / or the nuclear Overhouse effect spectrum characteristic information, determine the relative configurational properties of at least one chiral center in the test sample; Step S223: Encode the relative configuration attribute into a stereochemical feature vector, and fuse the stereochemical feature vector with the two-dimensional connectivity descriptor of the test sample to generate a three-dimensional chemical feature descriptor of the test sample, which serves as the chemical feature descriptor of the test sample.

4. The compound transaction fulfillment status transition method based on quality document parsing according to claim 2, characterized in that, Step S23, which involves performing graph structure matching between the chemical feature descriptor of the test sample and the target chemical feature descriptor, and using the structure matching score as the measured index of structure conformity, includes: Step S231: Perform the first round of graph structure matching to obtain the initial matching score and the correspondence between feature peaks and structural segments; Step S232: Based on the correspondence between the characteristic peak and the structural fragment, the set of characteristic peak parameters extracted in step S21 is subjected to attribution verification to identify at least one characteristic peak belonging to a specific structural fragment in the target chemical characteristic descriptor and at least one residual characteristic peak not belonging to the target chemical characteristic descriptor. Step S233: Re-extract the characteristic peak parameters for the residual characteristic peaks, and correct the chemical characteristic descriptor of the test sample according to the attribution verification result to obtain the corrected chemical characteristic descriptor of the test sample; Step S234: Perform a second round of graph structure matching between the modified chemical feature descriptor of the test sample and the target chemical feature descriptor, and use the structure matching score obtained from the second round of matching as the measured index of structure conformity.

5. The compound transaction fulfillment status transition method based on quality document parsing according to claim 4, characterized in that, After identifying at least one characteristic peak belonging to a specific structural fragment in the target chemical feature descriptor in step S232, the method further includes: Step S232a: Obtain the signal-to-noise ratio and peak shape symmetry parameter for each of the residual characteristic peaks; Step S232b: When the signal-to-noise ratio value of any residual feature peak is lower than a preset signal-to-noise ratio threshold, or the peak shape symmetry parameter exceeds a preset symmetry tolerance range, the residual feature peak is marked as a resolution artifact. Step S232c: Remove the residual feature peaks marked as parsing artifacts from the residual feature peaks, and input only the verified residual feature peaks into step S233.

6. The compound transaction fulfillment status transition method based on quality document parsing according to claim 1, characterized in that, Step S4, after the quality compliance judgment result is non-compliant and the project performance object is moved to the quality anomaly handling state or deviation handling state, further includes: Step S41: Extract the chemical quality deviation characteristic parameters that cause non-compliance judgment from the quality compliance judgment results; Step S42: Match the chemical quality deviation characteristic parameters with the abnormal patterns in the preset preparation process abnormality knowledge base to obtain at least one candidate abnormal process step and its matching confidence level. Step S43: Associate the candidate abnormal process step and its corresponding matching confidence level with the project performance object that has been migrated to the quality abnormality handling state or deviation handling state, and generate an abnormality tracing record.

7. The compound transaction fulfillment status transition method based on quality document parsing according to claim 6, characterized in that, The extraction of chemical quality deviation characteristic parameters that lead to non-compliance judgment in step S41 includes: Step S411: Obtain at least one deviation index item in the quality compliance judgment result that leads to non-compliance judgment, and identify whether the deviation index item contains the superimposed contribution of multiple independent deviation sources. Step S412: When the deviation index is identified as a superimposed contribution containing multiple independent deviation sources, chemical semantic decoupling is performed on the deviation index item to decompose the composite deviation into at least two independent deviation components, wherein each of the independent deviation components corresponds to a chemical structural fragment or an impurity source type. Step S413: Extract the characteristic parameters of each independent deviation component to obtain a set of independent deviation characteristic parameters, which are used as the chemical quality deviation characteristic parameters.

8. The compound transaction fulfillment status transition method based on quality document parsing according to claim 7, characterized in that, Step S412, which involves chemically and semantically decoupling the deviation index term and decomposing the composite deviation into at least two independent deviation components, includes: Step S4121: Obtain the original spectral signal data corresponding to the deviation index item, input the original spectral signal data into the independent component analysis model, and obtain a set of independent signal components by maximizing the statistical independence of each output signal component; Step S4122: Perform feature peak identification on each of the independent signal components and extract the peak position, peak area and peak shape parameters corresponding to each of the independent signal components; Step S4123: Based on the extracted peak position, peak area and peak shape parameters, each of the independent signal components is mapped to an independent deviation component.

9. The compound transaction fulfillment status transition method based on quality document parsing according to claim 7, characterized in that, After decomposing the composite deviation into at least two independent deviation components as described in step S412, the method further includes: Step S412a: Obtain the characteristic peak parameters corresponding to each independent deviation component, and match the characteristic peak parameters with a preset impurity standard spectral library, wherein the impurity standard spectral library contains standard spectral characteristic parameters of known impurity compounds; Step S412b: When the characteristic peak parameter of any of the independent deviation components successfully matches a known impurity compound in the impurity standard spectral library, the chemical identity of the known impurity compound is obtained, and the chemical identity is assigned to the independent deviation component. Step S412c: Based on whether each independent deviation component has been assigned a chemical identity, each independent deviation component is marked, and the marking result is passed to step S413 so that the feature parameters of the independent deviation components that have been assigned a chemical identity are extracted first in step S413.

10. The compound transaction fulfillment status transition method based on quality document parsing according to claim 6, characterized in that, Step S42, which involves matching the chemical quality deviation characteristic parameters with anomaly patterns in a preset preparation process anomaly knowledge base to obtain at least one candidate anomalous process step and its matching confidence level, includes: Step S421: When there is no abnormal pattern record of the supplier for the target compound identifier in the preset preparation process abnormal knowledge base, obtain the chemical structure descriptor of the target compound identifier; Step S422: Calculate the structural similarity between the chemical structure descriptor and the compound structure descriptors associated with the existing abnormal pattern records in the preset preparation process abnormal knowledge base; Step S423: Select at least one similar compound whose structural similarity exceeds a preset similarity threshold, and extract the supplier's historical process anomaly pattern records on the similar compound; Step S424: Match the chemical quality deviation feature parameters with the extracted historical process anomaly pattern records to obtain the candidate abnormal process steps and their matching confidence levels.

11. The compound transaction fulfillment status transition method based on quality document parsing according to claim 1, characterized in that, Before triggering the preset performance state machine to perform state transition for the project performance object in step S4, the following steps are also included: Step S401: After generating the project performance object, obtain the chemical structure attributes corresponding to the target compound identifier; Step S402: Select at least one specific verification state node from a preset state node library based on the chemical structure properties. Step S403: Inject the selected specific verification state node into the preset basic performance state machine topology to generate the performance state machine corresponding to the project performance object.

12. The compound transaction fulfillment status transition method based on quality document parsing according to claim 11, characterized in that, Step S402, which involves selecting at least one specific verification state node from a preset state node library based on the chemical structure properties, includes: Step S4021: Input the chemical structure attributes into a pre-trained chemical structure-risk association model to obtain the predicted risk impurity spectrum corresponding to the target compound, wherein the chemical structure-risk association model is trained based on the correspondence between the known compound structure and its typical impurities or degradation products; Step S4022: Based on the detection requirements of each impurity or degradation product in the predicted risk impurity spectrum, select the corresponding risk verification state node from the preset state node library. Step S4023: The risk verification state node is injected into the basic performance state machine topology of step S403 as the specific verification state node.

13. The compound transaction fulfillment status transition method based on quality document parsing according to claim 11, characterized in that, Step S403, which involves injecting the selected specific verification state node into a preset basic performance state machine topology to generate the performance state machine corresponding to the project performance object, includes: Step S4031: Obtain the timing constraints and / or resource constraints of at least two of the specific verification status nodes selected in step S402. Step S4032: Detect whether there are timing conflicts or resource conflicts between each of the specific verification state nodes and between the specific verification state nodes and existing nodes in the basic performance state machine topology. Step S4033: When a conflict is detected, a preset conflict resolution strategy library is invoked to merge, split, or rearrange the specific verification state nodes according to the conflict type, so as to obtain the resolved state node set. Step S4034: Inject the parsed set of state nodes into the basic performance state machine topology to generate the performance state machine.

14. The compound transaction fulfillment status transition method based on quality document parsing according to any one of claims 1-13, characterized in that, The sources of the quality documents received in step S2 include: suppliers, third-party testing institutions, platform quality control systems, logistics monitoring systems, purchaser review systems, or experimental instrument systems.

15. The compound transaction fulfillment status transition method based on quality document parsing according to any one of claims 1-13, characterized in that, The generation of quality compliance judgment results in step S3 also includes: When the judgment result is non-compliant, the non-compliance type is determined to be missing documents or substandard quality indicators; Step S4, which describes migrating the project performance object to a document supplementation state, a quality anomaly handling state, or a deviation handling state when the quality compliance judgment result is non-compliant, includes: When the non-compliance type is "file missing", the project fulfillment object will be migrated to the "file needs to be supplemented" state. When the non-compliance type is that the quality indicator fails to meet the standard, the project performance object will be moved to the quality anomaly handling status or deviation handling status.

16. The compound transaction fulfillment status transition method based on quality document parsing according to any one of claims 1-13, characterized in that, Step S4, after migrating the project performance object to the initial screening status of quality documents upon receiving a "compliant" result, further includes: Generate a review task and send it to the purchasing party or quality control role; Upon receiving confirmation instructions from the purchasing party or quality control role regarding the review task, the project performance object is moved from the initial screening status of the quality documents to the pending shipment status or the quality confirmation completion status.

17. The compound transaction fulfillment status transition method based on quality document parsing according to any one of claims 1-13, characterized in that, After triggering the preset performance state machine to perform state transitions on the project performance object in step S4, the method further includes: Generate and store state transition audit records, which include file upload information, parsing results, threshold comparison results, file overwrite judgment results, state identifiers before and after state transition, trigger rule version, trigger time, and operation subject identifier.

18. A compound transaction fulfillment status transition system based on quality document parsing, characterized in that, The project performance data is processed based on the compound transaction performance status transition method based on quality document parsing as described in any one of claims 1-17, including: The performance object generation module is used to obtain the supplier response version confirmed by the purchaser, order data, project start data or performance start instructions, and generate project performance objects based on the obtained data; wherein, the project performance object is associated with the target compound identifier, the chemical quality threshold extracted from the obtained data and the set of requirement document types; The quality document parsing module is used to receive the quality document corresponding to the project performance object, parse the content of the quality document, extract the structural conformity test indicators, and determine the document type of the quality document; generate a quality document association key based on at least one of the following: document hash, target compound identifier, batch number, test type, supplier identifier, order identifier, response version identifier, and key quality features, and bind the quality document association key to the project performance object; The quality compliance judgment module is used to compare the measured structural compliance indicators with the chemical quality thresholds associated with the project performance object, and to determine whether the file type of the received quality documents covers the set of required file types, and to generate a quality compliance judgment result. The state machine transition execution module is used to trigger a preset performance state machine to perform state transitions on the project performance object based on the quality compliance judgment result. Specifically, when the quality compliance judgment result is compliant, the project performance object is transitioned to the quality document initial screening pass state, the state awaiting purchaser confirmation state, or the state awaiting quality control review state. When the quality compliance judgment result is non-compliant, the project performance object is transitioned to the document supplementation state, the quality anomaly handling state, or the deviation handling state.

19. An electronic device, characterized in that, include: At least one processor; And a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the compound transaction fulfillment status transition method based on quality document parsing as described in any one of claims 1-17.

20. A computer-readable storage medium, characterized in that, It stores computer instructions that, when executed by a processor, implement the compound transaction fulfillment status transition method based on quality document parsing as described in any one of claims 1-17.

21. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the compound transaction fulfillment status transition method based on quality document parsing as described in any one of claims 1-17.

Images