Method and system for tracing faults of test equipment based on multi-modal logs and knowledge base

Abstract

This application relates to a method and system for tracing the faults of laboratory equipment based on multimodal logs and a knowledge base, belonging to the field of clinical laboratory equipment fault analysis technology. It collects underlying operational logs of the laboratory equipment and quality control data from the laboratory information system, completing temporal feature alignment; constructs a hybrid retrieval knowledge base containing sparse and dense indexes; integrates data features and knowledge fragments into structured prompts with physical consistency constraints; inputs these prompts into a large medical language model to complete fault reasoning, outputs fault root causes and troubleshooting guidance, and incrementally updates the knowledge base based on user feedback. This approach breaks through data silos, enabling fault tracing even in the sub-healthy stage of equipment, improving the accuracy of fault knowledge retrieval, suppressing the illusion of large model reasoning, achieving localized self-evolution of the knowledge base, and effectively improving the fault response speed of laboratory equipment and the laboratory's autonomous maintenance capabilities.

Description

Technical Field

[0001] This application relates to the field of clinical laboratory equipment failure analysis technology, and in particular to a method and system for tracing the source of laboratory equipment failures based on multimodal logs and knowledge bases. Background Technology

[0002] Currently, clinical testing laboratories primarily rely on the traditional model of "LIS system alarms + manual manual consultation" for internal quality control and equipment maintenance. This existing technology has the following significant drawbacks: First, fault information suffers from data silos and a lack of collaborative analysis capabilities: In existing technologies, quality control data (LJ quality control charts and Westgard rule states stored in the LIS system) is completely separated from the instrument's underlying operational logs (Log files stored in the instrument's industrial control computer, containing raw sensor data such as pressure, voltage, and light intensity). When quality control exhibits "drift" or "trend changes," inspectors can only see changes in the shape of the quality control charts and cannot correlate them with subtle changes in the instrument's underlying physical parameters at the same time point. This leads to delayed fault diagnosis, and the cause usually requires a complete shutdown of the instrument to report an error before it can be located.

[0003] Second, fault attribution is highly dependent on human experience and lacks standardized tools: interpreting equipment alarm codes relies on manually flipping through hundreds of pages of maintenance manuals or the experience of senior personnel. The instrument's underlying unstructured log contains a large number of manufacturer-encrypted codes and abbreviations, which ordinary inspectors cannot interpret. Complex faults require waiting for the manufacturer's engineers to come to the site, which seriously affects the timeliness of testing.

[0004] Third, there is a lack of knowledge-based intelligent reasoning mechanisms: existing LIS systems or middleware only rely on simple "threshold rules" for alarms, cannot understand contextual semantics, do not have the ability to reason logically about multi-source heterogeneous data from large language models, and cannot infer the root cause of faults by comprehensively considering multiple independent events.

[0005] Therefore, there is an urgent need for a technology that can solve the problems of fragmented multi-source data, difficulty in utilizing underlying logs, and over-reliance on human experience in the troubleshooting of existing clinical testing equipment. Summary of the Invention

[0006] Therefore, it is necessary to provide a method and system for tracing equipment faults based on multimodal logs and knowledge bases to address the aforementioned technical problems.

[0007] Firstly, this application provides a method for tracing the source of equipment failures based on multimodal logs and a knowledge base. The method includes: Collect the underlying operation logs of the testing equipment and the quality control data of the laboratory information system, and perform time-series feature alignment on the two types of data; Construct a hybrid retrieval knowledge base that includes both sparse and dense indexes; The aligned data features are fused with the retrieved knowledge fragments to form structured prompt words; The structured prompts are input into the medical big language model to perform fault reasoning, output the root cause of the fault and troubleshooting guidance, and the knowledge base is incrementally updated based on user feedback.

[0008] Optionally, in one embodiment of this application, performing temporal feature alignment on the two types of data includes: Extract statistical features from quality control data, extract time-series data from operation logs and perform parameter cleaning and feature aggregation, and establish a correlation mapping between physical parameters of operation logs and features of quality control data.

[0009] Optionally, in one embodiment of this application, constructing a hybrid retrieval knowledge base comprising sparse and dense indexes includes: Semantic segmentation and metadata annotation are performed on the maintenance manual text. A sparse index for keyword matching and a dense index for vector similarity calculation are constructed. The two search results are then weighted, fused, and reordered.

[0010] Optionally, in one embodiment of this application, the structured prompt includes a role setting slot, a static knowledge slot, a dynamic state slot, and a task instruction slot, and injects physical consistency reasoning constraints into the prompt to suppress model reasoning illusions.

[0011] Optionally, in one embodiment of this application, the construction of a hybrid retrieval knowledge base including sparse and dense indexes further includes: Valid repair cases are encapsulated into verification cases and vectorized. Verification cases are incrementally updated to the knowledge base, and subsequent retrieval processes prioritize the use of verification cases.

[0012] Secondly, this application also provides a fault tracing system for inspection equipment based on multimodal logs and a knowledge base. This includes a multi-source data acquisition module, a dynamic knowledge base construction module, a multimodal reasoning module, and an interactive feedback module, wherein: The multi-source data acquisition module is used to connect the testing equipment and the laboratory information system, perform time-series data alignment, extract statistical features of quality control data, and clean and aggregate physical parameters of the operation log. The dynamic knowledge base construction module is used to perform semantic segmentation and metadata annotation on the maintenance manual text, and to build sparse and dense indexes to generate a hybrid retrieval index. The multimodal reasoning module is used to fuse data features and knowledge fragments, employ structured prompts and inject physical consistency constraints to complete fault attribution reasoning; The interactive feedback module is used to output troubleshooting guidance and collect user feedback, incrementally update the knowledge base with valid repair cases, and prioritize the use of verified repair cases during the retrieval process.

[0013] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the steps of the methods described in the various embodiments above.

[0014] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the methods described in the various embodiments above.

[0015] The above-mentioned method and system for tracing equipment faults based on multimodal logs and knowledge bases have the following advantages compared with existing technologies: First, break through data silos and improve the ability to detect minor faults early: By aligning multi-source heterogeneous data in time series and combining the analysis of subtle changes in the underlying Log physical parameters of the instrument with the statistical trends of the upper-level LIS quality control chart, the root cause of the fault can be identified in the "sub-health" stage of the equipment, enabling predictive maintenance and reducing unexpected downtime.

[0016] Second, it solves the problem of inaccurate RAG retrieval in vertical fields: it adopts a hybrid retrieval strategy of "sparse index + dense index" to take into account both precise matching of fault codes and fuzzy retrieval of fault semantics, which greatly improves the accuracy and precision of fault knowledge retrieval.

[0017] Third, suppress large model illusions and ensure the safety of medical applications: By using physically consistent structured prompts, the model is forced to verify the logical consistency between physical evidence in the Log and quality control statistical phenomena, thus avoiding the generation of erroneous troubleshooting suggestions that violate common sense in physics.

[0018] Fourth, achieve localized self-evolution of the knowledge base: construct a human feedback loop, automatically encapsulate successful local repair cases and incrementally update them to the knowledge base, adapt the system to the specific laboratory environment, and the longer it is used, the stronger its diagnostic capabilities become. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a method for tracing equipment faults based on multimodal logs and a knowledge base in one embodiment. Figure 2 This is a structural block diagram of a fault tracing system for inspection equipment based on multimodal logs and a knowledge base, as shown in one embodiment. Figure 3 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0021] In one embodiment, such as Figure 1 As shown, a method for tracing the source of equipment faults based on multimodal logs and knowledge bases is provided, including the following steps: S101: Collect the underlying operation logs of the testing equipment and the quality control data of the laboratory information system, and perform time sequence feature alignment on the two types of data.

[0022] In this embodiment, the system monitors the LIS quality control status and instrument alarm port in real time. When any of the following conditions are met—instrument alarm, quality control violation of Westgard rules, or unidirectional drift of continuous quality control points—a timestamp is generated. The trigger signal. Based on this, 30 days of historical quality control data were retrieved to calculate the drift value and bias statistical characteristics. The drift value is calculated by taking the slope of the most recent 5-7 data points, which quantifies the rate of change of the systematic error. The bias is calculated by taking the percentage of the relative deviation between the current test value and the target value.

[0023] In one embodiment of this application, performing temporal feature alignment on the two types of data includes: Extract statistical features from quality control data, extract time-series data from operation logs and perform parameter cleaning and feature aggregation, and establish a correlation mapping between physical parameters of operation logs and features of quality control data.

[0024] In one embodiment of this application, an event-based time window backtracking algorithm is employed to... To retrieve logs from the end point, non-critical text is filtered using regular expressions, retaining only physical parameter fields related to flow path, optics, and temperature control (such as Vacuum_Pressure, Lamp_Voltage, Bath_Temperature). The retrieved time-series data is then statistically reduced in dimensionality, and the mean, variance, and slope of change of the physical parameters are calculated to form a health status feature vector.

[0025] S102: Construct a hybrid retrieval knowledge base that includes sparse and dense indexes.

[0026] In this embodiment, an intelligent knowledge base capable of simultaneously understanding "precise error codes" and "fuzzy fault semantics" is constructed. Through a dual-path indexing mechanism, high-precision retrieval of multimodal maintenance knowledge is achieved.

[0027] In one embodiment of this application, constructing a hybrid retrieval knowledge base comprising sparse and dense indexes includes: Semantic segmentation and metadata annotation are performed on the maintenance manual text. A sparse index for keyword matching and a dense index for vector similarity calculation are constructed. The two search results are then weighted, fused, and reordered.

[0028] In one embodiment of this application, firstly, documents such as the "Maintenance Service Manual," "Fault Code Table," and "Reagent Instructions" (usually in PDF format) provided by the medical device manufacturer are imported. OCR recognition is then performed on the imported text, and semantic segmentation is performed based on chapter titles and fault code indexes. Segmentation is not only based on page numbers but also logically divided into blocks according to the document's "chapter titles" and "fault code indexes." For example, "E-2035 Vacuum Error" and its corresponding "Fault Description," "Possible Causes," and "Troubleshooting Steps" are treated as a complete contextual semantic block. Each segment is automatically labeled with metadata such as the applicable model, subsystem, and release date.

[0029] Subsequently, a sparse index was constructed based on TF-IDF / BM25, mainly used to store proper nouns and error codes such as "E-2035", "pressure sensor", and "24V voltage" to ensure accurate matching of fault codes. A dense index was constructed using a medical pre-trained embedding model, which was used to capture semantic features such as "slow pressure decrease" and "absorbance baseline drift". Even if the exact same words do not appear in the manual, relevant content can be recalled by calculating the vector distance.

[0030] Finally, keyword matching and vector similarity calculation are performed in parallel. Keyword matching is performed in the sparse index, while vector similarity calculation is performed in the dense index, retrieving the Top-K candidate slices for each. The two results are merged using a weighted linear combination or reciprocal ranking fusion algorithm. A Cross-Encoder model is then used to refine the scoring of the fused candidate list, eliminating low-relevance content. The final output is the Top-5 most relevant repair manual fragments as the inference context.

[0031] S103: Integrate the aligned data features with the retrieved knowledge fragments into structured prompt words.

[0032] In this embodiment, instead of relying on the free divergence capability of a large model, a "modular prompt word template based on slot filling" is adopted to force the model to perform reasoning under constraints.

[0033] Specifically, in one embodiment of this application, the structured prompt includes a role setting slot, a static knowledge slot, a dynamic state slot, and a task instruction slot, and physical consistency reasoning constraints are injected into the prompt to suppress model reasoning illusions.

[0034] In one embodiment of this application, a pre-defined JSON-Schema style prompt framework is provided, comprising four semantic slots: role setting slot, static knowledge slot, dynamic state slot, and task instruction slot. The role setting slot defines the model's identity as a "senior laboratory maintenance engineer with physical logic reasoning ability" and loads meta-knowledge of a specific brand of instrument. The static knowledge slot is used to fill in retrieved Top-K maintenance manual slices. The dynamic state slot is used to fill in extracted instrument Log features (mean, slope) and quality control chart trend descriptions. The task instruction slot contains thought chain guidance and output format constraints. Retrieved manual fragments are injected into the knowledge slots, and log physical parameters and quality control features are textualized and injected into the dynamic state slots. This forces the model to follow the "phenomenon-physical mechanism-root cause" reasoning path, performs reverse verification, eliminates mismatched fault hypotheses, and suppresses model illusions.

[0035] S104: Input structured prompts into the medical big language model to perform fault reasoning, output the root cause of the fault and troubleshooting guidance, and complete the incremental update of the knowledge base based on user feedback.

[0036] In this embodiment of the application, the constructed structured prompt words are input into the medical big language model MedGemma. The model performs calculations based on the preset thought chain (CoT) path and outputs the root cause of the fault, confidence level, and evidence anchor point.

[0037] Meanwhile, to lower the technical threshold for operators, the analyzed fault causes are automatically converted into "tiered troubleshooting guidelines": Action sequenced: abstract maintenance suggestions are transformed into ordered action instructions. For example, if the cause is "gas leak," the generated SOP is: Step 1. Execute the 'system cleaning' procedure; Step 2. Check if the O-ring above the sampling needle is worn; Step 3. Tighten the vacuum pump interface nut. Risk assessment is also performed: safety warnings are automatically added based on the fault type (e.g., "High-pressure components, please disconnect power first").

[0038] After the user confirms that the fault has been resolved, the system automatically encapsulates the fault feature vector and the solution as a verified case, and then incrementally stores it in the knowledge base after vectorization, so that it can be called first in subsequent searches.

[0039] In one embodiment of this application, the construction of a hybrid retrieval knowledge base comprising sparse and dense indexes further includes: Valid repair cases are encapsulated into verification cases and vectorized. Verification cases are incrementally updated to the knowledge base, and subsequent retrieval processes prioritize the use of verification cases.

[0040] In one embodiment of this application, the automatically captured fault feature vector and successful solution are encapsulated as a "Verified Case." This case, after de-identification and vectorization, is inserted into the incremental partition of the vector database. In the next retrieval, the algorithm is configured to prioritize retrieving verified cases from the local incremental partition. This means that any fault that has occurred and been resolved once will achieve 100% accurate matching the second time it is encountered, thus achieving "getting smarter with use."

[0041] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0042] Based on the same inventive concept, this application also provides a fault tracing system for inspection equipment based on multimodal logs and knowledge bases for implementing the aforementioned fault tracing method for inspection equipment based on multimodal logs and knowledge bases. The solution provided by this system is similar to the implementation scheme described in the above method. Therefore, the specific limitations of one or more embodiments of the fault tracing system for inspection equipment based on multimodal logs and knowledge bases provided below can be found in the limitations of the fault tracing method for inspection equipment based on multimodal logs and knowledge bases described above, and will not be repeated here.

[0043] In one embodiment, such as Figure 2 As shown, a fault tracing system 200 for inspection equipment based on multimodal logs and a knowledge base is provided, including: a multi-source data acquisition module 201, a dynamic knowledge base construction module 202, a multimodal reasoning module 203, and an interactive feedback module 204, wherein: The multi-source data acquisition module 201 is used to connect the testing equipment and the laboratory information system, perform time-series data alignment, extract statistical features of quality control data, and clean and aggregate physical parameters of the operation log.

[0044] The dynamic knowledge base construction module 202 is used to perform semantic segmentation and metadata annotation on the maintenance manual text, and to build sparse and dense indexes to generate a hybrid retrieval index.

[0045] The multimodal reasoning module 203 is used to fuse data features and knowledge fragments, employ structured prompts and inject physical consistency constraints to complete fault attribution reasoning.

[0046] The interactive feedback module 204 is used to output troubleshooting guidance and collect user feedback, incrementally update the knowledge base with valid repair cases, and prioritize the use of verified repair cases during the retrieval process.

[0047] In one embodiment of this application, performing temporal feature alignment on the two types of data includes: Extract statistical features from quality control data, extract time-series data from operation logs and perform parameter cleaning and feature aggregation, and establish a correlation mapping between physical parameters of operation logs and features of quality control data.

[0048] In one embodiment of this application, constructing a hybrid retrieval knowledge base comprising sparse and dense indexes includes: Semantic segmentation and metadata annotation are performed on the maintenance manual text. A sparse index for keyword matching and a dense index for vector similarity calculation are constructed. The two search results are then weighted, fused, and reordered.

[0049] In one embodiment of this application, the structured prompt includes a role setting slot, a static knowledge slot, a dynamic state slot, and a task instruction slot. Physical consistency reasoning constraints are injected into the prompt to suppress model reasoning illusions.

[0050] In one embodiment of this application, the construction of a hybrid retrieval knowledge base comprising sparse and dense indexes further includes: Valid repair cases are encapsulated into verification cases and vectorized. Verification cases are incrementally updated to the knowledge base, and subsequent retrieval processes prioritize the use of verification cases.

[0051] The modules in the aforementioned fault tracing system for inspection equipment based on multimodal logs and knowledge bases can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0052] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 3As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a fault tracing method for inspection equipment based on multimodal logs and a knowledge base. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0053] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0054] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0055] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0056] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0057] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0058] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0059] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0060] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for tracing the source of equipment failures based on multimodal logs and knowledge bases, characterized in that, The method includes: Collect the underlying operation logs of the testing equipment and the quality control data of the laboratory information system, and perform time-series feature alignment on the two types of data; Construct a hybrid retrieval knowledge base that includes both sparse and dense indexes; The aligned data features are fused with the retrieved knowledge fragments to form structured prompt words; The structured prompts are input into the medical big language model to perform fault reasoning, output the root cause of the fault and troubleshooting guidance, and the knowledge base is incrementally updated based on user feedback.

2. The method according to claim 1, characterized in that, The time-series feature alignment of the two types of data includes: Extract statistical features from quality control data, extract time-series data from operation logs and perform parameter cleaning and feature aggregation, and establish a correlation mapping between physical parameters of operation logs and features of quality control data.

3. The method according to claim 1, characterized in that, The construction of a hybrid retrieval knowledge base comprising sparse and dense indexes includes: Semantic segmentation and metadata annotation are performed on the maintenance manual text. A sparse index for keyword matching and a dense index for vector similarity calculation are constructed. The two search results are then weighted, fused, and reordered.

4. The method according to claim 1, characterized in that, The structured prompts include role setting slots, static knowledge slots, dynamic state slots, and task instruction slots. Physical consistency reasoning constraints are injected into the prompts to suppress model reasoning illusions.

5. The method according to claim 1, characterized in that, The construction of a hybrid retrieval knowledge base including sparse and dense indexes also includes: Valid repair cases are encapsulated into verification cases and vectorized. Verification cases are incrementally updated to the knowledge base, and subsequent retrieval processes prioritize the use of verification cases.

6. A fault tracing system for inspection equipment based on multimodal logs and a knowledge base, characterized in that, It includes a multi-source data acquisition module, a dynamic knowledge base construction module, a multimodal reasoning module, and an interactive feedback module, among which: The multi-source data acquisition module is used to connect the testing equipment and the laboratory information system, perform time-series data alignment, extract statistical features of quality control data, and clean and aggregate physical parameters of the operation log. The dynamic knowledge base construction module is used to perform semantic segmentation and metadata annotation on the maintenance manual text, and to build sparse and dense indexes to generate a hybrid retrieval index. The multimodal reasoning module is used to fuse data features and knowledge fragments, employ structured prompts and inject physical consistency constraints to complete fault attribution reasoning; The interactive feedback module is used to output troubleshooting guidance and collect user feedback, incrementally update the knowledge base with valid repair cases, and prioritize the use of verified repair cases during the retrieval process.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images