A medical equipment repair quotation verification method and system

Abstract

This invention provides a method and system for verifying medical equipment maintenance quotations, belonging to the field of intelligent operation and maintenance and technical auditing of medical equipment. The method includes: collecting multi-source data before the failure and screening for auditability based on anomaly degree, generating evidence atoms containing Mahalanobis distance anomalies; calculating necessity scores based on evidence atoms, equipment structure trees, and the set of competing minimum replaceable units, and performing hard constraints and verification judgments on the BOM line items of the quotation; for items requiring supplementary verification, solving for the minimum supplementary verification path based on distinguishing gain and testing cost, and calculating the marginal recovery contribution in conjunction with post-maintenance verification data, performing item-by-item causal attribution and rule updates. This invention reduces occasional noise and upper-level assembly interpretation envelope interference, ensures closed-loop verification hard constraints, suppresses over-maintenance, and reduces label contamination problems.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance and technical auditing of medical equipment, specifically to a method and system for verifying medical equipment repair quotations. Background Technology

[0002] In intelligent operation and maintenance and asset management of medical equipment, high-value imaging equipment such as MRI and CT usually have a complex hierarchical structure of maintainable components and multi-source heterogeneous operating data. Management requires both rapid response to faults and restoration of equipment operation, and precise matching between the maintenance quotation BOM and the actual root cause of the fault. However, in the actual operation and maintenance process in hospitals, there is a serious information barrier between third-party maintenance service providers and hospitals, which leads to difficulties in fault assessment and a high risk of over-maintenance. The problems in the existing technology are mainly concentrated in the following aspects: (1) Multi-source feedback merging relies heavily on the subjective experience of maintenance personnel or simple log summaries of a single device. In traditional maintenance methods, subjective or untimely records are easily affected by occasional noise interference and are very easy to lose key degradation features. This non-objective data tracing method makes it difficult for the quotation items to form a tight causal logical closed loop with the actual fault manifestation, resulting in difficulties in the ledger evidence. (2) The management method that relies solely on the static threshold alarm of a single component is difficult to cover the real fault scenarios of complex system coupling. Due to the limited data dimension and high noise in a single channel, the existing system is difficult to accurately identify the smallest replaceable unit that competes with each other. When faced with similar fault symptoms, it is easy to misjudge the root cause and cause distribution drift, and lacks the ability to make accurate judgments based on multi-source data. (3) In the hierarchical equipment maintenance, if the maintenance party is assumed to directly attribute the local fault to the high-value module above and replace it as a whole, it will trigger high over-maintenance costs; if the replacement of large components is blocked indiscriminately, the hospital cannot provide a convincing test plan for the replacement of lower-level local components. This management status quo makes the system fall into the contradiction between budget compliance constraints and equipment operation safety, which is manifested in the single and passive means of preventing over-maintenance. (4) Existing solutions generally lack a standardized closed-loop verification mechanism for multi-source evidence before the fault and causal attribution after maintenance. In a maintenance task that includes multiple replacements, some invalid maintenance actions are often uniformly marked as valid by the system, which can easily lead to label pollution in the maintenance history sample library. This continuous degradation of the underlying data seriously affects the reliability of subsequent fault mode library matching, supplementary evidence testing and quotation support judgment.

[0003] Therefore, there is an urgent need to propose a method and system for verifying medical equipment maintenance quotations. By identifying the minimum replaceable unit through competition, suppressing hierarchical replacement, solving the minimum supplementary verification path, and updating the closed loop, the objectivity and reproducibility of technical verification and the reliability of preventing over-maintenance can be improved. Summary of the Invention

[0004] The purpose of this invention is to solve the technical problems mentioned above and to propose a method for verifying medical equipment repair quotes, comprising the following steps: S1. The technical audit server receives the Bill of Materials (BOM) submitted by the maintenance quotation receiver, determines the time of occurrence of the current fault event, and extracts multi-source data from the IoT acquisition device on the equipment side and the hospital's operation and maintenance management system within a preset time window before the fault occurred. S2. The technical audit server performs preprocessing on multi-source data, forms a candidate evidence set based on the degree of anomaly, performs audit admissibility screening on the candidate evidence set, and registers audit admissibility evidence atoms. S3. The technical audit server maintains the equipment structure tree. The technical audit server performs standardization processing on each item in the quotation BOM and maps each item to the corresponding smallest replaceable unit in the equipment structure tree. When the mapping cannot be uniquely determined, a supplementary information request is output and the corresponding item is marked as a mapping uncertainty item. The mapping uncertainty item does not enter the necessity determination process before the mapping is completed. S4. The technical audit server determines the set of competing minimum replaceable units for the minimum replaceable unit corresponding to each row of items, introduces hierarchical replacement suppression rules and applies hierarchical replacement suppression to the replacement items of the superior module; the technical audit server performs hard constraint screening for each row of items to obtain the gating quantity, obtains the necessity score based on the gating quantity and the atomic of auditable evidence, and divides each row of items into items that support replacement, items that need to be supplemented, and items that do not support replacement according to the necessity score; S5. The technical audit server defines a set of candidate supplementary verification actions for the items to be verified, solves the minimum supplementary verification path that satisfies the minimum discrimination threshold constraint based on the test cost and discrimination gain of the candidate supplementary verification actions, and outputs the minimum supplementary verification path as the supplementary verification execution sequence of the items to be verified.

[0005] In the preferred embodiment, determining the fault occurrence time in step S1 includes: when the hospital's operation and maintenance management system has records of downtime, alarm time, or repair request time, the technical audit server selects the earliest and most reliable record as the fault occurrence time; when the hospital's operation and maintenance management system does not have a clear record of the fault occurrence time, the technical audit server infers the fault occurrence time based on the change point detection results of the key monitoring channel; the preset time window consists of a short window, a medium window, and a long window; when the target medical equipment has periodic operating conditions, a control window that matches the historical normal cycle is extracted at the same time, and an individualized health baseline is constructed based on the control window.

[0006] In the preferred scheme, the preprocessing in step S2 includes time alignment, unit normalization, missing value imputation, abnormal noise removal, and channel identifier standardization; when multi-source data does not use a unified clock stamp, time axis reconstruction is performed based on the acquisition gateway timestamp, message sequence number, system log, and co-occurring events; anomaly degree includes the Mahalanobis distance anomaly degree of the feature vector relative to the healthy distribution; auditable admissible evidence atoms include occurrence time interval, evidence source, feature type, anomaly degree, credibility, and original data index; when the same physical phenomenon is repeatedly observed by multiple acquisition sources, multiple auditable admissible evidence atoms are merged to obtain a composite auditable admissible evidence atom.

[0007] In the preferred embodiment, the standardization process in step S3 includes deambiguation of project names, merging of component codes, replacement of aliases, identification of model substitution relationships, and identification of maintenance actions. Maintenance actions include overall replacement, board-level replacement, partial repair, calibration, cleaning, and tightening. The output of supplementary information requests includes requests for supplementary codes, pictures, or maintenance manual page numbers.

[0008] In the preferred embodiment, the competing minimum replaceable unit set in step S4 satisfies the following conditions: the competing minimum replaceable unit can explain at least some of the key symptoms in the current fault manifestation; the competing minimum replaceable unit is located in the same fault propagation chain or an adjacent fault propagation chain as the minimum replaceable unit; and the audit-admissible evidence atom set cannot sufficiently distinguish the competing minimum replaceable unit from the minimum replaceable unit. The hierarchical replacement suppression rule in step S4 takes effect when the lower-level minimum replaceable unit set can already explain the current fault manifestation and there is no cross-level propagation evidence in the audit-admissible evidence atom set. Cross-level propagation evidence includes upper-level power supply chain anomalies and upper-level control... Chain anomalies, upper-level thermal field diffusion anomalies, and multi-channel synchronization mismatch anomalies; hard constraint screening in step S4 includes: the project has been uniquely mapped to the smallest replaceable unit, there are leading evidence atoms in the auditable evidence atom set that have a physical relationship with the smallest replaceable unit, there is a verifiable structural propagation relationship between the smallest replaceable unit and the current fault manifestation, the hierarchical replacement inhibition rule has not excluded the project, and there is no strong conflicting evidence in the auditable evidence atom set that directly excludes the fault of the smallest replaceable unit; the necessity score is obtained based on a combination of gating quantity, posterior fault probability, structural dependency score, historical consistency score, conflict coefficient, and hierarchical replacement inhibition coefficient.

[0009] In the preferred embodiment, after the technical audit server outputs the supplementary verification execution sequence in step S5, the following steps are also included: After the maintenance service provider completes the maintenance, the technical audit server collects post-maintenance verification data, which includes power-on self-test results, fault disappearance status, key operating parameter recovery status, waveform or timing feature recovery status, and recurrence status within a preset observation period; the technical audit server performs item-by-item causal attribution for multiple actual replacement items, calculates marginal recovery contribution and obtains attribution weights, updates the effective sample library of maintenance history based on the attribution weights, and updates the fault mode library, the set of competing minimum replaceable units, the minimum supplementary verification path library, the set of conflicting evidence, and the conflict coefficient based on the supplementary verification results, the post-maintenance verification data, and the item-by-item causal attribution results.

[0010] This invention also provides a medical equipment repair quotation verification system, comprising: The fault window and multi-source data module receive the Bill of Materials (BOM) submitted by the maintenance quotation receiver, determine the time of the fault occurrence, and extract multi-source data within a preset time window before the fault occurs. The evidence atom generation module performs preprocessing on multi-source data, forms a candidate evidence set based on the degree of anomaly, performs audit-admissibility screening on the candidate evidence set, and registers audit-admissible evidence atoms. The quotation BOM mapping module maintains the equipment structure tree, performs standardization processing on each item in the quotation BOM and maps it to the smallest replaceable unit. When the mapping cannot be uniquely determined, it outputs a supplementary information request and marks the corresponding item as a mapping uncertainty item. The verification and judgment module determines the minimum set of competitive substitutable units, introduces hierarchical substitution suppression rules and applies hierarchical substitution suppression to the replacement items of the superior module, performs hard constraint screening to obtain the gate quantity, obtains the necessity score based on the gate quantity and the atomic evidence that can be accepted by the audit, and outputs the items that support the replacement, the items that need to be supplemented, and the items that do not support the replacement. The minimum supplementary evidence path solution module defines a set of candidate supplementary evidence actions for the item to be supplemented, solves the minimum supplementary evidence path that satisfies the minimum discrimination threshold constraint based on the test cost and discrimination gain, and outputs the supplementary evidence execution sequence.

[0011] In the preferred embodiment, the audit-admissible screening performed by the evidence atom generation module includes checking data integrity, sampling continuity, clock consistency, reliability of the collection source, and traceability of the original record. The audit-admissible evidence atom includes the time interval of occurrence, evidence source, feature type, anomaly degree, credibility, and original data index. When the same physical phenomenon is repeatedly observed by multiple collection sources, the evidence atom generation module merges multiple audit-admissible evidence atoms to obtain a composite audit-admissible evidence atom.

[0012] In the preferred embodiment, the minimum supplementary evidence path solving module sets a test cost for the candidate supplementary evidence actions. The minimum supplementary evidence path solving module selects the combination of supplementary evidence actions with the minimum total test cost and that satisfies the minimum discrimination threshold constraint from the set of candidate supplementary evidence actions as the minimum supplementary evidence path, and calculates the discrimination gain based on the cumulative Kullback-Leibler divergence between the minimum replaceable unit and the competing minimum replaceable unit set.

[0013] In the preferred embodiment, the module further includes a post-repair causal attribution and sample and rule update module. This module collects post-repair verification data and calculates the marginal recovery contribution and attribution weight for multiple actual replacement items. The module also updates the effective sample library of repair history based on the attribution weight, and updates the fault mode library, the set of competing minimum replaceable units, the minimum supplementary evidence path library, the set of conflict evidence, and the conflict coefficient.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Precise source tracing and interference resistance enhance evidence reliability: This invention extracts multi-source data within a preset time window before the fault occurs, forms a candidate evidence set based on anomaly degree, performs audit-admissible screening, and finally registers audit-admissible evidence atoms. This mechanism effectively reduces the interference of sporadic noise, sampling gaps, and weakly reliable sources on subsequent quotation verification; simultaneously, by calculating the posterior fault probability, heterogeneous evidence such as discrete fault codes, continuous waveform features, temperature, vibration, and pressure are uniformly converted into the same posterior probability space, laying an objective verification foundation.

[0015] 2. Refined Identification and Suppression Limitations to Prevent Over-Repair: This invention identifies the minimum replaceable unit for row item mapping, determines the set of competing minimum replaceable units, and introduces hierarchical replacement suppression rules to suppress replacement items in higher-level modules. This mechanism effectively reduces the situation where higher-value assemblies naturally dominate under conditions of larger explanatory envelopes; simultaneously, by combining counterfactual substitution verification and hard constraint screening to calculate necessity scores, it further suppresses over-repair from the system mechanism level by strictly determining items that support replacement, require supplementary verification, and do not support replacement.

[0016] 3. Low-cost optimization and accurate supplementary verification sequence output: For the item to be supplemented, this invention solves for the minimum supplementary verification path based on the discrimination gain and test cost from the set of candidate supplementary verification actions. This mechanism can select the set of supplementary verification actions with the minimum test cost under the condition of satisfying the minimum discrimination threshold constraint, and output a clear supplementary verification execution sequence, avoiding blind screening and invalid testing.

[0017] 4. Objective Attribution and Rule Adaptation for Closed-Loop Verification: This invention collects post-repair verification data after maintenance, calculates the marginal recovery contribution and attribution weight of each replacement item using the predicted residual failure loss value, and performs item-by-item causal attribution after maintenance. This mechanism effectively reduces the label contamination problem of all replacement items being uniformly marked as valid after a single multi-replacement maintenance; simultaneously, it continuously updates the historical valid sample library and various verification rules based on supplementary verification results, verification data, and attribution weights, forming a data-driven closed-loop correction. Attached Figure Description

[0018] Figure 1 This is a flowchart of the method.

[0019] Figure 2 This is a system architecture diagram. Detailed Implementation

[0020] Example 1 like Figure 1 As shown, this embodiment of the invention provides a method for verifying medical equipment maintenance quotations based on multi-source evidence before failure, identification of the least replaceable unit in competition, hierarchical substitution suppression, solution of the least supplementary evidence path, and item-by-item causal attribution after maintenance. A technical audit server is deployed on the hospital side. The technical audit server is communicatively connected to the equipment-side IoT acquisition device, the hospital's operation and maintenance management system, the maintenance work order system, and the maintenance quotation receiving end. The equipment-side IoT acquisition device continuously collects the operating parameters, sensor time-series data, waveform data, and log data of the target medical equipment. The hospital's operation and maintenance management system provides historical maintenance records, parts replacement records, equipment history, and power-on / off records. The maintenance quotation receiving end receives the Bill of Materials (BOM) submitted by the maintenance service provider. The target medical equipment includes MRI equipment, CT equipment, anesthesia machines, monitors, and other medical equipment with a maintainable component hierarchy. Preferably, the target medical equipment is a high-value imaging device with a cooling circuit, electrical control module, and actuator. The technical audit server establishes a closed-loop technical verification system around each item in the Bill of Materials (BOM) that includes multi-source evidence before failure, equipment structure tree, set of competing minimum replaceable units, hierarchical replacement suppression rules, minimum supplementary evidence path, and post-repair causal attribution.

[0021] After receiving the Bill of Materials (BOM) for the target medical equipment, the technical audit server determines the time of occurrence of the current fault event. When the hospital's operation and maintenance management system has records of downtime, alarm times, or repair requests, the technical audit server selects the earliest and most reliable record as the time of the fault occurrence. When the hospital's internal operation and maintenance management system does not have a clear record of the time of the fault occurrence, the technical audit server infers the time of the fault occurrence based on the change point detection results of key monitoring channels. When the fault occurred Once identified, the technical audit server extracts multi-source data from the IoT acquisition devices on the equipment side and the hospital's operation and maintenance management system within a preset time window before the fault occurred. The preset time window consists of short, medium, and long windows. The short window corresponds to the extraction of sudden anomalies, the medium window corresponds to the extraction of fault evolution trajectories, and the long window corresponds to the extraction of slow degradation trends. When the target medical equipment has periodic operating conditions, the technical audit server simultaneously extracts a control window that matches the historical normal cycle and constructs an individualized health baseline based on the control window.

[0022] The technical audit server performs time alignment, unit normalization, missing value imputation, anomaly noise removal, and channel identifier standardization on multi-source data. When multi-source data uses a unified clock stamp, the technical audit server directly performs time alignment. When multi-source data does not use a unified clock stamp, the technical audit server performs timeline reconstruction based on the acquisition gateway timestamp, message sequence number, and co-occurring events in the system log. After preprocessing, the technical audit server constructs a health baseline for each monitoring channel. The health baseline is derived from normal samples from the same model of equipment and historical stable operating samples from the current target medical equipment. Continuous feature channels use feature mean vectors and covariance matrices to represent the health distribution. The feature vector extracted at the current time... Corresponding anomaly As shown in equation (1): (1) in, This represents the feature mean vector of the monitoring channel in a healthy state. This represents the covariance matrix of the monitoring channel under healthy conditions. This represents the Mahalanobis distance anomaly degree of the current feature vector relative to the healthy distribution. Equation (1) gives the comprehensive anomaly strength under the condition that multiple features are correlated.

[0023] Technical audit server based on anomaly level A candidate evidence set is formed. Candidate evidence fragments satisfy the condition that the anomaly level consistently exceeds the corresponding channel threshold and precedes the fault event in time. The technical audit server further performs audit-admissibility screening on the candidate evidence set. Audit-admissibility screening checks data integrity, sampling continuity, clock consistency, acquisition source reliability, and original record traceability. After candidate evidence passes audit-admissibility screening, the technical audit server registers audit-admissible evidence atoms. Each audit-admissible evidence atom includes the occurrence time interval, evidence source, feature type, anomaly level, credibility, and original data index. Feature types include vibration frequency band energy anomalies, temperature rise slope anomalies, pressure decay anomalies, current ripple anomalies, fault code co-occurrence anomalies, self-test failure item anomalies, and operating load deviation anomalies. When the same physical phenomenon is repeatedly observed by multiple acquisition sources, the technical audit server merges multiple audit-admissible evidence atoms into a composite audit-admissible evidence atom. The audit-admissibility screening mechanism reduces the interference of occasional noise, sampling gaps, and weakly reliable sources on subsequent quotation verification.

[0024] The technical audit server pre-maintains the equipment structure tree for the target medical device. The equipment structure tree includes a complete device layer, a subsystem layer, a functional module layer, and a minimum replaceable unit layer. The minimum replaceable unit is an independent replacement object in the current model's maintenance specifications, and it establishes a one-to-one or many-to-one correspondence with specific items in the Bill of Materials (BOM). Minimum replaceable units include boards, pump bodies, valve assemblies, sensor assemblies, connection assemblies, and local piping assemblies. After receiving the BOM, the technical audit server performs standardization processing on each line of the BOM. Standardization processing includes deambiguity of item names, merging of part codes, alias replacement, identification of model substitution relationships, and identification of maintenance actions. Maintenance actions include overall replacement, board-level replacement, local repair, calibration, cleaning, and tightening. After standardization processing is complete, the technical audit server maps each line of the BOM to the corresponding minimum replaceable unit in the equipment structure tree. When an item in the Bill of Materials (BOM) cannot be uniquely mapped to the smallest replaceable unit, the technical audit server marks the corresponding item as a mapping uncertainty and triggers a request process for supplementary code, image, or service manual page number. Mapping uncertainty items do not enter the necessity determination process until mapping is completed.

[0025] The technical audit server further establishes a failure mode library. Each failure mode is associated with a set of supporting evidence atomic templates, a set of conflicting evidence, a set of candidate minimum replaceable units, and a set of supplementary evidence actions. The technical audit server addresses each minimum replaceable unit... Record prior fault probability It supports evidence distribution and structural dependencies. Structural dependencies reflect the propagation impact of the current smallest replaceable unit failure on higher-level modules, adjacent units, and associated monitoring channels. Given the set of auditable evidence atoms corresponding to the current failure... At that time, the smallest replaceable unit The posterior failure probability of the root cause unit is shown in equation (2): (2) in, This represents the set of candidate minimum replaceable units corresponding to the current fault. Indicates the first A single auditable piece of evidence. Indicates the first The weight of each auditable evidence atom. Represents the smallest replaceable unit When a failure occurs, credible evidence atoms for auditing are observed. The conditional probability. Equation (2) unifies heterogeneous evidence such as discrete fault codes, continuous waveform features, temperature, vibration and pressure into the same posterior probability space.

[0026] The technical audit server targets the first item in the quotation BOM. Determine the minimum replaceable unit set for each project. . No. Line items are mapped to the smallest replaceable unit. Subsequently, the technical audit server selects the candidate minimum replaceable unit set. The process involves selecting the minimum replaceable unit from the competing elements. A minimum replaceable unit must satisfy three conditions: it must be able to explain at least a portion of the key symptoms in the current fault manifestation; it must be located near the minimum replaceable unit. The same or adjacent failure propagation chains; the existing set of admissible evidence atoms in the audit. Unable to match the competing minimum substitutable unit with the minimum substitutable unit Fully differentiate. Compete for the minimum replaceable unit set. After establishment, the technical audit server will simultaneously determine two issues when performing quotation verification: the smallest replaceable unit. Is there a possibility of failure, and what is the set of atomic evidence that can be accepted by the existing audit? Is it sufficient to replace the smallest unit? Competition for the smallest replaceable unit set distinguish.

[0027] The technical audit server also introduces hierarchical replacement suppression rules. These rules apply to replacement items in the higher-level modules of the Bill of Materials (BOM). The minimum replaceable unit set at the lower level is sufficient to fully explain the current fault symptoms, and the set of atomic evidence that can be accepted by the audit is also considered valid. When no evidence of cross-layer propagation exists, the technical audit server applies hierarchical substitution suppression to the replacement items of the higher-level modules. Evidence of cross-layer propagation includes anomalies in the higher-level power supply chain, control chain, thermal field diffusion, and multi-channel synchronization mismatch. When the hierarchical substitution suppression rule is in effect, replacement items of higher-level modules cannot directly enter the supported replacement items; they can only enter the items awaiting supplementary verification or the unsupported replacement items. The hierarchical substitution suppression rule reduces the situation where higher-value assemblies naturally have an advantage under conditions of a larger explanatory envelope.

[0028] The technical audit server establishes a valid sample library of maintenance history. Samples in this library have undergone equipment self-testing, operational recovery testing, performance verification testing, or equivalent post-testing, and the sample results demonstrate that a particular replacement action effectively eliminated the fault or restored critical performance. This constitutes the atomic set of auditable evidence corresponding to the current fault. Compared with historical samples Corresponding set of evidence The similarity between them is used to calculate the smallest replaceable unit. Historical consistency score Historical consistency score As shown in equation (3): (3) in, This indicates the smallest replaceable unit in the valid sample library of maintenance history. A relevant and posteriorly validated set of samples. This function represents the similarity between the current set of evidence and the historical set of sample evidence. Representing historical samples Replace the smallest replaceable unit The tag indicating whether the device's functionality has been effectively restored afterwards. The value can be either 1 or 0. Historical consistency score. It reflects the degree of consistency between the current failure evidence pattern and existing valid maintenance samples.

[0029] Before performing a necessity determination, the technical audit server performs hard constraint screening. Hard constraint screening includes the following conditions: the first one in the quotation BOM. The row item has been uniquely mapped to the smallest replaceable unit. Atom set of admissible evidence in audit There exists at least one unit that is the smallest replaceable unit. Atoms with physical correlation as leading evidence; minimal substitutable unit A verifiable structural propagation relationship exists between the current fault manifestation and the hierarchical replacement suppression rule; the smallest replaceable unit is not included. Exclusion of the relevant project; Atom set of admissible evidence in audit There is no direct exclusion of the smallest replaceable unit. Strongly contradictory evidence of the fault. The technical audit server uses gating. This indicates the results of hard constraint filtering. The first item in the quotation BOM... When the project satisfies all hard constraints, The first item in the quotation BOM When a project does not meet any hard constraints, .

[0030] After the hard constraint screening is completed, the technical audit server calculates the first item in the quotation BOM. Necessity score for each line item Necessity score As shown in equation (4): (4) in, Represents the Logistic mapping function. The posterior failure probability calculated by expression (2) Indicates the structural dependency score. The historical consistency score is calculated using expression (3). Indicates the conflict coefficient. Indicates the hierarchical substitution inhibition coefficient. , , , and This represents the weight coefficients determined through training or calibration. (Structure dependency score) Reflecting the smallest replaceable unit The strength of causal support for the current fault symptoms. Conflict coefficient. Reflecting the impact of reverse evidence, lifespan discrepancies, and maintenance record discrepancies on the first The weakening effect of line items. Hierarchical replacement inhibition coefficient. Reflects the smallest replaceable unit or local repair action replacing the first The feasibility of the project. The technical audit server scores based on necessity. The first item in the quotation BOM The items are divided into those that can be replaced, those that need to be replaced, and those that cannot be replaced.

[0031] The first item in the quotation BOM When an item is determined to be an item requiring supplementary verification, the technical audit server targets the smallest replaceable unit. and the set of minimum substitutable units Find the minimum path for supplementary evidence. Technical audit server definition and the first... Set of candidate supplementary certification actions related to the project Candidate supplementary certification action set The actions include local isolation testing, directional sensor re-sampling, self-test sub-process execution, power-on waveform sampling, valve pre- and post-valve differential pressure measurement, replacement board insertion / removal verification, and target interface thermal imaging acquisition. The technical audit server performs each candidate re-certification action. Set test cost And calculate candidate supplementary evidence actions. In the smallest replaceable unit Competition for the smallest replaceable unit The distinguishing gain generated between them. Minimum proof path. As shown in equation (5): (5) in, Indicates the first The minimum threshold that must be met to determine whether a project has been completed and is eligible for approval. Represents the set of supplementary evidence actions Relative to the set of competing minimum replaceable units The minimum distinguishing gain produced. As shown in equation (6): (6) in, Indicates the execution of supplementary certification procedures. The subsequent observation results, This represents the Kullback-Leibler divergence. The solutions corresponding to equations (5) and (6) are in the candidate complement action set. The set of supplementary verification actions with the minimum total test cost and satisfying the discrimination threshold constraint is selected. Minimum Supplementary Verification Path Once formed, the technical audit server will implement the minimum necessary verification path. As the first Output the supplementary certificate execution sequence for the line project.

[0032] The technical audit server also performs counterfactual substitution verification. Counterfactual substitution verification applies to the higher-level module replacement items in the Bill of Materials (BOM). The technical audit server constructs an alternative repair set for each higher-level module replacement item. This alternative repair set consists of repair actions for the smallest replaceable unit, partial repair actions, and parameter recalibration actions. The technical audit server evaluates the explanatory power of the alternative repair set for the current fault symptoms and its predictive power for critical performance recovery, provided that the higher-level module replacement item is not executed. If the alternative repair set is sufficient to explain the current fault symptoms and predicts that critical performance can be restored to a threshold range, the technical audit server increases the hierarchical substitution inhibition coefficient corresponding to the higher-level module replacement item. When the alternative repair set cannot explain the current fault symptoms or restore critical performance, the technical audit server maintains the current level of replacement inhibition coefficient. Counterfactual surrogate verification further curbs over-maintenance.

[0033] After the repair service provider completes the repair, the technical audit server continues to collect post-repair verification data. This data includes power-on self-test results, fault disappearance status, key operating parameter recovery status, waveform or timing characteristic recovery status, and recurrence status within a preset observation period. When the Bill of Materials (BOM) contains multiple actual replacement items, the technical audit server performs a causal attribution for each item. The technical audit server defines the set of actual replacement items for this specific instance. Define the post-repair verification dataset Define the residual fault loss function . No. Each replacement item corresponds to the smallest replaceable unit. Marginal recovery contribution As shown in equation (7): (7) in, This means keeping other replacement items unchanged and removing the replacement items. Predicted residual failure loss at that time Represents the actual set of replacement items Corresponding residual failure loss prediction. Marginal recovery contribution. When the value is large, replace the item. It makes a high contribution to fault elimination and performance recovery; marginal recovery contribution When the size is small, replace the item. Its contribution to the success of this repair was relatively low. Attribution weight. As shown in equation (8): (8) in, Indicates replacement item The relative contribution percentage to the success of this repair. The technical audit server assigns this based on attribution weights. Update the valid sample database of maintenance history. Attribution weights. When the high contribution threshold is exceeded, the technical audit server will replace the item. Write positive samples; attribution weights When the contribution level falls below the low contribution threshold, the technical audit server will replace the item. Write inefficient or invalid samples; attribution weights When the data is in the middle range, the technical audit server will replace the item. Write weak positive samples or samples to be confirmed. The item-by-item causal attribution mechanism reduces the problem of all replaced items being uniformly marked as valid labels after a single multi-item replacement or repair is completed.

[0034] The technical audit server continuously updates the failure mode library, the set of competing minimum replaceable units, the minimum supplementary evidence path library, the set of conflicting evidence, and the library of valid samples from maintenance history based on the supplementary evidence results, post-repair verification data, and item-by-item causal attribution results. If a conclusion is reversed after supplementary evidence is obtained for an item pending supplementary verification, the technical audit server updates the corresponding evidence template and the set of competing minimum replaceable units. and minimum proof path When a replacement item in a higher-level module that was rejected due to hierarchical replacement suppression subsequently generates cross-level propagation evidence, the technical audit server increases the weight of the corresponding cross-level propagation evidence template. If the post-repair causal attribution results show that a particular replacement item has a low contribution, the technical audit server increases the conflict coefficient of the corresponding replacement item in subsequent similar fault scenarios. The technical audit server thus forms a closed-loop processing flow, including multi-source data extraction before failure, atomic screening of auditable evidence, BOM mapping of quotation, identification of the smallest replaceable unit in competition, screening out hard constraints, necessity scoring, solving the minimum supplementary evidence path, and post-repair causal attribution—sample and rule updating.

[0035] In a scenario of an abnormal cooling circuit in a certain magnetic resonance imaging (MRI) device, the hospital-side technical audit server received a Bill of Materials (BOM). The BOM included four items: "Replacement of the cold head assembly," "Replacement of the cooling pump assembly," "Replacement of the main control board," and "Replacement of local piping components." The technical audit server used vibration, pressure, temperature, current ripple, alarm logs, and previous maintenance records from the seven days prior to the repair request as input for analysis, and extracted three auditable evidence atoms based on Equation (1): abnormal vibration frequency band of the cooling pump, slow attenuation of cooling circuit pressure, and accelerated local temperature rise. The technical audit server also confirmed that no abnormal power supply, self-test failure, or control register abnormality of the main control board occurred. Based on Equation (2), the technical audit server calculated that the posterior failure probability of the minimum replaceable unit corresponding to the cooling pump assembly was the highest, and based on Equation (3), it found that the current evidence pattern was highly similar to the historical valid sample of "cooling pump degradation leading to pressure attenuation." In the current scenario, the technical audit server determined that the local valve group and local piping components belonged to the set of competing minimum replaceable units corresponding to the cooling pump assembly. The technical audit server continues to determine the replacement item of the upper-level module corresponding to "cold head assembly replacement". The set of the smallest replaceable units at the lower level is sufficient to explain the current fault symptoms, and there is no evidence of cross-layer propagation in the current scenario. Therefore, the technical audit server applies hierarchical replacement suppression to "cold head assembly replacement". The technical audit server performs minimum verification path solving for "local pipeline component replacement", and finds the minimum verification path among the three candidate verification actions: "sampling the pressure difference curve before and after the valve", "performing a local isolation test", and "collecting the flow pulsation waveform after the pump". The technical audit server ultimately outputs a supplementary evidence execution sequence of "first perform local isolation testing, then re-sample the pressure difference curve before and after the valve." When performing hard constraint screening for the "main control board replacement" item, the technical audit server discovered the set of atomic evidence that could be accepted by the audit. There are no leading evidence atoms physically associated with the main control board, and no verifiable structural propagation relationship exists. Therefore, the technical audit server directly determines "main control board replacement" as an unsupported replacement item. After the repair service provider only replaces the coolant pump assembly and completes local cleaning, the current fault disappears. The technical audit server calculates, based on equations (7) and (8), that the marginal recovery contribution and attribution weight of the replacement item corresponding to the coolant pump assembly are significantly higher than those of the local cleaning action. Based on this, the technical audit server writes "coolant pump assembly replacement" into the positive sample, writes "main control board replacement is unnecessary" into the conflict sample library, and sets the minimum supplementary evidence path corresponding to the current scenario. Write the minimum replacement path library for the corresponding fault mode.

[0036] Example 2 like Figure 2 As shown, the medical equipment maintenance quotation verification system is deployed on the hospital side. The system includes a technical audit server, which communicates with the equipment-side IoT data acquisition device, the hospital's operation and maintenance management system, the maintenance work order system, and the maintenance quotation receiving terminal. The equipment-side IoT data acquisition device continuously collects the operating parameters, sensor time-series data, waveform data, and log data of the target medical equipment. The hospital's operation and maintenance management system provides historical maintenance records, parts replacement records, equipment history, and power-on / off records. The maintenance quotation receiving terminal receives the Bill of Materials (BOM) submitted by the maintenance service provider. The target medical equipment includes MRI equipment, CT equipment, anesthesia machines, monitors, and medical equipment with a maintainable component hierarchy. The technical audit server includes a fault window and multi-source data module, an evidence atom generation module, a BOM mapping module, a verification judgment module, a minimum supplementary evidence path solution module, and a post-maintenance causal attribution and sample and rule update module.

[0037] The fault window and multi-source data module include: The quotation BOM receiving unit is used to receive the quotation BOM from the maintenance quotation receiving end and output the quotation BOM to the fault occurrence time determination unit and the multi-source data extraction unit before the fault.

[0038] The fault occurrence time determination unit is used to obtain downtime records, alarm times, and repair request times from the hospital's operation and maintenance management system, and output the fault occurrence time based on the earliest and most reliable record. When the hospital's operation and maintenance management system does not have downtime records, alarm times, and repair request times, the fault occurrence time determination unit is used to perform change point detection on key monitoring channels and output the fault occurrence time based on the change point detection results.

[0039] The pre-fault multi-source data extraction unit is used to extract multi-source data from the IoT acquisition device on the equipment side and the hospital operation and maintenance management system within a preset time window before the fault occurs, and output the multi-source data to the multi-source data preprocessing unit. The preset time window consists of a short window, a medium window, and a long window. The short window corresponds to the extraction of sudden anomalies, the medium window corresponds to the extraction of fault evolution trajectory, and the long window corresponds to the extraction of slow degradation trend.

[0040] The control window extraction and individualized health baseline triggering unit is used to extract a control window that matches the historical normal cycle when the target medical device has periodic operating conditions and output the control window to the health baseline construction unit.

[0041] The evidence atom generation module includes: The multi-source data preprocessing unit performs time alignment, unit normalization, missing value imputation, abnormal noise removal, and channel identifier standardization on multi-source data, and outputs the preprocessed multi-source data to the anomaly calculation unit. If the IoT acquisition device on the equipment side and the hospital's operation and maintenance management system use a unified clock stamp, the multi-source data preprocessing unit directly performs time alignment. If the IoT acquisition device on the equipment side and the hospital's operation and maintenance management system do not use a unified clock stamp, the multi-source data preprocessing unit performs timeline reconstruction and completes time alignment based on the acquisition gateway timestamp, message sequence number, and co-occurring events in the system log.

[0042] The health baseline construction unit is used to construct the health baseline for each monitoring channel based on normal samples of the same model of equipment and historical stable operation samples of the target medical equipment, and output the health baseline to the anomaly calculation unit. The health baseline of the continuous feature channel is characterized by the feature mean vector and covariance matrix to represent the health distribution.

[0043] The anomaly calculation unit is used to extract the feature vector of the current time from the preprocessed multi-source data and calculate the Mahalanobis distance anomaly output based on the healthy baseline. The Mahalanobis distance anomaly is shown in Equation (1): (1) The candidate evidence set generation unit is used to generate a candidate evidence set output based on the Mahalanobis distance anomaly. The candidate evidence set satisfies the constraint that the Mahalanobis distance anomaly continuously exceeds the monitoring channel threshold and precedes the fault event in time.

[0044] The audit-admissibility screening unit is used to check the data integrity, sampling continuity, clock consistency, source reliability, and original record traceability of the candidate evidence set, and outputs the candidate evidence that passes the audit-admissibility screening to the evidence atom registration unit.

[0045] The evidence atom registration unit is used to register audit-admissible evidence atoms and output the audit-admissible evidence atom set to the posterior fault probability calculation unit, the competition minimum replaceable unit set determination unit, and the hard constraint screening unit. Audit-admissible evidence atoms include occurrence time interval, evidence source, feature type, anomaly degree, credibility, and original data index. Feature types include vibration frequency band energy anomaly, temperature rise slope anomaly, pressure decay anomaly, current ripple anomaly, fault code co-occurrence anomaly, self-test failure item anomaly, and operating load deviation anomaly.

[0046] The composite auditable evidence atom merging unit is used to merge multiple auditable evidence atoms when the same physical phenomenon is repeatedly observed by multiple acquisition sources to obtain composite auditable evidence atoms and output the composite auditable evidence atoms to the auditable evidence atom set.

[0047] The quotation BOM mapping module includes: The equipment structure tree maintenance unit is used to maintain the equipment structure tree of the target medical equipment and output the equipment structure tree to the minimum replaceable unit mapping unit. The equipment structure tree includes the whole machine layer, subsystem layer, functional module layer, and minimum replaceable unit layer. The minimum replaceable unit is an independent replacement object in the maintenance specification and establishes a one-to-one or many-to-one correspondence with the items in the quotation BOM. The minimum replaceable unit includes boards, pump bodies, valve groups, sensor assemblies, connection assemblies, and local pipeline assemblies.

[0048] The quotation BOM line-by-line standardization processing unit is used to perform item name deambiguation, component code merging, alias replacement, model substitution relationship identification, and maintenance action identification on each line of the quotation BOM, and output the standardized line item to the smallest replaceable unit mapping unit. Maintenance actions include overall replacement, board-level replacement, partial repair, calibration, cleaning, and tightening.

[0049] The minimum replaceable unit mapping unit is used to map standardized row items to the corresponding minimum replaceable units in the device structure tree and output the mapping results to the competing minimum replaceable unit set determination unit and the hard constraint elimination unit. When the mapping results cannot be uniquely determined, the minimum replaceable unit mapping unit is used to output the mapping uncertainty item marker to the mapping uncertainty item management unit.

[0050] The Mapping Uncertainty Management Unit is used to output a supplementary information request to the maintenance service provider and mark the line item as a mapping uncertainty when the mapping result cannot be uniquely determined. The supplementary information request includes supplementary codes, pictures, and maintenance manual page numbers. The Mapping Uncertainty Management Unit is used to block the mapping uncertainty from entering the necessity determination process output before the mapping uncertainty is completed.

[0051] The verification and judgment module includes: The fault mode library maintenance unit is used to maintain the fault mode library and output the fault mode library to the posterior fault probability calculation unit and the minimum supplementary evidence path solution module. The fault mode association supports the evidence atom template set, conflict evidence set, candidate minimum replaceable unit set, and supplementary evidence action set. For each minimum replaceable unit, the fault mode library records the prior fault probability, supported evidence distribution, and structural dependency relationship. The posterior fault probability calculation unit is used to calculate the posterior fault probability output of each minimum replaceable unit as the root cause unit based on the prior fault probability and the audit-admissible evidence atom set. The posterior fault probability is shown in Equation (2): (2) The competitive minimum replaceable unit set determination unit is used to filter the competitive minimum replaceable unit set from the candidate minimum replaceable unit set for the minimum replaceable unit obtained by mapping the minimum replaceable unit of the i-th row of the quotation BOM, and output the competitive minimum replaceable unit set to the distinguishing gain calculation unit and the necessity score calculation unit. The competitive minimum replaceable unit set satisfies the constraint that the competitive minimum replaceable unit can explain the key symptoms in the failure manifestation, the constraint that the competitive minimum replaceable unit is located in the same failure propagation chain or adjacent failure propagation chain as the minimum replaceable unit, and the constraint that the competitive minimum replaceable unit set cannot be sufficiently distinguished from the minimum replaceable unit by the set of atomic evidence admissible by audit.

[0052] The hierarchical replacement suppression determination unit is used to check the completeness of the explanation of the fault manifestation by the set of the smallest replaceable units for the upper-level module replacement item in the quotation BOM, and to check the cross-level propagation evidence in the set of auditable evidence atoms. The hierarchical replacement suppression coefficient is output to the necessity score calculation unit. Hierarchical replacement suppression is triggered when the set of the smallest replaceable units for the upper-level module replacement item can completely explain the fault manifestation and there is no cross-level propagation evidence in the set of auditable evidence atoms. Cross-level propagation evidence includes upper-level power supply chain anomaly, upper-level control chain anomaly, upper-level thermal field diffusion anomaly, and multi-channel synchronization mismatch anomaly.

[0053] The counterfactual substitution verification unit is used to construct an alternative maintenance set for the replacement item of the upper-level module, consisting of the repair action of the smallest replaceable unit at the lower level, the local repair action, and the parameter recalibration action. It evaluates the degree of explanation of the fault phenomenon and the degree of prediction of key performance recovery of the alternative maintenance set. When the alternative maintenance set is sufficient to explain the fault phenomenon and the prediction of key performance recovery meets the threshold range, the counterfactual substitution verification unit is used to increase the output of the hierarchical substitution inhibition coefficient corresponding to the replacement item of the upper-level module. When the alternative maintenance set cannot explain the fault phenomenon or the prediction of key performance recovery cannot meet the threshold range, the counterfactual substitution verification unit is used to maintain the output of the hierarchical substitution inhibition coefficient.

[0054] The maintenance history valid sample library maintenance unit is used to maintain the maintenance history valid sample library and output the maintenance history valid sample library to the history consistency score calculation unit. The samples in the maintenance history valid sample library have undergone equipment self-inspection, operation recovery test, performance confirmation test, and equivalent post-verification, and the sample results can prove that the replacement action effectively eliminates the fault or restores the key performance.

[0055] The historical consistency score calculation unit is used to calculate the similarity between the current audit admissible evidence atomic set and the corresponding evidence set of historical samples in the maintenance history valid sample library, and calculate the historical consistency score of the smallest replaceable unit and output it to the necessity score calculation unit. The historical consistency score is shown in equation (3): (3) The hard constraint removal unit is used to perform hard constraint removal on the i-th row of the quotation BOM and output the gating value to the necessity score calculation unit. Hard constraint removal includes: the row item has been uniquely mapped to the minimum replaceable unit; the auditable evidence atom set has a leading evidence atom that is physically related to the minimum replaceable unit; there is a verifiable structural propagation relationship between the minimum replaceable unit and the fault manifestation; the hierarchical replacement inhibition judgment unit has not excluded the row item; and the auditable evidence atom set does not have strong conflicting evidence that directly excludes the fault of the minimum replaceable unit. The gating value is 1 when all hard constraints are satisfied and 0 when any hard constraint is not satisfied.

[0056] The necessity score calculation unit is used to calculate the necessity score based on the gating quantity, posterior failure probability, structural dependency score, historical consistency score, conflict coefficient, and hierarchical replacement inhibition coefficient, and output the necessity score to the verification conclusion output unit. The necessity score is shown in equation (4): (4) The verification conclusion output unit is used to divide the items in the i-th row of the quotation BOM into items that can be replaced, items that need to be recertified, and items that cannot be replaced based on the necessity score, and output the division results to the maintenance quotation receiving end and the minimum recertification path solving module.

[0057] The minimum proof path solution module includes: The candidate supplementary certification action set definition unit is used to define the candidate supplementary certification action set for the minimum replaceable unit and the competing minimum replaceable unit set of the item to be supplemented and output the candidate supplementary certification action set to the test cost setting unit and the differentiation gain calculation unit. The candidate supplementary certification action set includes local isolation test, directional sensor supplementary sampling, self-test sub-process execution, power-on waveform sampling, valve pre-valve and post-valve differential pressure measurement, replacement board insertion and removal verification, and target interface thermal imaging acquisition.

[0058] The test cost setting unit is used to set the test cost for each candidate supplementary proof action and output the test cost to the minimum supplementary proof path solving unit.

[0059] The discrimination gain calculation unit is used to calculate the minimum discrimination gain of the set of supplementary actions relative to the competing minimum substitutable unit set based on the Kullback-Leibler divergence accumulation, and outputs the minimum discrimination gain to the minimum supplementary path solving unit. The minimum discrimination gain is shown in Equation (5): (5) The minimum supplementary evidence path solving unit is used to select the set of supplementary evidence actions with the minimum total test cost and satisfy the minimum discrimination threshold constraint from the candidate set of supplementary evidence actions. The set of supplementary evidence actions is shown in Equation (6): (6) The certificate supplement execution sequence output unit is used to output the certificate supplement action set as the certificate supplement execution sequence of the item to be supplemented and output the certificate supplement execution sequence to the maintenance service provider and the maintenance work order system.

[0060] The post-repair causal attribution and sample and rule update module includes: The post-maintenance verification data acquisition unit is used to collect post-maintenance verification data from the IoT acquisition device on the equipment side and the hospital operation and maintenance management system after the maintenance service provider completes the maintenance, and outputs the post-maintenance verification data set to the marginal recovery contribution calculation unit. The post-maintenance verification data includes power-on self-test results, fault disappearance status, key operating parameter recovery status, waveform or timing feature recovery status, and recurrence status within the preset observation period.

[0061] The actual execution replacement item set acquisition unit is used to obtain the actual execution replacement item set from the maintenance work order system and output the actual execution replacement item set to the marginal recovery contribution calculation unit.

[0062] The marginal recovery contribution calculation unit is used to calculate the marginal recovery contribution output of the minimum replaceable unit corresponding to each replacement item based on the residual failure loss prediction value of the post-maintenance verification data set and the actual replacement item set. The marginal recovery contribution is shown in Equation (7): (7) The attribution weight calculation unit is used to calculate the attribution weight output based on the marginal recovery contribution. The attribution weight is shown in equation (8): (8) The maintenance history valid sample library update unit is used to update the maintenance history valid sample library based on attribution weight and high contribution threshold and low contribution threshold, and output the update result to the maintenance history valid sample library maintenance unit. The high contribution threshold satisfies the trigger condition for attribution weight to write positive samples, the low contribution threshold satisfies the trigger condition for attribution weight to write inefficient or invalid samples, and the intermediate interval satisfies the trigger condition for attribution weight to write weak positive samples or unconfirmed samples.

[0063] The sample and rule update unit is used to continuously update the fault mode library, the set of competing minimum replaceable units, the minimum supplementary evidence path library, the set of conflicting evidence, and the conflict coefficient based on the supplementary evidence results, the post-repair verification data set, and the attribution weight. It then outputs the updated fault mode library, the set of competing minimum replaceable units, the set of minimum supplementary evidence path, the set of conflicting evidence, and the conflict coefficient to the verification judgment module and the minimum supplementary evidence path solving module. When the conclusion of the supplementary evidence item is reversed after supplementary evidence is obtained, the sample and rule update unit is used to update the evidence template, the set of competing minimum replaceable units, and the minimum supplementary evidence path output. When the replacement item of the superior module that was rejected due to hierarchical replacement suppression appears to propagate evidence across layers in subsequent scenarios, the sample and rule update unit is used to increase the weight output of the cross-layer propagation evidence template. When the attribution weight shows that the contribution of the replacement item is low, the sample and rule update unit is used to increase the output of the conflict coefficient under similar fault scenarios.

Claims

1. A method for verifying medical equipment repair quotes, characterized in that, Includes the following steps: S1. The technical audit server receives the Bill of Materials (BOM) submitted by the maintenance quotation receiver, determines the time of occurrence of the current fault event, and extracts multi-source data from the IoT acquisition device on the equipment side and the hospital's operation and maintenance management system within a preset time window before the fault occurred. S2. The technical audit server performs preprocessing on multi-source data, forms a candidate evidence set based on the degree of anomaly, performs audit admissibility screening on the candidate evidence set, and registers audit admissibility evidence atoms. S3. The technical audit server maintains the equipment structure tree. The technical audit server performs standardization processing on each item in the quotation BOM and maps each item to the corresponding smallest replaceable unit in the equipment structure tree. When the mapping cannot be uniquely determined, a supplementary information request is output and the corresponding item is marked as a mapping uncertainty item. The mapping uncertainty item does not enter the necessity determination process before the mapping is completed. S4. The technical audit server determines the set of competing minimum replaceable units for each line of items, introduces hierarchical replacement suppression rules, and applies hierarchical replacement suppression to the replacement items of the upper-level module. The technical audit server performs hard constraint screening on each line of items to obtain a gating quantity. Based on the gating quantity and the atomic evidence that can be accepted by the audit, a necessity score is obtained. Based on the necessity score, each line of items is divided into items that support replacement, items that need to be supplemented, and items that do not support replacement. S5. The technical audit server defines a set of candidate supplementary verification actions for the items to be verified, solves the minimum supplementary verification path that satisfies the minimum discrimination threshold constraint based on the test cost and discrimination gain of the candidate supplementary verification actions, and outputs the minimum supplementary verification path as the supplementary verification execution sequence of the items to be verified.

2. The method for verifying medical equipment repair quotations according to claim 1, characterized in that, In step S1, determining the time of failure includes: when the hospital's operation and maintenance management system has records of downtime, alarm time, or repair request time, the technical audit server selects the earliest and most reliable record as the time of failure; when the hospital's operation and maintenance management system does not have a clear record of the time of failure, the technical audit server infers the time of failure based on the change point detection results of the key monitoring channel; the preset time window consists of a short window, a medium window, and a long window. When the target medical equipment has periodic operating conditions, a control window that matches the historical normal cycle is extracted at the same time, and an individualized health baseline is constructed based on the control window.

3. The method for verifying medical equipment repair quotations according to claim 1, characterized in that, The preprocessing in step S2 includes time alignment, unit normalization, missing value imputation, abnormal noise removal, and channel identifier standardization. When multi-source data does not use a unified clock stamp, time axis reconstruction is performed based on the acquisition gateway timestamp, message sequence number, system log, and co-occurring events. Anomaly degree includes the Mahalanobis distance anomaly degree of the feature vector relative to the healthy distribution. Auditable admissible evidence atoms include the occurrence time interval, evidence source, feature type, anomaly degree, credibility, and original data index. When the same physical phenomenon is repeatedly observed by multiple acquisition sources, multiple auditable admissible evidence atoms are merged to obtain a composite auditable admissible evidence atom.

4. The method for verifying medical equipment repair quotations according to claim 1, characterized in that, The standardization process in step S3 includes deambiguity of project names, merging of component codes, replacement of aliases, identification of model substitution relationships, and identification of maintenance actions. Maintenance actions include overall replacement, board-level replacement, partial repair, calibration, cleaning, and tightening. The output of supplementary information requests includes requests for supplementary codes, pictures, or maintenance manual page numbers.

5. The method for verifying medical equipment repair quotations according to claim 1, characterized in that, In step S4, the competing minimum replaceable unit set satisfies the following conditions: the competing minimum replaceable unit can explain at least some of the key symptoms in the current fault manifestation; the competing minimum replaceable unit is located in the same fault propagation chain or an adjacent fault propagation chain as the minimum replaceable unit; and the auditable evidence atom set cannot fully distinguish the competing minimum replaceable unit from the minimum replaceable unit. The hierarchical replacement suppression rule in step S4 takes effect when the lower-level minimum replaceable unit set can explain the current fault manifestation and there is no cross-layer propagation evidence in the auditable evidence atom set. Cross-layer propagation evidence includes upper-level power supply chain anomalies, upper-level control chain anomalies, upper-level thermal field diffusion anomalies, and multi-channel synchronization mismatch anomalies. The hard constraint screening in step S4 includes: the project has been uniquely mapped to the smallest replaceable unit; there is a leading evidence atom in the auditable evidence atom set that is physically related to the smallest replaceable unit; there is a verifiable structural propagation relationship between the smallest replaceable unit and the current failure manifestation; the hierarchical replacement suppression rule has not excluded the project; and there is no strong conflicting evidence in the auditable evidence atom set that directly excludes the failure of the smallest replaceable unit. The necessity score is obtained by combining gating quantity, posterior failure probability, structural dependency score, historical consistency score, conflict coefficient and hierarchical replacement inhibition coefficient.

6. The method for verifying medical equipment repair quotations according to claim 1, characterized in that, After the technical audit server outputs the supplementary verification execution sequence in step S5, the process also includes: after the maintenance service provider completes the maintenance, the technical audit server collects post-maintenance verification data, which includes power-on self-test results, fault disappearance status, key operating parameter recovery status, waveform or timing feature recovery status, and recurrence status within a preset observation period; the technical audit server performs item-by-item causal attribution for multiple actual replacement items, calculates marginal recovery contribution and obtains attribution weights, updates the effective sample library of maintenance history based on the attribution weights, and updates the fault mode library, the set of competing minimum replaceable units, the minimum supplementary verification path library, the set of conflicting evidence, and the conflict coefficient based on the supplementary verification results, the post-maintenance verification data, and the item-by-item causal attribution results.

7. A medical equipment repair quotation verification system, characterized in that, include: The fault window and multi-source data module receive the Bill of Materials (BOM) submitted by the maintenance quotation receiver, determine the time of the fault occurrence, and extract multi-source data within a preset time window before the fault occurs. The evidence atom generation module performs preprocessing on multi-source data, forms a candidate evidence set based on the degree of anomaly, performs audit-admissibility screening on the candidate evidence set, and registers audit-admissible evidence atoms. The quotation BOM mapping module maintains the equipment structure tree, performs standardization processing on each item in the quotation BOM and maps it to the smallest replaceable unit. When the mapping cannot be uniquely determined, it outputs a supplementary information request and marks the corresponding item as a mapping uncertainty item. The verification and judgment module determines the minimum set of competitive substitutable units, introduces hierarchical substitution suppression rules and applies hierarchical substitution suppression to the replacement items of the superior module, performs hard constraint screening to obtain the gate quantity, obtains the necessity score based on the gate quantity and the atomic evidence that can be accepted by the audit, and outputs the items that support the replacement, the items that need to be supplemented, and the items that do not support the replacement. The minimum supplementary evidence path solution module defines a set of candidate supplementary evidence actions for the item to be supplemented, solves the minimum supplementary evidence path that satisfies the minimum discrimination threshold constraint based on the test cost and discrimination gain, and outputs the supplementary evidence execution sequence.

8. A medical equipment repair quotation verification system according to claim 7, characterized in that, The audit admissibility screening performed by the evidence atom generation module includes checking data integrity, sampling continuity, clock consistency, reliability of the collection source, and traceability of the original record. Audit admissible evidence atoms include the time interval of occurrence, evidence source, feature type, anomaly degree, credibility, and original data index. When the same physical phenomenon is repeatedly observed by multiple collection sources, the evidence atom generation module merges multiple audit admissible evidence atoms to obtain composite audit admissible evidence atoms.

9. A medical equipment repair quotation verification system according to claim 7, characterized in that, The minimum supplementary evidence path solving module sets the test cost for candidate supplementary evidence actions. The minimum supplementary evidence path solving module selects the combination of supplementary evidence actions with the minimum test cost and that satisfies the minimum discrimination threshold constraint from the set of candidate supplementary evidence actions as the minimum supplementary evidence path, and calculates the discrimination gain based on the cumulative Kullback-Leibler divergence between the minimum replaceable unit and the competing minimum replaceable unit set.

10. A medical equipment repair quotation verification system according to claim 7, characterized in that, It also includes a post-repair causal attribution and sample and rule update module. This module collects post-repair verification data and calculates marginal recovery contribution and attribution weight for multiple actual replacement items. The post-repair causal attribution and sample and rule update module updates the effective sample library of repair history according to the attribution weight, and updates the fault mode library, the set of competing minimum replaceable units, the minimum supplementary evidence path library, the set of conflict evidence, and the conflict coefficient.

Images