Operation and maintenance data management system and method based on large model

Abstract

The application discloses a big model-based operation and maintenance data management system and method, and relates to the technical field of data management.The system comprises a data acquisition module, a data processing module, a rule mining module, a dynamic adaptation module and a feedback optimization module; the data acquisition module acquires environment sensitive parameters and operation threshold data and calibrates and verifies; the data processing module constructs environment-threshold association data sets; the rule mining module generates a dynamic anchoring rule library through a pre-trained big model; the dynamic adaptation module matches rules to generate adjustment instructions and operation and maintenance suggestions; and the feedback optimization module records evaluation adjustment effects and supports rule library updating.The application realizes dynamic anchoring adaptation of operation and maintenance thresholds and environment changes, and improves equipment operation stability and operation and maintenance accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data management, in particular to an operation and maintenance data management system and method based on a large model. BACKGROUND

[0002] In the field of intelligent operation and maintenance and data management, special equipment that requires high adaptability to the operating environment and needs to be dynamically regulated and controlled through parameters to ensure core functions faces special challenges in operation and maintenance management. Such special equipment is widely used in industrial production, infrastructure protection and other scenarios, and the core function output is easily disturbed by environmental factors such as wind, temperature and humidity, precipitation, and light. Wind can change the range of action propagation, temperature and humidity affect the sensitivity of components, and precipitation can interfere with the triggering accuracy, resulting in deviations between actual operation results and expectations. Therefore, optimizing equipment operation parameters and strategies based on dynamic changes in the environment has become a key requirement to ensure that the equipment functions stably.

[0003] The existing operation and maintenance data management scheme based on a large model has obvious deficiencies: first, the running threshold is fixed, and the preset fixed parameters cannot adapt to dynamic changes in the environment, and the device function is easily attenuated under complex working conditions; second, the environmental data and device data are separated, and no linkage analysis logic is established, and the environmental data is not effectively converted into parameter adjustment basis; third, there is a lack of mining of the implicit correlation between the environment and the threshold, and the operation and maintenance decision relies on artificial experience, with low precision and efficiency.

[0004] The existing technology cannot meet the environmental adaptation type dynamic operation and maintenance requirements of such equipment, resulting in insufficient equipment operation stability and high operation and maintenance costs, and there is an urgent need for an operation and maintenance data management system and method that can dynamically associate environmental sensitive parameters with running thresholds and accurately adapt operation and maintenance strategies to solve the pain points of existing technology. SUMMARY

[0005] The purpose of the present application is to provide an operation and maintenance data management system and method based on a large model to solve the problems raised in the background.

[0006] In order to solve the above technical problems, the present application provides the following technical solutions:

[0007] The operation and maintenance data management method based on a large model comprises the following steps:

[0008] S1. Collecting environmental sensitive parameters and running threshold data of the target device, the environmental sensitive parameters including quantized data of wind level, temperature and humidity, precipitation intensity, and light intensity, and the running threshold data including real-time values of action output effective threshold, component working safety threshold, and trigger response threshold;

[0009] S2. Quantize the collected environmental sensitive parameters according to the preset interval to form an environmental sensitive parameter quantization interval; label the running threshold data with a feature tag according to the functional effectiveness, and complete the two-dimensional alignment processing of the environmental sensitive parameter quantization interval and the running threshold data through feature mapping to form an environment-threshold association data set;

[0010] S3. Input the association data set into a pre-trained large model, mine the dynamic anchoring rules of different environmental sensitive parameter quantization intervals and threshold adjustment amplitudes through the large model, and generate a dynamic anchoring rule library; the dynamic anchoring rule library clearly defines the one-to-one quantitative relationship between the environmental sensitive parameter change amplitude and the threshold adjustment amount;

[0011] S4. Real-time acquisition of the current environmental sensitive parameter and conversion into the corresponding quantization interval, reverse matching of the corresponding threshold adjustment standard based on the dynamic anchoring rule library, automatic generation of quantitative adjustment instructions and operation and maintenance strategy adaptation suggestions, output to the device control system for execution, and realization of the dynamic anchoring adaptation of the operation and maintenance threshold and the environmental change.

[0012] Further, S1 includes the following contents:

[0013] Through the multi-sensor array deployed on the target device, the instantaneous value of each environmental sensitive parameter is monitored in real time, the parameter change rate ΔP in the adjacent acquisition period is calculated, and when ΔP≥preset change rate threshold P0, the environmental sensitive parameter acquisition of the current period is started; and the calculation formula of the parameter change rate ΔP is: ΔP=|P n -P n-1 | / Δt, wherein P n is the instantaneous value of the environmental sensitive parameter at the current time, P n-1 is the instantaneous value of the environmental sensitive parameter in the previous acquisition period, Δt is the time interval of the adjacent acquisition period, and P0 is the inherent threshold value calibrated based on the sensor response sensitivity before the device is shipped;

[0014] Error correction is performed on the collected original environmental sensitive parameters, and calibrated quantitative data is calculated based on the sensor calibration data. The calibration formula is: Pcal=Praw×α+β, wherein Pcal is the calibrated environmental sensitive parameter quantitative data, Praw is the sensor collected original data, α is the sensor error correction coefficient, and β is the sensor zero drift offset. α and β are obtained by fitting the least squares method through multiple calibration tests of the sensor and the standard metering device under the same environmental conditions;

[0015] After collecting the running threshold real-time value, the deviation rate R of the running threshold real-time value and the rated threshold of the device is calculated, and when R≤preset deviation threshold R0, it is determined that the data is valid and is retained; the calculation formula of the deviation rate R is: R=|Vreal-Vrated| / Vrated×100%, wherein Vreal is the running threshold real-time value, Vrated is the rated threshold set by the device when it is shipped, and R0 is the upper limit of the allowed deviation determined based on the reliability requirement of the device operation;

[0016] The calibrated environmental sensitive parameter quantization data and the verified running threshold data are added with time stamp information of uniform precision, respectively, to ensure that the time stamp error of the two types of data is ≤10 ms, and an initial collection data set that is time-synchronized is formed.

[0017] Further, S2 includes the following content:

[0018] Based on the statistical distribution characteristics of the environmental sensitive parameter quantization data in the initial collection data, a K-means clustering algorithm is used to determine the quantization classification interval boundary; the clustering number k is determined by calculating the silhouette coefficient, the calculation formula of the silhouette coefficient L is: L=(b-a) / max(a,b), wherein a is the average distance of the sample and other samples in the same cluster, b is the average distance of the sample and all samples in the nearest different cluster, and the k corresponding to the maximum value of L is selected as the optimal clustering number, and then k non-overlapping environmental sensitive parameter quantization intervals are divided;

[0019] The function compliance rate η corresponding to each running threshold real-time value is calculated, and η=(actual function satisfaction time length / statistical cycle total time length)×100%, wherein the actual function satisfaction time length is the cumulative time length during which the running threshold is in the effective interval; the label level is divided according to the distribution range of η, and the label level includes excellent, qualified and to be adjusted, and the label level and the corresponding running threshold data are bound, wherein the label level of excellent needs to satisfy η≥95%, the label level of to be adjusted needs to satisfy 80%≤η<95%, and the label level of to be adjusted needs to satisfy η<80%;

[0020] The feature vector of the environmental sensitive parameter quantization interval and the label vector of the running threshold data are constructed, the correlation degree cosθ between the two is calculated by cosine similarity, and cosθ=(X·Y) / (||X||×||Y||), wherein X is the feature vector of the environmental sensitive parameter quantization interval, and Y is the label vector of the running threshold data; when cosθ≥preset correlation degree threshold θ0, it is determined that there is an effective correlation between the two, θ0 is obtained by statistical analysis of historical effective operation and maintenance data, and the value range is 0.7-0.85; all environmental sensitive parameter quantization intervals and running threshold data that have an effective correlation are ranked in descending order of correlation degree, data pairs with a correlation degree lower than θ0 are removed, the remaining data pairs are retained and integrated into an environment-threshold correlation data set.

[0021] Furthermore, S3 includes the following:

[0022] For the environment-threshold related dataset, statistical features of the quantification interval of each environmental sensitive parameter are extracted. These statistical features include the interval mean μ, standard deviation σ, distribution skewness γ, and mean rate of change Δμ. Candidate features with |r|≥0.6 are screened using the Pearson correlation coefficient r. The dynamic weight wj of each candidate feature is calculated using the entropy weight method. The weight calculation formula is: wj=(1-Hi) / ∑(1-Hj), where Hi=-∑(Pij×lnPij), Pij is the normalized proportion of the j-th sample of the i-th feature, and Hj is the information entropy of the j-th feature. Features with weight wj≥w0 are selected to form the core feature set, where w0 represents the weight threshold and is determined based on the mean of all feature weights + 1 standard deviation.

[0023] The core feature set is input into the pre-trained large model, and the output is the threshold adjustment sample set {ΔVi} corresponding to the quantization interval of each environmental sensitive parameter. The baseline adjustment Vb, anchoring coefficient λ, and dynamic fluctuation range are calculated in sequence. Vb = ∑(ΔVi × wj), anchoring coefficient λ = |r_avg| × (1 - CV), where r_avg is the average correlation coefficient between the core feature and ΔVi, CV is the coefficient of variation of ΔVi, and CV = σ_ΔV / Vb, where σ_ΔV is the standard deviation of ΔVi. The dynamic fluctuation range is [Vb - λ × σ_ΔV, Vb + λ × σ_ΔV];

[0024] Initial rules are generated based on the correspondence between the quantification intervals of each environmentally sensitive parameter and the baseline adjustment amount Vb, dynamic fluctuation range, and anchoring coefficient λ. The support Sup, confidence Conf, and comprehensive score S of each rule are calculated, and S = 0.4 × Sup + 0.4 × Conf + 0.2 × λ, where Sup = number of valid samples of the rule / total number of samples in the associated dataset × 100%, and Conf = number of correctly matched samples of the rule / total number of matched samples of the rule × 100%. For conflicting rules with overlap, the rule with the highest comprehensive score S and a sample size greater than or equal to a preset number is retained, and the remaining rules are removed. The preset number is determined based on the statistical significance level to ensure that the rule samples are representative.

[0025] Furthermore, S4 includes the following:

[0026] The raw data of the current environmental sensitive parameters of the target device are acquired in real time. The error is corrected using the calibration formula in step S1 to obtain the quantitative data of the current environmental sensitive parameters. The quantitative data of the current environmental sensitive parameters is matched with the quantitative interval of the environmental sensitive parameters divided in step S2 to determine the corresponding target quantitative interval. If the current quantitative data is within ±5% of the critical value of the two intervals, the interval with the larger anchoring coefficient λ is taken as the target quantitative interval.

[0027] Based on the target quantization range, the corresponding baseline adjustment amount Vb, dynamic fluctuation range, and anchoring coefficient λ are matched from the dynamic anchoring rule base; the real-time adjustment amount ΔVreal = Vb + λ × (Pcur - Pavg) is calculated, where Pcur is the current environmental sensitive parameter quantization data, and Pavg is the interval mean μ of the target quantization range; if ΔVreal exceeds the dynamic fluctuation range, the critical value of the fluctuation range is taken as the final real-time adjustment amount.

[0028] Based on the final real-time adjustment amount, a quantitative adjustment instruction is generated, and an operation and maintenance strategy adaptation suggestion is generated simultaneously. The operation and maintenance strategy adaptation suggestion includes the adjustment execution priority, and the priority calculation formula is: Pri=λ×(1-CV)+0.3×(1-R'), where CV is the coefficient of variation of the corresponding threshold adjustment amount, and R' is the deviation rate of the current operating threshold data. When Pri≥0.7, the priority is set to the highest, and the quantitative adjustment instruction is executed first. The quantitative adjustment instruction and the operation and maintenance strategy adaptation suggestion are output to the equipment control system for execution, and the instruction execution time, the operating threshold data before and after execution, and the environmental sensitive parameters are recorded. The functional compliance rate η' after the instruction execution is calculated, η'=(functional compliance time after execution / total statistical period after execution)×100%, and η' is compared with the functional compliance rate η before execution to form adjustment effect feedback data, which is used for subsequent S3 rule base updates.

[0029] The operation and maintenance data management system based on a large model includes: a data acquisition module, a data processing module, a rule mining module, a dynamic adaptation module, and a feedback optimization module.

[0030] The data acquisition module is used to acquire environmental sensitive parameters and operating threshold data of the target device in real time, and to complete data calibration and validity verification.

[0031] The data processing module is used to quantify, classify, and label the collected data, and to construct an environment-threshold related dataset.

[0032] The rule mining module is used to input associated datasets into a pre-trained large model, mine dynamic anchoring patterns, and generate a dynamic anchoring rule library.

[0033] The dynamic adaptation module is used to generate quantitative adjustment instructions and operation and maintenance strategy adaptation suggestions based on a real-time environmental parameter matching rule base.

[0034] The feedback optimization module is used to record the execution data of adjustment instructions, evaluate the adjustment effect, and generate feedback data to support rule base updates.

[0035] Furthermore, the data acquisition module includes a parameter monitoring unit and a data calibration unit;

[0036] The parameter monitoring unit monitors the instantaneous values ​​of environmentally sensitive parameters and the real-time values ​​of operating thresholds in real time through a multi-sensor array, and triggers targeted acquisition actions based on the parameter change rate; the data calibration unit performs error correction and validity verification on the raw acquired data, and forms an initial acquired dataset with time synchronization through timestamp alignment processing.

[0037] Furthermore, the data processing module includes an interval partitioning unit and an association construction unit. The interval partitioning unit adaptively partitions quantized intervals based on the statistical distribution characteristics of environmentally sensitive parameters using a clustering algorithm to determine the optimal number of intervals. The association construction unit provides validity labels for the threshold data annotation function, calculates the correlation between environmental intervals and threshold data, filters valid association data, and integrates them to form an environment-threshold association dataset.

[0038] Furthermore, the rule mining module includes a feature filtering unit and a rule generation unit;

[0039] The feature selection unit extracts environmental interval statistical features from the associated dataset and selects the core feature set through correlation analysis and dynamic weighting. The rule generation unit inputs the core feature set into the pre-trained large model, quantitatively mines the dynamic anchoring pattern between the environmental interval and the threshold adjustment amount, generates a dynamic anchoring rule library, and eliminates conflicting rules.

[0040] Furthermore, the dynamic adaptation module includes an interval matching unit and an instruction generation unit;

[0041] The interval matching unit performs interval matching on the environmental parameters after real-time calibration, and matches the target interval based on the anchoring coefficient; the instruction generation unit calculates the real-time threshold adjustment amount based on the corresponding rules of the target interval matching, and generates quantitative adjustment instructions and operation and maintenance strategy adaptation suggestions including execution priority.

[0042] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention breaks through the limitation of the separation between environmental data and equipment data in existing technologies by establishing a dual-dimensional linkage analysis logic of environmentally sensitive parameters and operating thresholds. It effectively transforms dynamic environmental changes into precise bases for adjusting operation and maintenance parameters, avoiding the problem that fixed thresholds cannot adapt to complex working conditions. This invention mines implicit dynamic anchoring patterns between environmental ranges and threshold adjustments through a pre-trained large model, replacing the traditional manual experience-based decision-making mode, and constructing a statistically valid dynamic anchoring rule base, significantly improving the accuracy and automation level of operation and maintenance decisions. This invention forms a refined data processing system throughout the entire process through targeted screening and calibration in the data collection stage, effective correlation construction in the data processing stage, and intelligent adaptation to critical scenarios in the rule application stage, solving the problem of insufficient data quality leading to operation and maintenance adaptation deviations in existing technologies. This invention achieves a closed-loop operation and maintenance logic of "data collection - rule mining - dynamic adaptation - effect iteration" through real-time feedback of operation and maintenance adjustment effects and a dynamic optimization mechanism for the rule base, continuously improving the equipment's adaptability to environmental changes, reducing operation and maintenance costs, and ensuring stable equipment operation. This invention quantifies the reliability of the correlation between the environment and thresholds, and optimizes interval matching and adjustment strategies based on this, breaking through the limitations of mechanically dividing intervals and rigidly executing adjustments in traditional operation and maintenance, and realizing the intelligent dynamic generation of environment-adaptive operation and maintenance strategies. Attached Figure Description

[0043] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0044] Figure 1 This is a schematic diagram of the modules of the operation and maintenance data management system based on a large model according to the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] Please see Figure 1 The present invention provides the following technical solution:

[0047] The operation and maintenance data management system based on a large model includes: a data acquisition module, a data processing module, a rule mining module, a dynamic adaptation module, and a feedback optimization module.

[0048] The data acquisition module is used to acquire environmental sensitive parameters and operating threshold data of the target device in real time, and to complete data calibration and validity verification.

[0049] The data processing module is used to quantify, classify, and label the collected data, and to construct an environment-threshold related dataset.

[0050] The rule mining module is used to input associated datasets into a pre-trained large model, mine dynamic anchoring patterns, and generate a dynamic anchoring rule library.

[0051] The dynamic adaptation module is used to generate quantitative adjustment instructions and operation and maintenance strategy adaptation suggestions based on a real-time environmental parameter matching rule base.

[0052] The feedback optimization module is used to record the execution data of adjustment instructions, evaluate the adjustment effect, and generate feedback data to support rule base updates.

[0053] The data acquisition module includes a parameter monitoring unit and a data calibration unit;

[0054] The parameter monitoring unit monitors the instantaneous values ​​of environmentally sensitive parameters and the real-time values ​​of operating thresholds in real time through a multi-sensor array, and triggers targeted acquisition actions based on the parameter change rate; the data calibration unit performs error correction and validity verification on the raw acquired data, and forms an initial acquired dataset with time synchronization through timestamp alignment processing.

[0055] The data processing module includes an interval partitioning unit and an association construction unit. The interval partitioning unit adaptively divides quantized intervals based on the statistical distribution characteristics of environmentally sensitive parameters using a clustering algorithm to determine the optimal number of intervals. The association construction unit provides validity labels for the threshold data annotation function, calculates the correlation between environmental intervals and threshold data, filters valid association data, and integrates them to form an environment-threshold association dataset.

[0056] The rule mining module includes a feature filtering unit and a rule generation unit;

[0057] The feature selection unit extracts environmental interval statistical features from the associated dataset and selects the core feature set through correlation analysis and dynamic weighting. The rule generation unit inputs the core feature set into the pre-trained large model, quantitatively mines the dynamic anchoring pattern between the environmental interval and the threshold adjustment amount, generates a dynamic anchoring rule library, and eliminates conflicting rules.

[0058] The dynamic adaptation module includes a range matching unit and an instruction generation unit;

[0059] The interval matching unit performs interval matching on the environmental parameters after real-time calibration, and matches the target interval based on the anchoring coefficient; the instruction generation unit calculates the real-time threshold adjustment amount based on the corresponding rules of the target interval matching, and generates quantitative adjustment instructions and operation and maintenance strategy adaptation suggestions including execution priority.

[0060] The operation and maintenance data management method based on large models includes the following steps:

[0061] S1. Collect environmental sensitive parameters and operating threshold data of the target equipment. The environmental sensitive parameters include quantitative data of wind force level, temperature and humidity, precipitation intensity and light intensity. The operating threshold data includes real-time values ​​of effective output threshold, component working safety threshold and trigger response threshold.

[0062] S2. The collected environmentally sensitive parameters are quantified and classified according to preset intervals to form the environmentally sensitive parameter quantification intervals; the operating threshold data are labeled with feature tags according to functional effectiveness, and the environmentally sensitive parameter quantification intervals and operating threshold data are aligned in two dimensions through feature mapping to form an environment-threshold related dataset.

[0063] S3. Input the associated dataset into the pre-trained large model, and use the large model to mine the dynamic anchoring rules between the quantization range of different environmental sensitive parameters and the threshold adjustment range, and generate a dynamic anchoring rule library; the dynamic anchoring rule library clearly defines the one-to-one quantization relationship between the change range of environmental sensitive parameters and the threshold adjustment amount;

[0064] S4. Real-time acquisition of current environmental sensitive parameters and conversion into corresponding quantization ranges. Based on the dynamic anchoring rule base, reverse matching of the corresponding threshold adjustment standards, automatic generation of quantization adjustment instructions and operation and maintenance strategy adaptation suggestions, output to the equipment control system for execution, and realize dynamic anchoring adaptation of operation and maintenance thresholds with environmental changes.

[0065] S1 includes the following:

[0066] The system utilizes a multi-sensor array deployed on the target device to monitor the instantaneous values ​​of various environmentally sensitive parameters in real time. It calculates the parameter change rate ΔP within adjacent acquisition cycles. When ΔP ≥ a preset change rate threshold P0, the acquisition of environmentally sensitive parameters for the current cycle is initiated. The formula for calculating the parameter change rate ΔP is: ΔP = |P n -P n-1 | / Δt, where P n P is the instantaneous value of the environmentally sensitive parameter at the current moment. n-1 Δt is the instantaneous value of the environmental sensitive parameter in the previous acquisition cycle, Δt is the time interval between adjacent acquisition cycles, and P0 is the inherent threshold calibrated based on the sensor response sensitivity before the device leaves the factory.

[0067] Error correction is performed on the collected raw environmental sensitive parameters. Based on the sensor calibration data, the calibrated quantitative data is calculated. The calibration formula is: Pcal=Praw×α+β, where Pcal is the calibrated environmental sensitive parameter quantitative data, Praw is the raw data collected by the sensor, α is the sensor error correction coefficient, and β is the sensor zero drift offset. α and β are obtained by fitting the sensor with standard metrology equipment under the same environmental conditions through multiple sets of benchmark tests.

[0068] After collecting the real-time value of the operating threshold, the deviation rate R between it and the rated threshold of the equipment is calculated. When R ≤ the preset deviation threshold R0, the data is deemed valid and retained. The formula for calculating the deviation rate R is: R = |Vreal - Vrated| / Vrated × 100%, where Vreal is the real-time value of the operating threshold, Vrated is the rated threshold set by the equipment at the factory, and R0 is the upper limit of the allowable deviation determined based on the equipment's operational reliability requirements.

[0069] For the calibrated environmentally sensitive parameter quantification data and the verified effective operating threshold data, timestamp information with uniform precision is added to ensure that the acquisition timestamp error of the two types of data is ≤10ms, forming an initial acquisition dataset with time synchronization.

[0070] S2 includes the following:

[0071] Based on the statistical distribution characteristics of the quantification data of environmental sensitive parameters in the initial collected data, the K-means clustering algorithm is used to determine the boundaries of the quantification grading intervals. The number of clusters k is determined by the silhouette coefficient. The formula for calculating the silhouette coefficient L is: L=(ba) / max(a,b), where a is the average distance between the sample and other samples in the same cluster, and b is the average distance between the sample and all samples in the nearest heterogeneous cluster. The k corresponding to the maximum value of L is selected as the optimal number of clusters, thereby dividing k non-overlapping quantification intervals of environmental sensitive parameters.

[0072] Calculate the functional compliance rate η corresponding to the real-time values ​​of each operating threshold, where η = (actual functional compliance time / total statistical period) × 100%, and the actual functional compliance time is the cumulative time during which the operating threshold is within the valid range. Classify the tags according to the distribution range of η, including Excellent, Qualified, and Needing Adjustment. Bind the tag level to the corresponding operating threshold data. For tags of Excellent, η ≥ 95% is required; for tags of Needing Adjustment, 80% ≤ η < 95% is required; and for tags of Needing Adjustment, η < 80% is required.

[0073] A feature vector for the quantification range of environmental sensitive parameters and a label vector for the operational threshold data are constructed. The correlation degree cosθ between the two is calculated using cosine similarity, where cosθ = (X·Y) / (||X||×||Y||), where X is the feature vector of the quantification range of environmental sensitive parameters and Y is the label vector of the operational threshold data. When cosθ ≥ a preset correlation degree threshold θ0, a valid correlation is determined between the two. θ0 is obtained by statistical analysis of historical valid operation and maintenance data, with a value range of 0.7-0.85. All environmental sensitive parameter quantification ranges and operational threshold data with valid correlation are sorted in descending order of correlation degree. Data pairs with correlation degree lower than θ0 are removed, and the remaining data pairs are retained and integrated into an environment-threshold correlation dataset.

[0074] S3 includes the following:

[0075] For the environment-threshold related dataset, statistical features of the quantification interval of each environmental sensitive parameter are extracted. These statistical features include the interval mean μ, standard deviation σ, distribution skewness γ, and mean rate of change Δμ. Candidate features with |r|≥0.6 are screened using the Pearson correlation coefficient r. The dynamic weight wj of each candidate feature is calculated using the entropy weight method. The weight calculation formula is: wj=(1-Hi) / ∑(1-Hj), where Hi=-∑(Pij×lnPij), Pij is the normalized proportion of the j-th sample of the i-th feature, and Hj is the information entropy of the j-th feature. Features with weight wj≥w0 are selected to form the core feature set, where w0 represents the weight threshold and is determined based on the mean of all feature weights + 1 standard deviation.

[0076] The core feature set is input into the pre-trained large model, and the output is the threshold adjustment sample set {ΔVi} corresponding to the quantization interval of each environmental sensitive parameter. The baseline adjustment Vb, anchoring coefficient λ, and dynamic fluctuation range are calculated in sequence. Vb = ∑(ΔVi × wj), anchoring coefficient λ = |r_avg| × (1 - CV), where r_avg is the average correlation coefficient between the core feature and ΔVi, CV is the coefficient of variation of ΔVi, and CV = σ_ΔV / Vb, where σ_ΔV is the standard deviation of ΔVi. The dynamic fluctuation range is [Vb - λ × σ_ΔV, Vb + λ × σ_ΔV];

[0077] Initial rules are generated based on the correspondence between the quantification intervals of each environmentally sensitive parameter and the baseline adjustment amount Vb, dynamic fluctuation range, and anchoring coefficient λ. The support Sup, confidence Conf, and comprehensive score S of each rule are calculated, and S = 0.4 × Sup + 0.4 × Conf + 0.2 × λ, where Sup = number of valid samples of the rule / total number of samples in the associated dataset × 100%, and Conf = number of correctly matched samples of the rule / total number of matched samples of the rule × 100%. For conflicting rules with overlap, the rule with the highest comprehensive score S and a sample size greater than or equal to a preset number is retained, and the remaining rules are removed. The preset number is determined based on the statistical significance level to ensure that the rule samples are representative.

[0078] In this embodiment, the rule library is dynamically calibrated and stored. The rule library update trigger condition is set as follows: when the amount of newly added environment-threshold data reaches 30% of the original associated dataset, or when the actual hit rate of any rule is less than 85%, the rule update step S3 is repeated. Finally, the dynamic anchoring rule library is stored in the structure of "environment-sensitive parameter quantification range - core feature weight - benchmark adjustment amount - dynamic fluctuation range - anchoring coefficient - applicable function label". The applicable function label corresponds one-to-one with the excellent, qualified and unadjustable labels in S2.

[0079] S4 includes the following:

[0080] The raw data of the current environmental sensitive parameters of the target device are acquired in real time. The error is corrected using the calibration formula in step S1 to obtain the quantitative data of the current environmental sensitive parameters. The quantitative data of the current environmental sensitive parameters is matched with the quantitative interval of the environmental sensitive parameters divided in step S2 to determine the corresponding target quantitative interval. If the current quantitative data is within ±5% of the critical value of the two intervals, the interval with the larger anchoring coefficient λ is taken as the target quantitative interval.

[0081] In this embodiment, assuming the target device is a bird deterrent device, the "wind force level" among its environmental sensitive parameters is divided into two adjacent quantization intervals in step S2. The dynamic anchoring rule base generated in step S3 has recorded the anchoring coefficients λ corresponding to the two intervals, as follows:

[0082] Interval 1 (Low Wind Interval): Wind force level range [0,10] (unit: m / s), corresponding anchoring coefficient λ1=0.80;

[0083] Interval 2 (Medium Stroke Interval): Wind force level range [10, 20] (unit: m / s), corresponding anchoring coefficient λ2 = 0.95;

[0084] The critical value is 10 m / s, and its ±5% range is: 10×(1-5%)=9.5 m / s to 10×(1+5%)=10.5 m / s, that is, the critical range is [9.5,10.5] m / s.

[0085] The original wind force level data of the bird deterrent device is obtained in real time. After correction by the S1 calibration formula, the current quantitative data Pcur=10.2m / s is obtained. It is observed that 10.2m / s falls within the critical range [9.5,10.5]m / s, and the corresponding two adjacent intervals (interval 1 and interval 2) need to be matched.

[0086] Extract the anchoring coefficients for two intervals from the dynamic anchoring rule base: λ1=0.80, λ2=0.95;

[0087] Comparing the values ​​of λ: λ2 (0.95) > λ1 (0.80), therefore, interval 2 is selected as the target quantization interval for the current data.

[0088] Based on the target quantization range, the corresponding baseline adjustment amount Vb, dynamic fluctuation range, and anchoring coefficient λ are matched from the dynamic anchoring rule base; the real-time adjustment amount ΔVreal = Vb + λ × (Pcur - Pavg) is calculated, where Pcur is the current environmental sensitive parameter quantization data, and Pavg is the interval mean μ of the target quantization range; if ΔVreal exceeds the dynamic fluctuation range, the critical value of the fluctuation range is taken as the final real-time adjustment amount.

[0089] Based on the final real-time adjustment amount, a quantitative adjustment instruction is generated, and an operation and maintenance strategy adaptation suggestion is generated simultaneously. The operation and maintenance strategy adaptation suggestion includes the adjustment execution priority, and the priority calculation formula is: Pri=λ×(1-CV)+0.3×(1-R'), where CV is the coefficient of variation of the corresponding threshold adjustment amount, and R' is the deviation rate of the current operating threshold data. When Pri≥0.7, the priority is set to the highest, and the quantitative adjustment instruction is executed first. The quantitative adjustment instruction and the operation and maintenance strategy adaptation suggestion are output to the equipment control system for execution, and the instruction execution time, the operating threshold data before and after execution, and the environmental sensitive parameters are recorded. The functional compliance rate η' after the instruction execution is calculated, η'=(functional compliance time after execution / total statistical period after execution)×100%, and η' is compared with the functional compliance rate η before execution to form adjustment effect feedback data, which is used for subsequent S3 rule base updates.

[0090] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0091] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for managing operation and maintenance data based on a large model, characterized by: The method includes the following steps: S1. Collect environmental sensitive parameters and operating threshold data of the target equipment. The environmental sensitive parameters include quantitative data of wind force level, temperature and humidity, precipitation intensity and light intensity. The operating threshold data includes real-time values ​​of effective output threshold, component working safety threshold and trigger response threshold. S2. The collected environmentally sensitive parameters are quantified and classified according to preset intervals to form the environmentally sensitive parameter quantification intervals; the operating threshold data are labeled with feature tags according to functional effectiveness, and the environmentally sensitive parameter quantification intervals and operating threshold data are aligned in two dimensions through feature mapping to form an environment-threshold related dataset. S3. Input the associated dataset into the pre-trained large model, and use the large model to mine the dynamic anchoring rules between the quantization range of different environmental sensitive parameters and the threshold adjustment range, and generate a dynamic anchoring rule library; the dynamic anchoring rule library clearly defines the one-to-one quantization relationship between the change range of environmental sensitive parameters and the threshold adjustment amount; S3 includes the following: For the environment-threshold related dataset, statistical features of the quantification interval of each environmental sensitive parameter are extracted. These statistical features include the interval mean μ, standard deviation σ, distribution skewness γ, and mean rate of change Δμ. Candidate features with |r|≥0.6 are screened using the Pearson correlation coefficient r. The dynamic weight wj of each candidate feature is calculated using the entropy weight method. The weight calculation formula is: wj=(1-Hi) / ∑(1-Hj), where Hi=-∑(Pij×lnPij), Pij is the normalized proportion of the j-th sample of the i-th feature, and Hj is the information entropy of the j-th feature. Features with weight wj≥w0 are selected to form the core feature set, where w0 is the weight threshold. The core feature set is input into the pre-trained large model, and the output is the threshold adjustment sample set {ΔVi} corresponding to the quantization interval of each environmental sensitive parameter. The baseline adjustment Vb, anchoring coefficient λ, and dynamic fluctuation range are calculated in sequence. Vb = ∑(ΔVi × wj), anchoring coefficient λ = |r_avg| × (1 - CV), where r_avg is the average correlation coefficient between the core feature and ΔVi, CV is the coefficient of variation of ΔVi, and CV = σ_ΔV / Vb, where σ_ΔV is the standard deviation of ΔVi. The dynamic fluctuation range is [Vb - λ × σ_ΔV, Vb + λ × σ_ΔV]; Initial rules are generated based on the correspondence between the quantification intervals of each environmentally sensitive parameter and the baseline adjustment amount Vb, dynamic fluctuation range, and anchoring coefficient λ. The support Sup, confidence Conf, and comprehensive score S of each rule are calculated, and S = 0.4 × Sup + 0.4 × Conf + 0.2 × λ, where Sup = number of valid samples of the rule / total number of samples in the associated dataset × 100%, and Conf = number of correctly matched samples of the rule / total number of matched samples of the rule × 100%. For conflicting rules with overlap, the rule with the highest comprehensive score S and a sample size greater than or equal to the preset number is retained, and the rest of the rules are removed. S4. Real-time acquisition of current environmental sensitive parameters and conversion into corresponding quantization ranges. Based on the dynamic anchoring rule base, reverse matching of the corresponding threshold adjustment standards, automatic generation of quantization adjustment instructions and operation and maintenance strategy adaptation suggestions, output to the equipment control system for execution, and realize dynamic anchoring adaptation of operation and maintenance thresholds with environmental changes.

2. The operation and maintenance data management method based on a large model according to claim 1, characterized in that: S1 includes the following: The system utilizes a multi-sensor array deployed on the target device to monitor the instantaneous values ​​of various environmentally sensitive parameters in real time. It calculates the parameter change rate ΔP within adjacent acquisition cycles. When ΔP ≥ a preset change rate threshold P0, the acquisition of environmentally sensitive parameters for the current cycle is initiated. The formula for calculating the parameter change rate ΔP is: ΔP = |P n -P n-1 | / Δt, where P n P is the instantaneous value of the environmentally sensitive parameter at the current moment. n-1 Δt is the instantaneous value of the environmental sensitive parameter in the previous acquisition cycle, Δt is the time interval between adjacent acquisition cycles, and P0 is the inherent threshold calibrated based on the sensor response sensitivity before the device leaves the factory. Error correction is performed on the collected raw environmental sensitive parameters. Based on the sensor calibration data, the calibrated quantitative data is calculated. The calibration formula is: Pcal=Praw×α+β, where Pcal is the calibrated environmental sensitive parameter quantitative data, Praw is the raw data collected by the sensor, α is the sensor error correction coefficient, and β is the sensor zero drift offset. α and β are obtained by fitting the sensor with standard metrology equipment under the same environmental conditions through multiple sets of benchmark tests. After collecting the real-time value of the operating threshold, the deviation rate R between it and the rated threshold of the equipment is calculated. When R ≤ the preset deviation threshold R0, the data is deemed valid and retained. The formula for calculating the deviation rate R is: R = |Vreal - Vrated| / Vrated × 100%, where Vreal is the real-time value of the operating threshold, Vrated is the rated threshold set by the equipment at the factory, and R0 is the upper limit of the allowable deviation determined based on the equipment's operational reliability requirements. For the calibrated environmentally sensitive parameter quantification data and the verified effective operating threshold data, timestamp information with uniform precision is added to ensure that the acquisition timestamp error of the two types of data is ≤10ms, forming an initial acquisition dataset with time synchronization.

3. The operation and maintenance data management method based on a large model according to claim 2, characterized in that: S2 includes the following: Based on the statistical distribution characteristics of the quantification data of environmental sensitive parameters in the initial collected data, the K-means clustering algorithm is used to determine the boundaries of the quantification grading intervals. The number of clusters k is determined by the silhouette coefficient. The formula for calculating the silhouette coefficient L is: L=(ba) / max(a,b), where a is the average distance between the sample and other samples in the same cluster, and b is the average distance between the sample and all samples in the nearest heterogeneous cluster. The k corresponding to the maximum value of L is selected as the optimal number of clusters, thereby dividing k non-overlapping quantification intervals of environmental sensitive parameters. Calculate the functional compliance rate η corresponding to the real-time values ​​of each operating threshold, where η = (actual functional compliance time / total statistical period) × 100%, and the actual functional compliance time is the cumulative time during which the operating threshold is within the valid range. Classify the tags according to the distribution range of η, including Excellent, Qualified, and Needing Adjustment. Bind the tag level to the corresponding operating threshold data. For tags of Excellent, η ≥ 95% is required; for tags of Needing Adjustment, 80% ≤ η < 95% is required; and for tags of Needing Adjustment, η < 80% is required. A feature vector for the quantification range of environmental sensitive parameters and a label vector for the operational threshold data are constructed. The correlation degree cosθ between the two is calculated using cosine similarity, where cosθ = (X·Y) / (||X||×||Y||), where X is the feature vector of the quantification range of environmental sensitive parameters and Y is the label vector of the operational threshold data. When cosθ ≥ a preset correlation degree threshold θ0, a valid correlation is determined between the two. θ0 is obtained by statistical analysis of historical valid operation and maintenance data, with a value range of 0.7-0.

85. All environmental sensitive parameter quantification ranges and operational threshold data with valid correlation are sorted in descending order of correlation degree. Data pairs with correlation degree lower than θ0 are removed, and the remaining data pairs are retained and integrated into an environment-threshold correlation dataset.

4. The operation and maintenance data management method based on a large model according to claim 1, characterized in that: S4 includes the following: The raw data of the current environmental sensitive parameters of the target device are acquired in real time, and the error is corrected using the calibration formula in step S1 to obtain the quantitative data of the current environmental sensitive parameters. The current quantized data of environmental sensitive parameters is matched with the quantized interval of environmental sensitive parameters divided in step S2 to determine the corresponding target quantization interval. If the current quantized data is within ±5% of the critical value of the two intervals, the interval with the larger anchoring coefficient λ is taken as the target quantization interval. Based on the target quantization interval, the corresponding benchmark adjustment amount Vb, dynamic fluctuation range and anchoring coefficient λ are matched from the dynamic anchoring rule base; the real-time adjustment amount ΔVreal=Vb+λ×(Pcur-Pavg) is calculated, where Pcur is the quantization data of the current environmental sensitive parameter and Pavg is the interval mean μ of the target quantization interval. If ΔVreal exceeds the dynamic fluctuation range, the critical value of the fluctuation range is taken as the final real-time adjustment amount. Based on the final real-time adjustment amount, a quantitative adjustment instruction is generated, and an operation and maintenance strategy adaptation suggestion is generated simultaneously. The operation and maintenance strategy adaptation suggestion includes the adjustment execution priority. The priority calculation formula is: Pri=λ×(1-CV)+0.3×(1-R'), where CV is the coefficient of variation of the corresponding threshold adjustment amount, and R' is the deviation rate of the current operating threshold data. When Pri≥0.7, the priority is set to the highest, and the adjustment instruction is executed first. The quantitative adjustment instruction and the operation and maintenance strategy adaptation suggestion are output to the equipment control system for execution, and the instruction execution time, the operating threshold data before and after execution, and the environmental sensitive parameters are recorded. Calculate the functional compliance rate η' after the instruction is executed, where η' = (duration of functional compliance after execution / total duration of statistical period after execution) × 100%, and compare η' with the functional compliance rate η before execution to form adjustment effect feedback data, which is used for subsequent S3 rule base updates.

5. A large-model-based operation and maintenance data management system, applied to the large-model-based operation and maintenance data management method according to any one of claims 1-4, characterized in that: The system includes: a data acquisition module, a data processing module, a rule mining module, a dynamic adaptation module, and a feedback optimization module; The data acquisition module is used to acquire environmental sensitive parameters and operating threshold data of the target device in real time, and to complete data calibration and validity verification. The data processing module is used to quantify, classify, and label the collected data to construct an environment-threshold related dataset. The rule mining module is used to input the associated dataset into the pre-trained large model, mine dynamic anchoring patterns, and generate a dynamic anchoring rule library. The dynamic adaptation module is used to generate quantitative adjustment instructions and operation and maintenance strategy adaptation suggestions based on a real-time environmental parameter matching rule library. The feedback optimization module is used to record the execution data of adjustment instructions, evaluate the adjustment effect, and generate feedback data to support rule base updates.

6. The operation and maintenance data management system based on a large model according to claim 5, characterized in that: The data acquisition module includes a parameter monitoring unit and a data calibration unit; The parameter monitoring unit monitors the instantaneous values ​​of environmentally sensitive parameters and the real-time values ​​of operating thresholds in real time through a multi-sensor array, and triggers targeted acquisition actions based on the parameter change rate; the data calibration unit performs error correction and validity verification on the original acquired data, and forms an initial acquired dataset with time synchronization through timestamp alignment processing.

7. The operation and maintenance data management system based on a large model according to claim 5, characterized in that: The data processing module includes an interval partitioning unit and an association construction unit. The interval partitioning unit adaptively partitions quantized intervals based on the statistical distribution characteristics of environmentally sensitive parameters using a clustering algorithm to determine the optimal number of intervals. The association construction unit runs the threshold data annotation function with validity labels, calculates the correlation between environmental intervals and threshold data, filters valid association data, and integrates them to form an environment-threshold association dataset.

8. The operation and maintenance data management system based on a large model according to claim 5, characterized in that: The rule mining module includes a feature filtering unit and a rule generation unit; The feature filtering unit extracts environmental interval statistical features from the associated dataset and filters the core feature set through correlation analysis and dynamic weighting. The rule generation unit inputs the core feature set into the pre-trained large model, quantifies and mines the dynamic anchoring pattern between the environmental interval and the threshold adjustment amount, generates a dynamic anchoring rule library, and eliminates conflicting rules.

9. The operation and maintenance data management system based on a large model according to claim 5, characterized in that: The dynamic adaptation module includes an interval matching unit and an instruction generation unit; The interval matching unit performs interval matching on the environmental parameters after real-time calibration and matches the target interval based on the anchoring coefficient; the instruction generation unit calculates the real-time threshold adjustment amount based on the corresponding rules of the target interval matching, and generates a quantitative adjustment instruction and an operation and maintenance strategy adaptation suggestion with execution priority.

Images