An artificial intelligence-based temperature control device data management system and method

Abstract

The application discloses a temperature control equipment data management system and method based on artificial intelligence, and relates to the technical field of equipment data analysis.The method comprises the following steps: acquiring branch current, calling historical steady-state time length statistical parameters to adjust window width, determining confidence identification and boundary state refinement segmentation granularity according to the last uniformity judgment, and extracting current fluctuation characteristic parameters; calling historical branch weight parameters to obtain dispersion degree parameters through weighted processing, combining dynamic normalization benchmark to convert into relative deviation degree, and generating uniformity pre-score; taking historical uniformity score distribution quantile line as a dynamic judgment threshold, outputting an exception when the pre-score is lower than the threshold and the deviation degree is higher than the floating range, generating a suspected exception identification and executing secondary backtracking segmentation and score recalculation to output the final judgment when the pre-score is lower than the threshold and the deviation degree is not higher than the range; and synchronously and iteratively updating branch weight, judgment threshold and steady-state time length statistical parameters.

Description

Technical Field

[0001] This invention relates to the field of equipment data analysis technology, specifically to an artificial intelligence-based temperature control equipment data management system and method. Background Technology

[0002] During operation, the consistency of the output power of the cooling or heating units of a temperature control device directly affects the temperature control effect. Taking a temperature control device using a semiconductor cooling chip array as an example, due to factors such as manufacturing tolerances, differences in contact impedance, and aging over long-term use, the actual cooling or heating power of each cooling chip unit may be inconsistent, resulting in uneven temperature distribution.

[0003] In existing technologies, data analysis for temperature control equipment typically employs a fixed-parameter analysis mode. Key analytical parameters such as feature extraction window width and state judgment thresholds are often preset values ​​before shipment, making it difficult to adapt to the slow drift of performance parameters during long-term use, resulting in a decrease in analysis accuracy over time. Existing judgment logic often uses a single threshold comparison method, where an anomaly is determined when a feature value exceeds or falls below a certain fixed threshold. This method lacks a comprehensive judgment architecture based on multi-dimensional parameters, making it prone to misjudgments due to single data fluctuations and unable to distinguish between different degrees of abnormality. In existing solutions, the analysis results of each run are only used as the output for that run, without accumulating the analysis data and feeding it back to subsequent analysis stages. Data analysis between runs is independent, lacking a mechanism for iteratively optimizing analytical parameters based on historical data, and the data analysis accuracy cannot improve with the number of runs. Existing technologies have a relatively simple output format for analysis results, mostly consisting of simple normal or abnormal labels, lacking a structured data tagging system, making it difficult to support subsequent data retrieval, statistical analysis, and temperature control quality traceability. Summary of the Invention

[0004] The purpose of this invention is to provide a data management system and method for temperature control equipment based on artificial intelligence, so as to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a data management method for temperature control equipment based on artificial intelligence, the method comprising:

[0006] S100: Obtain the branch current data during the operation of the temperature control equipment, retrieve the steady-state duration statistics from the historical operation archive of the temperature control equipment, and adjust the feature extraction window width; obtain the confidence flag of the most recent uniformity determination; if the determination result is a boundary state, refine the segmented analysis granularity of the data; extract the current fluctuation feature parameters based on the adjusted window and segmented granularity.

[0007] S200: Retrieve the weight parameters of each branch from the historical operation archive, perform weighted processing on the current data of each branch, and obtain the dispersion parameter; retrieve the dynamic normalization benchmark from the historical operation archive, convert the dispersion parameter into a relative deviation, and generate a uniformity pre-score result.

[0008] S300: Retrieve the historical uniformity score distribution of the temperature control equipment and use the quantile line as the dynamic judgment threshold; when the pre-scoring result is lower than the dynamic judgment threshold and the relative deviation exceeds the floating range, output the abnormal judgment result and abnormal category; when the pre-scoring result is lower than the dynamic judgment threshold and the relative deviation does not exceed the floating range, generate a suspected abnormality mark, perform secondary backtracking segmentation and score recalculation for the corresponding time period, and output the final judgment result based on the recalculation result.

[0009] S400: Generate structured labels containing uniformity level, anomaly category, risk level and anomaly period based on the final judgment result, and archive the labels with the current data of the corresponding operating period to the historical operation archive.

[0010] S500. Based on the current dispersion parameter and uniformity score result, update the weight parameters of each branch; incorporate the current uniformity score result into the historical score distribution and update the dynamic judgment threshold; incorporate the current steady-state duration into the historical steady-state duration statistical parameter.

[0011] According to the above scheme, step S100 includes:

[0012] S110. Obtain the current sampling data of each branch of the temperature control equipment during the operating period, and use it as the branch current data.

[0013] S120. Retrieve the steady-state duration statistics of the temperature control equipment from the historical operation archives, and determine the initial width as the feature extraction window width based on the steady-state duration statistics.

[0014] S130. Obtain the confidence flag output by the most recent uniformity determination. When the confidence flag indicates that the most recent determination result is a boundary state, reduce the length of the segmented analysis interval of the current data and refine the granularity of the segmented analysis of the data.

[0015] S140. Based on the feature extraction window width and segmented analysis granularity, the branch current data is segmented and extracted from each segment to extract current fluctuation characteristic parameters.

[0016] According to the above scheme, step S200 includes:

[0017] S210. Retrieve the weight parameters corresponding to each branch of the temperature control equipment from the historical operation archive. The weight parameters are generated by the frequency of historical anomalies in each branch.

[0018] S220. The current data of each branch is weighted and calculated with the corresponding weight parameters of each branch to generate a dispersion parameter that reflects the dispersion of the current of each branch and the consistency of impedance.

[0019] S230. Retrieve the dynamic normalized benchmark from the historical operation archives. The dynamic normalized benchmark is generated from the historical uniformity score distribution.

[0020] S240. Compare and calculate the discreteness parameter with the dynamic normalization benchmark to obtain the relative deviation degree that characterizes the current uniformity deviation degree, and generate the uniformity pre-score result based on the relative deviation degree.

[0021] According to the above scheme, step S300 includes:

[0022] S310. Retrieve the historical uniformity score distribution of the temperature control equipment from the historical operation archive, and use the set quantile line of the historical uniformity score distribution as the dynamic judgment threshold.

[0023] S320. Obtain the uniformity pre-score result and relative deviation, compare the uniformity pre-score result with the dynamic judgment threshold, compare the relative deviation with the fluctuation range, and determine the uniformity status of the current temperature control operation.

[0024] S330. When the uniformity pre-score is lower than the dynamic judgment threshold and the relative deviation exceeds the floating range, output the anomaly judgment result and the corresponding anomaly category.

[0025] S340. When the uniformity pre-score is lower than the dynamic judgment threshold and the relative deviation does not exceed the floating range, a suspected anomaly marker is generated.

[0026] S350. Based on the suspected anomaly marker, locate the running segment corresponding to the uniformity pre-score result, perform secondary backtracking segmentation and fluctuation characteristic parameter extraction on the branch current data of the running segment, recalculate the dispersion parameter, relative deviation and uniformity pre-score result based on the re-extracted fluctuation characteristic parameter, and perform secondary comparison between the uniformity pre-score result and the dynamic judgment threshold, as well as the relative deviation and floating range, and output the final judgment result based on the secondary comparison result.

[0027] According to the above scheme, step S400 includes:

[0028] S410. Obtain the final determination result of the uniformity status, and extract information on uniformity level, anomaly category, risk level and anomaly period.

[0029] S420. Combine uniformity level, anomaly category, risk level and anomaly period to generate structured labels, associate the structured labels with the branch current data of the corresponding operating period to form associated archived data.

[0030] S430. Archive the associated structured tags and branch current data to the historical operation archive.

[0031] According to the above scheme, step S500 includes:

[0032] S510. Obtain the dispersion parameters and uniformity score results generated in this operation. Based on the dispersion parameters and uniformity score results, perform iterative correction on the weight parameters of each branch of the temperature control equipment in the historical operation file to complete the update of the weight parameters of each branch.

[0033] S520. The uniformity score result of this run is added to the statistical sequence of the historical uniformity score distribution to form an updated historical uniformity score distribution. Based on the set quantile line of the updated historical uniformity score distribution, the dynamic judgment threshold is updated accordingly.

[0034] S530. Incorporate the steady-state duration data of this run into the cumulative statistical dimension of the historical steady-state duration statistical parameters, and complete the update of the historical steady-state duration statistical parameters.

[0035] An artificial intelligence-based temperature control equipment data management system, comprising: a preprocessing module, a uniformity evaluation module, a judgment backtracking module, a label archiving module, and an iterative update module;

[0036] The preprocessing module is used to acquire branch current data during the operation of the temperature control equipment, retrieve steady-state duration statistics from historical operation files to adjust the feature extraction window width, refine the data segmentation analysis granularity under boundary conditions based on the confidence flag of the most recent uniformity determination, and extract current fluctuation feature parameters based on the adjusted window and segmentation granularity.

[0037] The uniformity assessment module is used to retrieve the weight parameters of each branch in the historical operation archive, perform weighted processing on the current data of each branch to obtain the dispersion parameter, retrieve the dynamic normalization benchmark to convert the dispersion parameter into a relative deviation, and generate a uniformity pre-score result.

[0038] The judgment backtracking module is used to retrieve the historical uniformity score distribution of the temperature control equipment, use the quantile line as the dynamic judgment threshold, and judge the uniformity status by comparing the pre-scoring result with the relative deviation under two conditions. It performs secondary backtracking segmentation and score recalculation for suspected abnormal periods, and outputs the final judgment result based on the recalculation result.

[0039] The tag archiving module is used to generate structured tags containing uniformity level, anomaly category, risk level and anomaly period based on the final judgment result, and to associate the tags with the current data of the corresponding running period and archive them to the historical running archive.

[0040] The iterative update module is used to update the weight parameters of each branch based on the current dispersion parameter and uniformity score result, incorporate the current uniformity score result into the historical score distribution to update the dynamic judgment threshold, and incorporate the current steady-state duration into the historical steady-state duration statistical parameter.

[0041] According to the above scheme, the preprocessing module includes a data acquisition unit, a window adjustment unit, a granularity adjustment unit, and a feature extraction unit;

[0042] The data acquisition unit is used to acquire current sampling data of each branch of the temperature control equipment during the operating period, as branch current data.

[0043] The window adjustment unit is used to retrieve the steady-state duration statistics of the temperature control equipment from the historical operation archives and determine the feature extraction window width based on the steady-state duration statistics.

[0044] The granularity adjustment unit is used to obtain the confidence flag output by the most recent uniformity determination. When the confidence flag indicates that the most recent determination result is a boundary state, the length of the segmented analysis interval is reduced, and the granularity of the segmented analysis of the data is refined.

[0045] The feature extraction unit is used to segment the branch current data based on the adjusted feature extraction window width and segmented analysis granularity, and extract current fluctuation feature parameters from each segment.

[0046] According to the above scheme, the uniform evaluation module includes a weight retrieval unit, a weighted calculation unit, a benchmark retrieval unit, and a score generation unit;

[0047] The weight retrieval unit is used to retrieve the weight parameters corresponding to each branch of the temperature control equipment from the historical operation archive. The weight parameters are generated by the frequency of historical anomalies in each branch.

[0048] The weighted calculation unit is used to perform weighted calculations on the current data of each branch with the corresponding weight parameters to generate a dispersion parameter that reflects the dispersion of the branch current and the consistency of the impedance.

[0049] The benchmark retrieval unit is used to retrieve the dynamic normalized benchmark from the historical operation archives. The dynamic normalized benchmark is generated from the historical uniformity score distribution.

[0050] The scoring generation unit is used to compare and calculate the dispersion parameter with the dynamic normalization benchmark to obtain the relative deviation, and generate the uniformity pre-scoring result based on the relative deviation.

[0051] According to the above scheme, the determination backtracking module includes a threshold generation unit, a state comparison unit, an anomaly output unit, and a backtracking recalculation unit.

[0052] The threshold generation unit is used to retrieve the historical uniformity score distribution of the temperature control equipment from the historical operation archives, and use the set quantile line of the historical uniformity score distribution as the dynamic judgment threshold.

[0053] The state comparison unit is used to obtain the uniformity pre-score result and the relative deviation, compare the uniformity pre-score result with the dynamic judgment threshold, and compare the relative deviation with the fluctuation range to determine the uniformity status of the current temperature control operation.

[0054] The abnormal output unit is used to output the abnormal judgment result and the corresponding abnormal category when the uniformity pre-score result is lower than the dynamic judgment threshold and the relative deviation exceeds the floating range.

[0055] The backtracking recalculation unit is used to generate a suspected anomaly flag when the uniformity pre-score result is lower than the dynamic judgment threshold and the relative deviation does not exceed the floating range. It locates the corresponding runtime segment to perform secondary backtracking segmentation and fluctuation feature parameter extraction, recalculates the dispersion parameter, relative deviation and uniformity pre-score result and completes secondary comparison, and outputs the final judgment result based on the secondary comparison result.

[0056] Compared with the prior art, the beneficial effects of the present invention are:

[0057] 1. This invention adopts an adaptive data preprocessing mechanism, which dynamically adjusts the feature extraction window width based on historical steady-state duration statistical parameters to adapt to data features under different operating conditions and effectively improve the extraction accuracy of current fluctuation features;

[0058] 2. This invention constructs a branch weight system based on historical anomaly frequency and a dynamic judgment threshold based on historical score distribution. Combined with multi-dimensional dispersion assessment and dynamic normalization processing, the uniformity assessment can be dynamically adapted to the equipment operating status, effectively reducing the assessment deviation caused by operating condition drift and equipment aging, and improving the accuracy and rationality of the assessment results.

[0059] 3. The present invention sets up a dual-condition judgment and secondary backtracking recalculation mechanism to backtrack and recalculate and perform secondary verification on suspected abnormal states at the judgment boundary, which can effectively reduce misjudgment and missed judgment caused by a single judgment and improve the reliability and rigor of uniformity anomaly judgment. Attached Figure Description

[0060] Figure 1 This is a flowchart illustrating the steps of a temperature control device data management method based on artificial intelligence according to the present invention.

[0061] Figure 2 This is a schematic diagram of the temperature control device structure of a temperature control device data management system based on artificial intelligence according to the present invention;

[0062] Figure 3This is a schematic diagram of a temperature control device according to an artificial intelligence-based temperature control device data management system and method of the present invention. Detailed Implementation

[0063] 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.

[0064] Example: Figures 1-3 As shown, this invention provides a technical solution: a data management method for temperature control devices based on artificial intelligence, applied to intelligent temperature-controlled face masks employing a dot-matrix TEC (thermal energy dissipation) array; such as... Figure 3 As shown, the main body of the intelligent temperature-controlled mask consists of a dual-layer structure: a heat dissipation layer and a dot-matrix TEC wafer functional layer. Multiple dot-matrix TEC wafers are evenly distributed in parallel within the functional layer and connected to the host control circuit via a magnetic interface. The heat dissipation layer employs a grid-like copper-clad design for rapid and even conduction of the cold or heat generated by the TEC, while the grid structure ensures the mask's breathability. Figure 2 As shown, the intelligent temperature-controlled mask consists of an MCU (microcontroller unit) module, a gate drive module, a bridge drive module, a TEC load module, and a current acquisition module.

[0065] In this embodiment, after the user wears the smart temperature-controlled mask and starts the device, they select the ice compress mode; the MCU module enters the cooling logic, outputs two sets of non-pulse width modulation signals, which are amplified by the gate driver and then control the bridge driver to output a positive voltage to both ends of each TEC; each TEC enters the cooling state, and the cold energy is conducted to the face through the mesh copper heat dissipation layer; the current acquisition module collects the current data of each TEC in the 8 branches in real time at a sampling frequency of 1000 Hz.

[0066] A data management method for temperature control equipment based on artificial intelligence includes the following steps:

[0067] S100: Obtain the branch current data during the operation of the temperature control equipment, retrieve the steady-state duration statistics from the historical operation archive of the temperature control equipment, and adjust the feature extraction window width; obtain the confidence flag of the most recent uniformity determination; if the determination result is a boundary state, refine the segmented analysis granularity of the data; extract the current fluctuation feature parameters based on the adjusted window and segmented granularity.

[0068] S200: Retrieve the weight parameters of each branch from the historical operation archive, perform weighted processing on the current data of each branch, and obtain the dispersion parameter; retrieve the dynamic normalization benchmark from the historical operation archive, convert the dispersion parameter into a relative deviation, and generate a uniformity pre-score result.

[0069] S300: Retrieve the historical uniformity score distribution of the temperature control equipment and use the quantile line as the dynamic judgment threshold; when the pre-scoring result is lower than the dynamic judgment threshold and the relative deviation exceeds the floating range, output the abnormal judgment result and abnormal category; when the pre-scoring result is lower than the dynamic judgment threshold and the relative deviation does not exceed the floating range, generate a suspected abnormality mark, perform secondary backtracking segmentation and score recalculation for the corresponding time period, and output the final judgment result based on the recalculation result.

[0070] S400: Generate structured labels containing uniformity level, anomaly category, risk level and anomaly period based on the final judgment result, and archive the labels with the current data of the corresponding operating period to the historical operation archive.

[0071] S500. Based on the current dispersion parameter and uniformity score result, update the weight parameters of each branch; incorporate the current uniformity score result into the historical score distribution and update the dynamic judgment threshold; incorporate the current steady-state duration into the historical steady-state duration statistical parameter.

[0072] Specifically, step S100 includes:

[0073] S110. Obtain the current sampling data of each branch of the temperature control equipment during the operating period, and use it as the branch current data.

[0074] S120. Retrieve the steady-state duration statistics of the temperature control equipment from the historical operation archives, and determine the initial width as the feature extraction window width based on the steady-state duration statistics.

[0075] S130. Obtain the confidence flag output by the most recent uniformity determination. When the confidence flag indicates that the most recent determination result is a boundary state, reduce the length of the segmented analysis interval of the current data and refine the granularity of the segmented analysis of the data.

[0076] S140. Based on the feature extraction window width and segmented analysis granularity, the branch current data is segmented and extracted from each segment to extract current fluctuation characteristic parameters.

[0077] Specifically, in this embodiment, the current sampling data of the eight branches obtained include: branch 1 at 2.05 amps, branch 2 at 2.12 amps, branch 3 at 1.98 amps, branch 4 at 2.01 amps, branch 5 at 2.15 amps, branch 6 at 1.92 amps, branch 7 at 2.08 amps, and branch 8 at 2.03 amps. The steady-state duration statistics of the device were retrieved from the historical operation archive, revealing that the temperature control device had previously completed 12 effective operations. The duration of each operation after entering steady state included 18... 0 seconds, 195 seconds, 172 seconds, 188 seconds, 165 seconds, 200 seconds, 178 seconds, 192 seconds, 185 seconds, 170 seconds, 196 seconds, and 182 seconds; the arithmetic mean of the steady-state duration statistical parameters is 183.6 seconds, and the standard deviation is 9.8 seconds; the feature extraction window width is determined proportionally to the average steady-state duration, with a scaling factor of 11 sampling points / second. The longer the steady-state duration, the larger the feature statistical window, and the stronger the data stability; in this embodiment, the corresponding window width is rounded to 2048 sampling points; simultaneously obtaining... The confidence value of the most recent uniformity determination is used. This confidence value is calculated from the relative difference between the previous determination result and the dynamic determination threshold, ranging from 0 to 1. A value closer to 1 indicates a clearer determination result. When the confidence value is between 0.65 and 0.75, it indicates that the previous determination was in a boundary state. The confidence value obtained this time is 0.72, corresponding to a boundary state. The granularity adjustment unit reduces the segmented analysis interval length of the current data from 1024 sampling points to 512 sampling points to refine the granularity of the segmented analysis. Based on the adjusted feature extraction window width and segmented analysis granularity, the current data of each branch is segmented. In this embodiment, the feature extraction window width is 2048 sampling points, and the segmented analysis granularity is 512 sampling points. According to the calculation method of dividing the window width by the segmented granularity, the current data of each branch is divided into 4 segments. The feature extraction unit extracts the current fluctuation feature parameters from each segment, including the maximum value, minimum value, average value, standard deviation and waveform slope of the current in each segment. This is only an example and is not a limitation.

[0078] Specifically, step S200 includes:

[0079] S210. Retrieve the weight parameters corresponding to each branch of the temperature control equipment from the historical operation archive. The weight parameters are generated by the frequency of historical anomalies in each branch.

[0080] S220. The current data of each branch is weighted and calculated with the corresponding weight parameters of each branch to generate a dispersion parameter that reflects the dispersion of the current of each branch and the consistency of impedance.

[0081] S230. Retrieve the dynamic normalized benchmark from the historical operation archives. The dynamic normalized benchmark is generated from the historical uniformity score distribution.

[0082] S240. Compare and calculate the discreteness parameter with the dynamic normalization benchmark to obtain the relative deviation degree that characterizes the current uniformity deviation degree, and generate the uniformity pre-score result based on the relative deviation degree.

[0083] Specifically, the weight parameters corresponding to the eight branches of the temperature control equipment are retrieved from the historical operation archives. The weight parameters of each branch are generated by weighting the base weight and the frequency of historical anomalies, with the sum of all weights being 1. The higher the frequency of anomalies, the greater the weight percentage of the branch. All branches are set with a uniform base weight, and then the weight is adjusted according to the frequency of historical anomalies. In 12 historical operations of the temperature control equipment, the frequencies of each branch being marked as anomaly are as follows: branch 1: 1 time, branch 2: 0 times, branch 3: 3 times, branch 4: 0 times, branch 5: 0 times, branch 6: 0 times, branch 7: 0 times, branch 8: 0 times, branch 9: 0 times, branch 10: 0 times, branch 10: 0 times, branch 11: 0 times, branch 12: 0 times, branch 13: 0 times, branch 14: 0 times, branch 15: 0 times, branch 16: 0 times, branch 17: 0 times, branch 18: 0 times, branch 19: 0 times, branch 10: 0 times, branch 12: 0 times, branch 19: 0 times, branch 10 ... Branch 1 had 0 occurrences, branch 2 had 2 occurrences, branch 7 had 0 occurrences, and branch 8 had 0 occurrences; the total number of anomalies was 6. After assigning weights according to the anomaly frequency, the weight parameters for the 8 branches were: branch 1: 0.14, branch 2: 0.08, branch 3: 0.30, branch 4: 0.08, branch 5: 0.08, branch 6: 0.24, branch 7: 0.08, and branch 8: 0.08. The current data of each branch were weighted and calculated with the corresponding weight parameters to generate the dispersion parameter. The formula for calculating the dispersion parameter is: D = (∑ i∈(1,n) w i ×(I i -I) 2 ) 1 / 2 Where D represents the weighted current dispersion; n represents the total number of branches, in this embodiment n=8; i represents the branch number, in this embodiment i=1, 2, ..., 8; w i I represents the weight parameter of the i-th branch; i Let I represent the current value of the i-th branch; let I represent the arithmetic mean of the currents in each branch; in this operation, the arithmetic mean of the currents in each branch is I = 2.0425 Amperes, and the weighted current dispersion D is calculated to be approximately 0.087 Amperes; the dynamic normalization benchmark is retrieved from the historical operation archive; this dynamic normalization benchmark is generated from the historical uniformity score distribution and is 0.075 Amperes; this dynamic normalization benchmark characterizes the average dispersion of the equipment under historical normal operation conditions; the scoring generation unit compares the weighted current dispersion D with the dynamic normalization benchmark B to obtain the relative deviation R, calculated as: R = (DB) / B; where R represents the relative deviation, D represents the weighted current dispersion in this operation, and B represents the dynamic normalization benchmark; the relative deviation R is obtained as: R = (0.087 - 0.075) / 0.075 = 0.16; the uniformity pre-scoring result P is generated based on the relative deviation R = 0.16; the pre-scoring result P is calculated as: P = P base -α×R; where P baseThe base score is 100 points, α represents the score adjustment coefficient (α=50 in this example), and R represents the relative deviation. The uniformity pre-score is obtained as P=100-50×0.16=92. Therefore, the uniformity pre-score for this run is 92 points. This is only an example and is not a limitation.

[0084] Specifically, step S300 includes:

[0085] S310. Retrieve the historical uniformity score distribution of the temperature control equipment from the historical operation archive, and use the set quantile line of the historical uniformity score distribution as the dynamic judgment threshold.

[0086] S320. Obtain the uniformity pre-score result and relative deviation, compare the uniformity pre-score result with the dynamic judgment threshold, compare the relative deviation with the fluctuation range, and determine the uniformity status of the current temperature control operation.

[0087] S330. When the uniformity pre-score is lower than the dynamic judgment threshold and the relative deviation exceeds the floating range, output the anomaly judgment result and the corresponding anomaly category.

[0088] S340. When the uniformity pre-score is lower than the dynamic judgment threshold and the relative deviation does not exceed the floating range, a suspected anomaly marker is generated.

[0089] S350. Based on the suspected anomaly marker, locate the running segment corresponding to the uniformity pre-score result, perform secondary backtracking segmentation and fluctuation characteristic parameter extraction on the branch current data of the running segment, recalculate the dispersion parameter, relative deviation and uniformity pre-score result based on the re-extracted fluctuation characteristic parameter, and perform secondary comparison between the uniformity pre-score result and the dynamic judgment threshold, as well as the relative deviation and floating range, and output the final judgment result based on the secondary comparison result.

[0090] Specifically, the historical uniformity score distribution of the temperature control equipment was retrieved from the historical operation records. The uniformity scores for 12 historical operations of the temperature control equipment were: 88, 91, 85, 89, 78, 92, 86, 90, 87, 84, 93, and 88. The 25th percentile of the historical score distribution, i.e., the lower quartile, was used as the dynamic judgment threshold. The quartile value corresponding to the 25th percentile of the 12 scores was 86. The uniformity pre-score result was obtained as P=92 and the relative deviation R=0. 16. Compare the uniformity pre-score P with the dynamic judgment threshold T=86 points, and compare the relative deviation R with the fluctuation range F; in this embodiment, the fluctuation range F is set to 0.30; since P=92 points is not lower than T=86 points, the uniformity pre-score result is not lower than the dynamic judgment threshold, and the uniformity status of the current temperature control operation is judged to be normal; in this embodiment, no abnormal or suspected abnormal judgment is triggered during this running period, and the judgment result of normal uniformity status is directly output without the need to perform secondary backtracking. This is only an example and is not a limitation.

[0091] Specifically, step S400 includes:

[0092] S410. Obtain the final determination result of the uniformity status, and extract information on uniformity level, anomaly category, risk level and anomaly period.

[0093] S420. Combine uniformity level, anomaly category, risk level and anomaly period to generate structured labels, associate the structured labels with the branch current data of the corresponding operating period to form associated archived data.

[0094] S430. Archive the associated structured tags and branch current data to the historical operation archive.

[0095] Specifically, the final judgment result of the uniformity status is normal, the uniformity level is extracted as good, there are no abnormal categories, the risk level is low, and there are no abnormal periods. Based on the final judgment result, the uniformity level is determined to be good, that is, a score ≥90 and <95 is good. Based on the abnormal category, the risk level is determined to be low, that is, the risk level is low when there are no abnormalities. Based on the running period, that is, the 12th run, the duration is 185 seconds, and the abnormal period is determined to be no. The uniformity level of good, the abnormal category of no abnormality, the risk level of low, and the abnormal period are combined to generate a structured tag, in the format: {Equipment number: TEC-2026-001, Run sequence number: 12, Uniformity level: good, Abnormal category: none, Risk level: low, Abnormal period: none, Score: 92}. The structured tag is associated with the current data of the 8 branches in this running period to form associated archived data, which is stored in the historical running archive. This is only an example and is not a limitation.

[0096] Specifically, step S500 includes:

[0097] S510. Obtain the dispersion parameters and uniformity score results generated in this operation. Based on the dispersion parameters and uniformity score results, perform iterative correction on the weight parameters of each branch of the temperature control equipment in the historical operation file to complete the update of the weight parameters of each branch.

[0098] S520. The uniformity score result of this run is added to the statistical sequence of the historical uniformity score distribution to form an updated historical uniformity score distribution. Based on the set quantile line of the updated historical uniformity score distribution, the dynamic judgment threshold is updated accordingly.

[0099] S530. Incorporate the steady-state duration data of this run into the cumulative statistical dimension of the historical steady-state duration statistical parameters, and complete the update of the historical steady-state duration statistical parameters.

[0100] Specifically, the dispersion parameter D=0.087 amperes and the uniformity score P=92 points generated in this operation are obtained; based on the dispersion parameter D=0.087 amperes and the uniformity score P=92 points, the weight parameters of each branch in the historical operation archive are iteratively corrected; in this embodiment, the weight parameter update adopts the exponential weighted moving average algorithm, and the calculation formula is: w i-new =β×w i-old +(1-β)×f i Among them, w i-new w represents the updated weight parameters of the i-th branch; i-old represents the weight parameters of the i-th branch before the update; β represents the smoothing coefficient, which is 0.85 in this embodiment; f i This represents the anomaly indicator factor for the i-th branch in this operation. If the branch is determined to be abnormal in this operation, then f... i =1, otherwise f i =0; During this operation, none of the 8 branches were identified as abnormal, therefore the f of all branches is 0. i=0; The weight parameters of each branch have been slightly reduced after iterative correction, and the update is complete; The uniformity score result P=92 points of this run is added to the statistical sequence of the historical uniformity score distribution; The updated historical uniformity score distribution contains 13 scores: 88, 91, 85, 89, 78, 92, 86, 90, 87, 84, 93, 88, and 92 points; Based on the 25th percentile of the updated historical uniformity score distribution, the corresponding... Update the dynamic judgment threshold; the 25th percentile of the 13 scores corresponds to a percentile value of 87, so the dynamic judgment threshold is updated from 86 to 87; the steady-state duration of this run, 185 seconds, is included in the cumulative statistical dimension of the historical steady-state duration statistical parameters; the updated steady-state duration statistical parameters contain 13 data points, with their arithmetic mean updated to 183.7 seconds and standard deviation updated to 9.5 seconds, completing the update of the historical steady-state duration statistical parameters; this is only an example and is not a limitation.

[0101] In another embodiment, when the user selects the heat therapy mode to operate the device, the microcontroller outputs two sets of non-pulse width modulation signals to control the bridge drive to output reverse voltage to both ends of each semiconductor cooling chip, and the device enters the heating state; the current acquisition module acquires current sampling data from 8 branches, including 2.11 amps for branch 1, 2.18 amps for branch 2, 2.04 amps for branch 3, 2.07 amps for branch 4, 2.21 amps for branch 5, 1.98 amps for branch 6, 2.14 amps for branch 7, and 2.09 amps for branch 8; the arithmetic mean of the current in each branch is I = 2.1025 amps; the weighted calculation unit calculates the weighted current discrete value based on the weight parameters of each branch. Degree D≈0.085 Amperes; Current dynamic normalized baseline B=0.075 Amperes; Relative deviation R=(0.085-0.075) / 0.075≈0.133; Pre-scoring result P=100-50×0.133≈93.35 points; Obtain the pre-scoring result P≈93.35 points and relative deviation R≈0.133; Current dynamic judgment threshold T=87 points, fluctuation range F=0.30; Compare P and T, 93.35 points is not lower than 87 points; Therefore, the pre-scoring result is not lower than the dynamic judgment threshold, and the uniformity status under the current heat therapy mode is judged to be normal, without triggering abnormal or suspected abnormal judgment. This is only an example and is not a limitation.

[0102] In another embodiment, after the user wears the smart temperature-controlled mask and starts the device, selecting the ice-cold compress mode, an anomaly detection is triggered during one operation. After the user selects the ice-cold compress mode and starts the device, the current acquisition module obtains current sampling data from eight branches, including: branch 1 at 2.35 amps, branch 2 at 1.45 amps, branch 3 at 2.28 amps, branch 4 at 2.31 amps, branch 5 at 2.40 amps, branch 6 at 1.38 amps, branch 7 at 2.33 amps, and branch 8 at 2. 26 Amperes; Arithmetic mean of current in each branch I = 2.095 Amperes; Weighted current dispersion D = 0.168 Amperes calculated by the weighted calculation unit; Current dynamic normalization benchmark B = 0.075 Amperes; Relative deviation R = (0.168 - 0.075) / 0.075 = 1.24; Preliminary score P = 100 - 50 × 1.24 = 38 points; Obtain the preliminary score P = 38 points and relative deviation R = 1.24; Current dynamic judgment threshold T = 87 points. The floating range F=0.30; comparing P and T: 38 points is lower than 87 points; comparing R and F: 1.24 exceeds 0.30; since the uniformity pre-score result is lower than the dynamic judgment threshold and the relative deviation exceeds the floating range, the abnormal judgment result is output. The abnormality category is that the current of branch 2 and branch 6 is significantly lower, and the local cooling is insufficient; the tag archiving module generates a structured tag according to the abnormal judgment result, with the format: {Equipment number: TEC-2026-001, operation sequence number: 14, uniformity level: poor, abnormality category: local cooling insufficient, risk level: high, abnormal time period: 30 seconds to 60 seconds after startup, score: 38}; the structured tag is associated with the current data of this operation period and archived to the historical operation archive; based on the current dispersion parameter D=0.168 amperes and the uniformity score P=38 points, the three-way parameter update is performed; in the weight parameter update, the abnormality indicator factors f2=1 and f6=1 for branch 2 and branch 6, respectively, and the f6=1 for the other branches. i =0. After being updated by the exponentially weighted moving average algorithm, the weight parameters of branch 2 and branch 6 are increased, and the weight parameters of the other branches are adjusted accordingly; the dynamic judgment threshold is updated to 88 points; the steady-state duration statistics parameters are updated synchronously. This is only an example and is not a limitation.

[0103] In another embodiment, a suspected anomaly detection and secondary backtracking process is triggered during a single operation. Specifically, after the user selects the ice pack mode to start the device, current sampling data for eight branches are obtained: branch 1 is 2.08 amps, branch 2 is 1.92 amps, branch 3 is 2.01 amps, branch 4 is 2.12 amps, branch 5 is 1.89 amps, branch 6 is 2.05 amps, branch 7 is 1.95 amps, and branch 8 is 2.10 amps; the arithmetic mean of the current in each branch is I = 2.015 amps; the weighted calculation unit calculates the weighted current dispersion D = 0.096 amps; the current dynamic normalization benchmark B = 0.075 amps; the relative deviation R = (0.096 - 0.075) / 0.075 = 0.28; Pre-scoring result P = 100 - 50 × 0.28 = 86 points; Obtain the pre-scoring result P = 86 points and relative deviation R = 0.28; Current dynamic judgment threshold T = 87 points, fluctuation range F = 0.30; Compare P and T, 86 points is lower than 87 points; Compare R and F, 0.28 does not exceed 0.30; Since the uniformity pre-scoring result is lower than the dynamic judgment threshold and the relative deviation does not exceed the fluctuation range, the current state is judged as a suspected anomaly, and a suspected anomaly label is generated; The backtracking recalculation unit locates the running segment corresponding to the pre-scoring result based on the suspected anomaly label, that is, the 45th second to the 75th second of this run, and performs secondary backtracking segmentation on the branch current data of this period; The window width of the secondary backtracking is increased from... The number of sampling points was reduced from 2048 to 1024, and the segmentation granularity was refined from 512 to 256. This allowed for finer segmentation and extraction of fluctuation characteristic parameters from data in suspected abnormal periods. Based on the re-extracted fluctuation characteristic parameters, the backtracking recalculation unit recalculated the dispersion parameter D' = 0.089 amperes, the relative deviation R' = (0.089 - 0.075) / 0.075 ≈ 0.187, and the uniformity pre-score P' = 100 - 50 × 0.187 = 90.65 points. The recalculated uniformity pre-score P' = 90.65 points was then compared with the dynamic judgment threshold T = 87 points. A score of 90.65 was not lower than 87 points. The recalculated relative deviation R' = A second comparison was performed between 0.187 and the floating range F=0.30. 0.187 did not exceed 0.30. Since the pre-scoring result in the second comparison was not lower than the dynamic judgment threshold and the relative deviation did not exceed the floating range, the final judgment result was output as normal based on the second comparison result. The tag archiving module generated a structured tag based on the final judgment result, in the format: {Equipment Number: TEC-2026-001, Operation Sequence Number: 15, Uniformity Level: Good, Anomaly Category: None, Risk Level: Low, Anomaly Period: None, Score: 90.65}. Based on the current dispersion parameter D'=0.089 amperes and the uniformity score P'=90.65, three-way parameter updates were performed, with the dynamic judgment threshold updated from 87 to 87.5 points; This embodiment effectively avoids misjudgments caused by single data fluctuations through a secondary backtracking and recalculation mechanism; if an anomaly is output directly based on the initial judgment result without secondary backtracking, it will cause a false alarm; after refined analysis through secondary backtracking, it is confirmed that the current operation is actually normal, improving the accuracy and reliability of the judgment result.

[0104] This invention provides a technical solution: an AI-based temperature control equipment data management system. This system is embedded within an MCU module as an intelligent temperature control data management functional module, comprising a preprocessing module, a uniformity evaluation module, a judgment backtracking module, a tag archiving module, and an iterative update module. The preprocessing module acquires branch current data during the operation of the temperature control equipment, retrieves steady-state duration statistics from historical operation records to adjust the feature extraction window width, refines the data segmentation analysis granularity under boundary conditions based on the confidence marker of the most recent uniformity judgment, and extracts current fluctuation feature parameters based on the adjusted window and segmentation granularity. The uniformity evaluation module retrieves the weight parameters of each branch from historical operation records, performs weighted processing on the current data of each branch to obtain a dispersion parameter, and retrieves a dynamic normalization benchmark to convert the dispersion parameter into a relative deviation. The system comprises four modules: a uniformity pre-score module, a judgment backtracking module, and a tag archiving module. The tag archiving module generates structured tags containing uniformity level, anomaly category, risk level, and anomaly period based on the final judgment result. The tags are associated with the current data of the corresponding operating period and archived in the historical operating archive. The iterative update module updates the weight parameters of each branch based on the current dispersion parameter and uniformity score result. The current uniformity score result is included in the historical score distribution to update the dynamic judgment threshold. The steady-state duration of the current operation is included in the historical steady-state duration statistical parameter.

[0105] Specifically, the preprocessing module includes a data acquisition unit, a window adjustment unit, a granularity adjustment unit, and a feature extraction unit. The data acquisition unit is used to acquire current sampling data of each branch of the temperature control equipment during operation, as branch current data. The window adjustment unit is used to retrieve the steady-state duration statistics of the temperature control equipment from historical operation records and determine the feature extraction window width based on the steady-state duration statistics. The granularity adjustment unit is used to acquire the confidence flag output by the most recent uniformity judgment. When the confidence flag indicates that the most recent judgment result is a boundary state, the length of the segmented analysis interval is reduced, and the granularity of the segmented analysis of the data is refined. The feature extraction unit is used to segment the branch current data based on the adjusted feature extraction window width and segmented analysis granularity, and extract current fluctuation feature parameters from each segment.

[0106] Specifically, the uniformity assessment module includes a weight retrieval unit, a weighted calculation unit, a benchmark retrieval unit, and a score generation unit. The weight retrieval unit retrieves the weight parameters corresponding to each branch of the temperature control equipment from historical operation records. The weight parameters are generated by the frequency of historical anomalies in each branch. The weighted calculation unit performs weighted calculations on the current data of each branch with the corresponding weight parameters to generate a dispersion parameter reflecting the dispersion of the branch current and the consistency of impedance. The benchmark retrieval unit retrieves a dynamically normalized benchmark from historical operation records. The dynamic normalized benchmark is generated by the historical uniformity score distribution. The score generation unit compares and calculates the dispersion parameter with the dynamic normalized benchmark to obtain the relative deviation, and generates a uniformity pre-score based on the relative deviation.

[0107] Specifically, the judgment backtracking module includes a threshold generation unit, a state comparison unit, an anomaly output unit, and a backtracking recalculation unit. The threshold generation unit retrieves the historical uniformity score distribution of the temperature control equipment from historical operation records, using the set quantile of the historical uniformity score distribution as the dynamic judgment threshold. The state comparison unit obtains the uniformity pre-score result and relative deviation, compares the uniformity pre-score result with the dynamic judgment threshold, and compares the relative deviation with the fluctuation range to determine the current uniformity status of the temperature control operation. The anomaly output unit outputs an anomaly judgment result and the corresponding anomaly category when the uniformity pre-score result is lower than the dynamic judgment threshold and the relative deviation exceeds the fluctuation range. The backtracking recalculation unit generates a suspected anomaly flag when the uniformity pre-score result is lower than the dynamic judgment threshold and the relative deviation does not exceed the fluctuation range. It then locates the corresponding operating segment, performs secondary backtracking segmentation and fluctuation feature parameter extraction, recalculates the dispersion parameter, relative deviation, and uniformity pre-score result, completes a secondary comparison, and outputs the final judgment result based on the secondary comparison result.

[0108] In this embodiment, the intelligent temperature-controlled mask consists of an MCU module, a gate drive module, a bridge drive module, a TEC load module, and a current acquisition module. The hardware of the intelligent temperature-controlled mask works in concert as follows:

[0109] After wearing the mask, the user selects either the ice compress mode or the hot compress mode using the mode selection button. When the user presses the cooling switch, the MCU unit enters the cooling logic and outputs two sets of non-pulse width modulation signals. After being amplified by the gate drive module, the two sets of signals control the bridge drive module to output a positive voltage to both ends of each semiconductor cooling chip. Under the positive voltage drive, each semiconductor cooling chip enters the cooling state, its cold end temperature decreases, and the cooling energy is quickly and evenly conducted to the facial skin through the grid-like copper-clad structure of the heat dissipation layer.

[0110] When the user presses the heating switch, the MCU unit enters the heating logic and outputs two sets of non-pulse width modulation signals. After being amplified by the gate drive, the signals control the bridge drive module to output a reverse voltage to both ends of each thermoelectric cooler. Under the reverse voltage drive, each thermoelectric cooler enters the heating state, and its hot end temperature rises. The heat is conducted to the facial skin through the grid-like copper-clad structure of the heat dissipation layer.

[0111] During cooling or heating, the current acquisition module continuously collects current data from each branch of the semiconductor cooling chip at a frequency of 1000 Hz and transmits the collected current data to the MCU unit. The microcontroller unit executes the temperature control equipment data management method through its built-in intelligent temperature control data management function module, dynamically adjusting the duty cycle of the pulse width modulation signal according to the analysis results to control the cooling or heating power and temperature. At the same time, the MCU unit feeds back the judgment result to the user through an indicator light or display screen. When the judgment result is abnormal, the indicator light turns red and emits a warning sound; when the judgment result is suspected to be abnormal, the indicator light turns yellow and prompts the user to pay attention; when the judgment result is normal, the indicator light turns green. This is only an example and is not a limitation.

[0112] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A data management method for temperature control equipment based on artificial intelligence, characterized in that: The method includes: S100: Obtain the branch current data during the operation of the temperature control equipment, retrieve the steady-state duration statistics from the historical operation archive of the temperature control equipment, and adjust the feature extraction window width; obtain the confidence flag of the most recent uniformity determination; if the determination result is a boundary state, refine the segmented analysis granularity of the data; extract the current fluctuation feature parameters based on the adjusted window and segmented granularity. S200: Retrieve the weight parameters of each branch from the historical operation archive, perform weighted processing on the current data of each branch to obtain the dispersion parameter; retrieve the dynamic normalization benchmark from the historical operation archive, convert the dispersion parameter into a relative deviation, and generate a uniformity pre-score result. S300: Retrieve the historical uniformity score distribution of the temperature control equipment and use the quantile line as the dynamic judgment threshold; when the pre-scoring result is lower than the dynamic judgment threshold and the relative deviation exceeds the floating range, output the abnormal judgment result and abnormal category; when the pre-scoring result is lower than the dynamic judgment threshold and the relative deviation does not exceed the floating range, generate a suspected abnormality mark, perform secondary backtracking segmentation and score recalculation for the corresponding time period, and output the final judgment result based on the recalculation result. S400. Generate a structured label containing uniformity level, anomaly category, risk level and anomaly period based on the final judgment result, and associate the label with the current data of the corresponding running period and archive it to the historical running archive. S500. Based on the current dispersion parameter and uniformity score result, update the weight parameters of each branch; incorporate the current uniformity score result into the historical score distribution and update the dynamic judgment threshold; incorporate the current steady-state duration into the historical steady-state duration statistical parameter.

2. The data management method for temperature control equipment based on artificial intelligence according to claim 1, characterized in that: Step S100 includes: S110. Obtain the current sampling data of each branch of the temperature control equipment during the operating period, and use it as the branch current data. S120. Retrieve the steady-state duration statistics of the temperature control device from the historical operation archive, and determine the initial width as the feature extraction window width based on the steady-state duration statistics. S130. Obtain the confidence flag output by the most recent uniformity determination. When the confidence flag indicates that the most recent determination result is a boundary state, reduce the length of the segmented analysis interval of the current data and refine the granularity of the segmented analysis of the data. S140. Based on the feature extraction window width and segmented analysis granularity, the branch current data is segmented and extracted from each segment to extract current fluctuation feature parameters.

3. The data management method for temperature control equipment based on artificial intelligence according to claim 1, characterized in that: Step S200 includes: S210. Retrieve the weight parameters corresponding to each branch of the temperature control equipment from the historical operation archive. The weight parameters are generated by the frequency of historical anomalies in each branch. S220. The current data of each branch is weighted and calculated with the corresponding weight parameters of each branch to generate a dispersion parameter that reflects the dispersion of the current of each branch and the consistency of impedance. S230. Retrieve the dynamic normalization benchmark from the historical operation archives. The dynamic normalization benchmark is generated from the historical uniformity score distribution. S240. The discreteness parameter is compared and calculated with the dynamic normalization benchmark to obtain the relative deviation degree characterizing the current uniformity deviation degree, and a uniformity pre-score result is generated based on the relative deviation degree.

4. The data management method for temperature control equipment based on artificial intelligence according to claim 1, characterized in that: Step S300 includes: S310. Retrieve the historical uniformity score distribution of the temperature control equipment from the historical operation archive, and use the set quantile line of the historical uniformity score distribution as the dynamic judgment threshold. S320. Obtain the uniformity pre-score result and relative deviation, compare the uniformity pre-score result with the dynamic judgment threshold, compare the relative deviation with the fluctuation range, and determine the uniformity status of the current temperature control operation. S330. When the uniformity pre-score is lower than the dynamic judgment threshold and the relative deviation exceeds the floating range, output the anomaly judgment result and the corresponding anomaly category. S340. When the uniformity pre-score is lower than the dynamic judgment threshold and the relative deviation does not exceed the floating range, a suspected anomaly marker is generated. S350. Locate the running segment corresponding to the uniformity pre-scoring result based on the suspected anomaly identifier, perform secondary backtracking segmentation and fluctuation characteristic parameter extraction on the branch current data of the running segment, recalculate the dispersion parameter, relative deviation and uniformity pre-scoring result based on the re-extracted fluctuation characteristic parameter, and perform secondary comparison between the uniformity pre-scoring result and the dynamic judgment threshold, as well as the relative deviation and floating range, and output the final judgment result based on the secondary comparison result.

5. The data management method for temperature control equipment based on artificial intelligence according to claim 1, characterized in that: Step S400 includes: S410. Obtain the final determination result of the uniformity status, and extract information on uniformity level, anomaly category, risk level and anomaly period. S420. Combine the uniformity level, the anomaly category, the risk level, and the anomaly period to generate a structured label, and associate the structured label with the branch current data of the corresponding operating period to form associated archived data. S430. Archive the associated structured tags and branch current data to the historical operation archive.

6. The data management method for temperature control equipment based on artificial intelligence according to claim 1, characterized in that: Step S500 includes: S510. Obtain the dispersion parameters and uniformity score results generated in this operation. Based on the dispersion parameters and uniformity score results, perform iterative correction on the weight parameters of each branch of the temperature control equipment in the historical operation file to complete the update of the weight parameters of each branch. S520. The uniformity score result of this run is added to the statistical sequence of the historical uniformity score distribution to form an updated historical uniformity score distribution. Based on the set quantile line of the updated historical uniformity score distribution, the dynamic judgment threshold is updated accordingly. S530. Incorporate the steady-state duration data of this run into the cumulative statistical dimension of the historical steady-state duration statistical parameters, and complete the update of the historical steady-state duration statistical parameters.

7. A data management system for temperature control equipment based on artificial intelligence, characterized in that: The system includes: a preprocessing module, a uniform evaluation module, a decision backtracking module, a label archiving module, and an iterative update module; The preprocessing module is used to acquire branch current data during the operation of the temperature control equipment, retrieve steady-state duration statistics from historical operation files to adjust the feature extraction window width, refine the data segmentation analysis granularity under boundary conditions based on the confidence flag of the most recent uniformity determination, and extract current fluctuation feature parameters based on the adjusted window and segmentation granularity. The uniformity evaluation module is used to retrieve the weight parameters of each branch in the historical operation archive, perform weighted processing on the current data of each branch to obtain the dispersion parameter, retrieve the dynamic normalization benchmark to convert the dispersion parameter into a relative deviation, and generate a uniformity pre-score result. The judgment backtracking module is used to retrieve the historical uniformity score distribution of the temperature control equipment, use the quantile line as the dynamic judgment threshold, and determine the uniformity status by comparing the pre-scoring result with the relative deviation under two conditions. It performs secondary backtracking segmentation and score recalculation for suspected abnormal periods, and outputs the final judgment result based on the recalculation result. The tag archiving module is used to generate structured tags containing uniformity level, anomaly category, risk level and anomaly period according to the final judgment result, and to associate the tags with the current data of the corresponding running period and archive them to the historical running archive. The iterative update module is used to update the weight parameters of each branch based on the current dispersion parameter and the uniformity score result, incorporate the current uniformity score result into the historical score distribution to update the dynamic judgment threshold, and incorporate the current steady-state duration into the historical steady-state duration statistical parameter.

8. The temperature control equipment data management system based on artificial intelligence according to claim 7, characterized in that: The preprocessing module includes a data acquisition unit, a window adjustment unit, a granularity adjustment unit, and a feature extraction unit; The data acquisition unit is used to acquire current sampling data of each branch of the temperature control equipment during the operation period, as branch current data; The window adjustment unit is used to retrieve the steady-state duration statistics of the temperature control equipment from the historical operation archives, and determine the feature extraction window width based on the steady-state duration statistics. The granularity adjustment unit is used to obtain the confidence flag output by the most recent uniformity determination. When the confidence flag indicates that the most recent determination result is a boundary state, the length of the segmented analysis interval is reduced and the granularity of the segmented analysis of the data is refined. The feature extraction unit is used to segment the branch current data based on the adjusted feature extraction window width and segmented analysis granularity, and extract current fluctuation feature parameters from each segment.

9. The temperature control equipment data management system based on artificial intelligence according to claim 7, characterized in that: The uniform evaluation module includes a weight retrieval unit, a weighted calculation unit, a benchmark retrieval unit, and a score generation unit; The weight retrieval unit is used to retrieve the weight parameters corresponding to each branch of the temperature control equipment from the historical operation archive. The weight parameters are generated by the frequency of historical anomalies in each branch. The weighted calculation unit is used to perform weighted calculations on the current data of each branch and the corresponding weight parameters to generate a dispersion parameter that reflects the dispersion of the branch current and the consistency of the impedance. The benchmark retrieval unit is used to retrieve a dynamic normalized benchmark from historical operation archives. The dynamic normalized benchmark is generated from the historical uniformity score distribution. The scoring generation unit is used to compare and calculate the dispersion parameter with the dynamic normalization benchmark to obtain the relative deviation, and generate a uniformity pre-scoring result based on the relative deviation.

10. A temperature control equipment data management system based on artificial intelligence according to claim 7, characterized in that: The determination and backtracking module includes a threshold generation unit, a state comparison unit, an anomaly output unit, and a backtracking recalculation unit; The threshold generation unit is used to retrieve the historical uniformity score distribution of the temperature control equipment from the historical operation archive, and use the set quantile line of the historical uniformity score distribution as the dynamic judgment threshold. The state comparison unit is used to obtain the uniformity pre-score result and the relative deviation, compare the uniformity pre-score result with the dynamic judgment threshold, compare the relative deviation with the fluctuation range, and determine the uniformity state of the current temperature control operation. The abnormal output unit is used to output the abnormal judgment result and the corresponding abnormal category when the uniformity pre-score result is lower than the dynamic judgment threshold and the relative deviation exceeds the floating range. The backtracking recalculation unit is used to generate a suspected anomaly flag when the uniformity pre-score result is lower than the dynamic judgment threshold and the relative deviation does not exceed the floating range. It then locates the corresponding runtime segment to perform secondary backtracking segmentation and fluctuation feature parameter extraction, recalculates the dispersion parameter, relative deviation and uniformity pre-score result and completes secondary comparison, and outputs the final judgment result based on the secondary comparison result.

Images