Air conditioning unit refrigerant quantity anomaly detection method based on digital twinning and AI robot

Abstract

The application discloses an air conditioning unit refrigerant quantity anomaly detection method based on digital twinning and an AI robot, and belongs to the technical field of air conditioning unit detection. The application constructs a digital twin corresponding to the air conditioning unit to be detected, controls the AI robot to move autonomously and collect operation state data of the air conditioning unit to be detected, calculates the residual error between the actual operation state data and ideal simulation data of the air conditioning unit to be detected, determines whether the air conditioning unit to be detected has refrigerant quantity anomaly, cooperatively collects operation state data by multiple devices, improves the overall accuracy of refrigerant flow state analysis, ensures the positioning accuracy in the complex environment of the machine room through the autonomous movement and double positioning of the AI robot, collects real-time working condition data to drive the digital twin to generate ideal simulation data to obtain comprehensive residual error, confirms the missing quantity through digital twin simulation verification, constructs a missing quantity mapping relationship library based on a large amount of experimental data, converts the corrected residual error into specific refrigerant missing quantity, and realizes anomaly degree quantification.

Description

Technical Field

[0001] This invention relates to the field of air conditioning unit testing technology, and in particular to a method for detecting abnormal refrigerant levels in air conditioning units based on digital twins and AI robots. Background Technology

[0002] During the operation of a computer room air conditioning system, the amount of refrigerant is a core parameter that directly affects the heat exchange efficiency and operational stability of the unit. Insufficient refrigerant or leakage can lead to problems such as uneven evaporator frosting, abnormal return pipe temperature, and abnormal compressor noise. Long-term operation will not only increase energy consumption, but may also cause unit failure and shutdown, resulting in serious consequences such as heat dissipation failure of computer room equipment and data loss.

[0003] Currently, the industry mainly relies on traditional manual inspections and single-point sampling methods to detect refrigerant levels in air conditioning units. Each air conditioning unit is checked one by one, and the refrigerant level is judged by empirical methods such as touching the return pipe to sense the temperature, observing the evaporator frost, and reading the pressure gauge. The inspection efficiency is low and it is difficult to meet the needs of real-time monitoring. The test results are highly dependent on the experience of the personnel. Different inspectors have different standards for judging the uniformity of frost and temperature abnormalities, which can easily lead to misjudgment or omission. Moreover, manual sampling is mostly single-point testing, which cannot obtain multi-dimensional data such as evaporator temperature distribution and overall refrigerant flow status, making it difficult to accurately locate the degree of refrigerant abnormality and the location of the leak.

[0004] Meanwhile, although fixed sensors are introduced for data collection in some scenarios, the sensors need to be pre-installed in specific locations on the unit, which cannot be flexibly adapted to different models of units. Furthermore, due to the limitations of the installation location, they can only monitor data in local areas and cannot fully reflect the overall refrigerant operating status of the unit. Summary of the Invention

[0005] The purpose of this invention is to provide a method for detecting abnormal refrigerant levels in air conditioning units based on digital twins and AI robots, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting abnormal refrigerant levels in air conditioning units based on digital twins and AI robots, comprising the following steps:

[0007] Obtain basic information about the air conditioning unit to be tested in the target computer room, and construct a digital twin of the air conditioning unit to be tested;

[0008] The AI ​​robot is controlled to carry the detection equipment and move autonomously within the target computer room, locate each air conditioning unit to be tested, and use the detection equipment to collect the operating status data of the air conditioning units to be tested;

[0009] The collected operating status data is input into the preset analysis model to obtain the refrigerant flow status analysis results of the air conditioning unit under test;

[0010] Retrieve ideal simulation data of the digital twin under the same operating conditions and standard refrigerant quantity as the air conditioning unit under test, calculate the residual between the actual operating status data of the air conditioning unit under test and the ideal simulation data; determine whether the residual continuously exceeds the preset normal threshold, if it does, determine that the air conditioning unit under test has an abnormal refrigerant quantity.

[0011] Furthermore, the step of obtaining basic information about the air conditioning unit to be tested in the target computer room and constructing a digital twin of the air conditioning unit to be tested includes:

[0012] The layout drawings of the target computer room and the equipment parameters of the air conditioning unit to be tested are retrieved through the computer room management system. The equipment parameters include the unit model, rated refrigerant capacity, evaporator specifications and return pipe size.

[0013] Based on the layout drawings, a three-dimensional spatial model of the target computer room is established, and the coordinates of the unit are marked in the three-dimensional spatial model corresponding to the actual installation position of the air conditioning unit to be tested.

[0014] Input the equipment parameters of the air conditioning unit to be tested into the three-dimensional space model to generate an initial digital model that includes the internal structure of the unit and the refrigerant circulation path;

[0015] Historical operating data of the air conditioning unit under test under standard refrigerant volume and different operating conditions are obtained, and the parameters of the initial digital model are calibrated to obtain a digital twin that is precisely matched with the air conditioning unit under test.

[0016] Furthermore, the AI ​​robot, carrying the detection equipment, autonomously moves within the target computer room, locating each air conditioning unit to be tested, including:

[0017] The three-dimensional spatial model of the target computer room and the unit coordinates of the air conditioning unit to be tested are imported into the navigation system of the AI ​​robot to generate an autonomous movement path. The autonomous movement path avoids equipment obstacles and personnel passages in the computer room.

[0018] The AI ​​robot activates its LiDAR to scan the surrounding environment, acquires environmental point cloud data in real time, compares it with a preset 3D spatial model, and corrects its movement trajectory.

[0019] When the AI ​​robot moves to the preset distance range of the air conditioning unit to be inspected, the camera is activated to capture images of the unit's appearance and extract its appearance features.

[0020] The collected unit appearance features are matched with the appearance parameters of the air conditioning unit to be tested stored in the digital twin. After confirming the target unit, the AI ​​robot is controlled to adjust its posture so that the testing equipment is aligned with the testing area of ​​the unit.

[0021] Furthermore, the step of collecting operating status data of the air conditioning unit under test using the detection equipment includes:

[0022] The thermal imaging camera is activated to acquire thermal imaging images of the evaporator and return pipe of the air conditioning unit under test, and to obtain temperature distribution data on the surface of the evaporator and real-time temperature value of the return pipe.

[0023] The evaporator surface is captured by a high-definition camera to identify the frost formation and record the area and distribution of the frost formation.

[0024] Turn on the audio acquisition device to acquire the sound signal of the air conditioning unit under test during operation within a preset time period, filter out ambient noise, and extract the characteristic audio of the unit operation;

[0025] The temperature distribution data, real-time temperature value of the return pipe, evaporator frost data, and characteristic audio are integrated to form the operating status data of the air conditioning unit under test.

[0026] Furthermore, the step of inputting the collected operating status data into a preset analysis model to obtain the refrigerant flow status analysis results of the air conditioning unit under test includes:

[0027] The thermal imaging images in the operating status data are preprocessed, and the temperature regions of the evaporator and return pipe are extracted by the image segmentation algorithm. The average temperature and temperature variance of the temperature regions are calculated.

[0028] The evaporator frost image is input into a convolutional neural network model to identify the uniformity of frost. When the proportion of uneven frost areas exceeds a preset ratio, it is marked as a candidate for abnormal refrigerant flow.

[0029] Perform Fourier transform on the characteristic audio to convert the time domain signal into a frequency domain signal, extract the frequency and amplitude features of the audio, and compare them with the preset refrigerant abnormal noise feature library.

[0030] Based on the combined temperature analysis results, frost uniformity identification results, and audio comparison results, a weighted voting algorithm is used to obtain the refrigerant flow status analysis results of the air conditioning unit under test. The analysis results include normal, suspected abnormal, and abnormal.

[0031] Furthermore, after obtaining the refrigerant flow state analysis results through a weighted voting algorithm, the following steps are also included:

[0032] Obtain the historical refrigerant flow status analysis results of the air conditioning unit under test within the past month, and statistically analyze the frequency, duration and corresponding operating conditions of historical suspected anomalies and abnormal results.

[0033] Construct a working condition-anomaly correlation model, input current working condition data and historical correlation data, and calculate the confidence level of the current refrigerant flow state analysis results;

[0034] If the confidence level is lower than the preset confidence threshold, the AI ​​robot is controlled to adjust the detection equipment parameters, including the shooting angle of the thermal imaging camera, the sampling frequency of the audio acquisition device, and the focal length of the high-definition camera, and the operating status data is re-acquired.

[0035] The newly collected operating status data is input into the preset analysis model for secondary analysis. The difference between the two analysis results is compared. If the difference is less than the preset difference threshold, the result with the higher risk level is taken as the final refrigerant flow status analysis result. If the difference is greater than or equal to the preset difference threshold, a third-party testing device is activated for verification, and the verification result is taken as the final analysis result.

[0036] Furthermore, the calculation of the residual between the actual operating status data and the ideal simulation data of the air conditioning unit under test includes:

[0037] Real-time operating data of the target computer room is collected by sensors, including ambient temperature, ambient humidity and the set operating parameters of the air conditioning unit;

[0038] Input real-time operating data into the digital twin to simulate the operating status of the air conditioning unit under test under standard refrigerant quantity, and output ideal temperature distribution data, ideal return pipe temperature value, ideal frost simulation data and ideal operating audio data.

[0039] The actual temperature distribution data of the air conditioning unit to be tested is compared with the ideal temperature distribution data point by point, and the temperature residual is calculated.

[0040] Similarly, calculate the temperature residual of the return air pipe, the frost characteristic residual, and the audio characteristic residual respectively;

[0041] Weighting coefficients are set based on the importance of each type of residual, and a comprehensive residual is obtained by weighted summation, which serves as the residual result between the actual operating data and the ideal simulation data of the air conditioning unit under test.

[0042] Furthermore, after calculating the comprehensive residual, it also includes:

[0043] Based on the operating years, maintenance records, and refrigerant type of the air conditioning unit under test, a residual correction model is constructed, and the residual correction coefficient is determined. The value range of the correction coefficient is [0.8, 1.2].

[0044] The composite residual and each individual residual are corrected using residual correction factors to obtain the corrected residual;

[0045] Establish a residual-refrigerant shortage mapping relationship library. The mapping relationship library is constructed based on the refrigerant shortage experimental data under different models and operating conditions, and includes the correspondence between the corrected residual and the refrigerant shortage.

[0046] Based on the corrected residual query mapping relationship library, the estimated refrigerant shortage of the air conditioning unit to be tested is determined;

[0047] The estimated refrigerant shortage is input into the digital twin, and the unit's operating status after replenishing the estimated refrigerant shortage is simulated. The simulation residual between the simulated operating data and the ideal simulation data is calculated. If the simulation residual is less than the preset qualified threshold, the accuracy of the estimated refrigerant shortage is confirmed. If the simulation residual is greater than or equal to the preset qualified threshold, the estimated refrigerant shortage is adjusted and the simulation is repeated until the simulation residual meets the requirements. The final adjusted estimated shortage is taken as the refrigerant shortage determination result.

[0048] Furthermore, determining whether the residual continuously exceeds a preset normal threshold includes:

[0049] Based on the model of the air conditioning unit to be tested and its historical operating data, set the normal threshold range for different types of residuals and the normal threshold for comprehensive residuals.

[0050] Real-time monitoring of the trend of changes in the comprehensive residual; when the comprehensive residual exceeds the normal threshold for the first time, start timing and continuously collect operating status data, and recalculate the residual.

[0051] If the overall residual continuously exceeds the normal threshold within the preset duration, and at least two types of individual residuals simultaneously exceed the corresponding normal threshold, then the air conditioning unit under test is determined to have an abnormal refrigerant quantity.

[0052] An anomaly detection report is generated, which includes the abnormal unit number, the time of the anomaly, the residual data exceeding the standard, and the suspected cause of the anomaly. The report is then sent to the data center management terminal.

[0053] Furthermore, after determining that the air conditioning unit under test has an abnormal refrigerant level, the following steps are also included:

[0054] Control the AI ​​robot to move to the refrigerant detection interface of the abnormal unit and activate the refrigerant pressure sensor to collect refrigerant pressure data;

[0055] Input refrigerant pressure data into the digital twin to simulate the unit's operating status under different refrigerant replenishment amounts and predict the optimal refrigerant replenishment amount;

[0056] Based on the anomaly detection report and the optimal refrigerant replenishment prediction results, maintenance suggestions are generated, which include maintenance steps, required tools, and precautions.

[0057] Regularly control AI robots to re-inspect the repaired air conditioning units, collect operating status data and calculate residuals to confirm that the refrigerant volume has returned to normal.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] 1. The present invention accurately constructs a digital twin by combining AI robot autonomous positioning and detection. By retrieving the layout drawings of the computer room and the parameters of the unit equipment, a three-dimensional spatial model is established. Combined with historical operating data, a digital twin that accurately matches the physical unit is generated. At the same time, the AI ​​robot is controlled to import the three-dimensional model and the unit coordinates to generate an obstacle avoidance movement path. The trajectory is corrected in real time by using LiDAR and the appearance feature is matched to confirm the target unit. This enables multiple devices to collaboratively collect operating status data. The autonomous movement and dual positioning of the AI ​​robot ensure positioning accuracy in the complex environment of the computer room. At the same time, it can enter high-risk and confined spaces to eliminate personnel safety hazards.

[0060] 2. This invention acquires evaporator temperature distribution, frost pattern, and operating audio data through a thermal imaging camera, a high-definition camera, and an audio acquisition device, respectively. After processing, it combines historical operating data to construct an association model to calculate the confidence level of the analysis results. When the confidence level is low, the device parameters are adjusted for secondary acquisition or third-party verification is initiated to improve the coverage of anomaly identification. The hierarchical analysis model first achieves accurate single-dimensional analysis through professional algorithms, then combines historical data to evaluate the confidence level, and finally corrects errors through secondary verification to improve the overall accuracy of refrigerant flow state analysis.

[0061] 3. This invention collects real-time operating data to drive the generation of ideal simulation data by a digital twin. It calculates and weights multi-dimensional residuals of temperature, frost, and audio to obtain a comprehensive residual. After correcting the residuals by combining the unit's operating years and maintenance records, it uses a mapping relationship library to back-calculate the refrigerant shortage. The shortage is then verified by digital twin simulation. Finally, maintenance suggestions are generated and regularly re-inspected. The dynamic residual calculation is based on synchronous operating conditions to ensure benchmark fairness. The corrected residuals closely match the actual state of the unit, controlling the error in back-calculating the shortage. The closed-loop management of abnormal repair uses digital twin to predict the optimal replenishment amount, generate maintenance guidance, and re-inspect to confirm the effect, shortening the maintenance response time. Attached Figure Description

[0062] Figure 1 This is a schematic diagram of the air conditioning unit refrigerant quantity abnormality detection 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] Please see Figure 1 The present invention provides the following technical solutions:

[0065] A method for detecting refrigerant quantity anomalies in air conditioning units based on digital twins and AI robots includes the following steps:

[0066] Obtain basic information about the air conditioning unit to be tested in the target computer room, and construct a digital twin of the air conditioning unit to be tested;

[0067] The AI ​​robot is controlled to carry the detection equipment and move autonomously within the target computer room, locate each air conditioning unit to be tested, and use the detection equipment to collect the operating status data of the air conditioning units to be tested;

[0068] The collected operating status data is input into the preset analysis model to obtain the refrigerant flow status analysis results of the air conditioning unit under test;

[0069] Retrieve ideal simulation data of the digital twin under the same operating conditions and standard refrigerant quantity as the air conditioning unit under test, calculate the residual between the actual operating status data of the air conditioning unit under test and the ideal simulation data; determine whether the residual continuously exceeds the preset normal threshold, if it does, determine that the air conditioning unit under test has an abnormal refrigerant quantity.

[0070] Obtain basic information about the air conditioning unit to be tested in the target computer room, and construct a digital twin of the air conditioning unit to be tested, including:

[0071] The layout drawings of the target computer room and the equipment parameters of the air conditioning unit to be tested are retrieved through the computer room management system. The equipment parameters include the unit model, rated refrigerant capacity, evaporator specifications and return pipe size.

[0072] Based on the layout drawings, a three-dimensional spatial model of the target computer room is established, and the coordinates of the unit are marked in the three-dimensional spatial model corresponding to the actual installation position of the air conditioning unit to be tested.

[0073] Input the equipment parameters of the air conditioning unit to be tested into the three-dimensional space model to generate an initial digital model that includes the internal structure of the unit and the refrigerant circulation path;

[0074] Historical operating data of the air conditioning unit under test under standard refrigerant volume and different operating conditions are obtained, and the parameters of the initial digital model are calibrated to obtain a digital twin that is precisely matched with the air conditioning unit under test.

[0075] In the above embodiments, by acquiring the layout of the computer room and the static parameters of the unit during the data acquisition stage, and incorporating key internal structural information such as the refrigerant circulation path, the digital model can truly reflect the flow logic of the refrigerant within the unit. This breaks through the limitations of traditional models that only focus on appearance and basic parameters. The combination of the three-dimensional spatial model and the unit coordinates achieves accurate mapping of the macro layout of the computer room and the micro position of the unit, providing a high-precision spatial reference for the autonomous positioning of AI robots. This avoids alignment errors of detection equipment caused by spatial positioning deviations. Historical operating data is dynamically calibrated, and the model parameters are adjusted in reverse using measured data under different operating conditions with standard refrigerant quantities. This controls the deviation between the simulation results of the digital twin and the actual operating characteristics of the physical unit. The calibrated digital twin can accurately output ideal temperature distribution and frosting status data under different operating conditions.

[0076] The AI ​​robot is controlled to autonomously move within the target server room, carrying detection equipment, and locate each air conditioning unit to be tested, including:

[0077] The three-dimensional spatial model of the target computer room and the unit coordinates of the air conditioning unit to be tested are imported into the navigation system of the AI ​​robot to generate an autonomous movement path. The autonomous movement path avoids equipment obstacles and personnel passages in the computer room.

[0078] The AI ​​robot activates its LiDAR to scan the surrounding environment, acquires environmental point cloud data in real time, compares it with a preset 3D spatial model, and corrects its movement trajectory.

[0079] When the AI ​​robot moves to the preset distance range of the air conditioning unit to be inspected, the camera is activated to capture images of the unit's appearance and extract its appearance features.

[0080] The collected unit appearance features are matched with the appearance parameters of the air conditioning unit to be tested stored in the digital twin. After confirming the target unit, the AI ​​robot is controlled to adjust its posture so that the testing equipment is aligned with the testing area of ​​the unit.

[0081] In the above embodiments, the generation of autonomous movement paths is not only based on the 3D model of the computer room, but also incorporates obstacle avoidance logic for equipment and personnel passages, significantly improving flexibility and adapting to dynamic scenarios such as the addition or removal of computer room equipment and temporary obstacles. Real-time point cloud comparison and trajectory correction by LiDAR control the positional deviation of the robot during movement within ±5cm, ensuring accurate arrival near the target unit. A secondary confirmation mechanism for appearance feature matching is introduced, comparing the appearance details of the unit (such as body markings, interface positions, and pipeline layout) collected by the camera with the stored parameters in the digital twin, avoiding misidentification caused by similar unit models or similar installation positions. Compared with manual guidance of robot detection, it saves auxiliary time. By combining the static parameters of the digital twin with the dynamic navigation of the robot, the pain points of traditional robots being prone to navigation deviation and identification errors in complex computer room environments are solved.

[0082] The testing equipment is used to collect operating status data of the air conditioning unit under test, including:

[0083] The thermal imaging camera is activated to acquire thermal imaging images of the evaporator and return pipe of the air conditioning unit under test, and to obtain temperature distribution data on the surface of the evaporator and real-time temperature value of the return pipe.

[0084] The evaporator surface is captured by a high-definition camera to identify the frost formation and record the area and distribution of the frost formation.

[0085] Turn on the audio acquisition device to acquire the sound signal of the air conditioning unit under test during operation within a preset time period, filter out ambient noise, and extract the characteristic audio of the unit operation;

[0086] The temperature distribution data, real-time temperature value of the return pipe, evaporator frost data, and characteristic audio are integrated to form the operating status data of the air conditioning unit under test.

[0087] In the above embodiments, targeting the core characteristics of uneven evaporator frosting and abnormal return pipe temperature when refrigerant is insufficient, a thermal imaging camera and a high-definition camera are used in synergy to accurately identify local low-temperature frosting areas. The high-definition camera captures the specific morphology (such as dotted frost, sheet-like frost) and distribution pattern (such as unilateral frost, local dense frost). The combination of the two upgrades the description of frost characteristics from presence or absence to multi-dimensional data combining morphology, distribution, and temperature. Audio acquisition and noise filtering are used to extract frequency domain features through Fourier transform for features such as compressor abnormal noise and airflow turbulence caused by refrigerant abnormalities. This fills the technical gap of traditional detection neglecting acoustic signals, making the anomaly judgment more comprehensive. The dataset collected by multiple devices can cover various manifestations of refrigerant flow anomalies. Compared with single temperature detection, it improves the anomaly identification coverage and provides a high-quality data foundation for accurate analysis of refrigerant flow status.

[0088] The collected operating status data is input into a preset analysis model to obtain the refrigerant flow status analysis results of the air conditioning unit under test, including:

[0089] The thermal imaging images in the operating status data are preprocessed, and the temperature regions of the evaporator and return pipe are extracted by the image segmentation algorithm. The average temperature and temperature variance of the temperature regions are calculated.

[0090] The evaporator frost image is input into a convolutional neural network model to identify the uniformity of frost. When the proportion of uneven frost areas exceeds a preset ratio, it is marked as a candidate for abnormal refrigerant flow.

[0091] Perform Fourier transform on the characteristic audio to convert the time domain signal into a frequency domain signal, extract the frequency and amplitude features of the audio, and compare them with the preset refrigerant abnormal noise feature library.

[0092] Based on the combined temperature analysis results, frost uniformity identification results, and audio comparison results, a weighted voting algorithm is used to obtain the refrigerant flow status analysis results of the air conditioning unit under test. The analysis results include normal, suspected abnormal, and abnormal.

[0093] After obtaining the refrigerant flow status analysis results through a weighted voting algorithm, the historical refrigerant flow status analysis results of the air conditioning unit under test within the past month are obtained, and the frequency, duration and corresponding operating conditions of historical suspected anomalies and abnormal results are statistically analyzed.

[0094] Construct a working condition-anomaly correlation model, input current working condition data and historical correlation data, and calculate the confidence level of the current refrigerant flow state analysis results;

[0095] If the confidence level is lower than the preset confidence threshold, the AI ​​robot is controlled to adjust the detection equipment parameters, including the shooting angle of the thermal imaging camera, the sampling frequency of the audio acquisition device, and the focal length of the high-definition camera, and the operating status data is re-acquired.

[0096] The newly collected operating status data is input into the preset analysis model for secondary analysis. The difference between the two analysis results is compared. If the difference is less than the preset difference threshold, the result with the higher risk level is taken as the final refrigerant flow status analysis result. If the difference is greater than or equal to the preset difference threshold, a third-party testing device is activated for verification, and the verification result is taken as the final analysis result.

[0097] In the above embodiments, specialized processing algorithms are used for temperature, frost, and audio data in the basic analysis stage. The image segmentation algorithm accurately extracts temperature regions to avoid interference from ambient temperature. The convolutional neural network model is trained with a large number of frost samples to identify frost unevenness areas of 0.1%. The Fourier transform converts the audio signal into quantifiable frequency domain features to achieve objective comparison of abnormal sounds. The synergy of the three algorithms improves the accuracy of single-dimensional analysis. Historical data is introduced to construct an operating condition-abnormality correlation model. By statistically analyzing the correspondence between historical suspected abnormalities, abnormal results, and operating conditions, it can be determined whether the current analysis results are consistent with the unit's operating rules. This adds an evaluation dimension of regularity to the analysis results and avoids misjudgments caused by occasional interference.

[0098] In the above embodiments, when the confidence level is insufficient, the data is re-collected by adjusting the parameters of the detection equipment to ensure data quality; when the difference between the two results is large, third-party verification is initiated to further reduce the risk of misjudgment and improve the overall accuracy of refrigerant flow state analysis. Real-time data analysis is combined with historical pattern mining, the reliability of the analysis results is quantified by confidence level assessment, and the error is corrected by dynamically adjusting the detection parameters and third-party verification. This overcomes the limitations of traditional models that rely solely on real-time data and cannot self-verify and optimize.

[0099] Calculate the residual between the actual operating data and the ideal simulation data of the air conditioning unit under test, including:

[0100] Real-time operating data of the target computer room is collected by sensors, including ambient temperature, ambient humidity and the set operating parameters of the air conditioning unit;

[0101] Input real-time operating data into the digital twin to simulate the operating status of the air conditioning unit under test under standard refrigerant quantity, and output ideal temperature distribution data, ideal return pipe temperature value, ideal frost simulation data and ideal operating audio data.

[0102] The actual temperature distribution data of the air conditioning unit to be tested is compared with the ideal temperature distribution data point by point, and the temperature residual is calculated.

[0103] Similarly, calculate the temperature residual of the return air pipe, the frost characteristic residual, and the audio characteristic residual respectively;

[0104] Weighting coefficients are set based on the importance of each type of residual, and the comprehensive residual is obtained by weighted summation, which serves as the residual result between the actual operating data and the ideal simulation data of the air conditioning unit under test.

[0105] After calculating the comprehensive residual, a residual correction model is constructed based on the operating years, maintenance records, and refrigerant type of the air conditioning unit under test, and the residual correction coefficient is determined. The value range of the correction coefficient is [0.8, 1.2]. The longer the operating years and the higher the maintenance frequency, the smaller the correction coefficient.

[0106] The composite residual and each individual residual are corrected using residual correction factors to obtain the corrected residual;

[0107] Establish a residual-refrigerant shortage mapping relationship library. The mapping relationship library is constructed based on the refrigerant shortage experimental data under different models and operating conditions, and includes the correspondence between the corrected residual and the refrigerant shortage.

[0108] Based on the corrected residual query mapping relationship library, the estimated refrigerant shortage of the air conditioning unit to be tested is determined;

[0109] The estimated refrigerant shortage is input into the digital twin, and the unit's operating status after replenishing the estimated refrigerant shortage is simulated. The simulation residual between the simulated operating data and the ideal simulation data is calculated. If the simulation residual is less than the preset qualified threshold, the accuracy of the estimated refrigerant shortage is confirmed. If the simulation residual is greater than or equal to the preset qualified threshold, the estimated refrigerant shortage is adjusted and the simulation is repeated until the simulation residual meets the requirements. The final adjusted estimated shortage is taken as the refrigerant shortage determination result.

[0110] In the above embodiments, by simulating real-time operating data and digital twins synchronously, it is ensured that ideal data and actual data are in the same environment and operating conditions, avoiding the artificially high or low residuals caused by differences in operating conditions. This makes the benchmark for residual calculation fairer and the results more reliable. Multi-dimensional residual calculation is not a simple summation, but rather sets weights based on the importance of each residual. For example, the temperature residual reflects the refrigerant heat exchange efficiency and has a weight of 0.4; the frosting residual reflects the refrigerant distribution and has a weight of 0.3; and the audio residual assists in verification and has a weight of 0.3. The comprehensive residual obtained by weighted summation can more comprehensively and objectively reflect the overall degree of anomaly and is more representative than a single residual. A residual correction model and a missing quantity mapping relationship library are introduced. The residual correction model considers actual factors such as the unit's operating years and maintenance records, making the corrected residuals more consistent with the actual state of the unit. The missing quantity mapping relationship library is built based on a large amount of experimental data and can directly convert the corrected residuals into specific refrigerant missing quantities, thereby quantifying the degree of anomaly.

[0111] Determining whether the residual continuously exceeds a preset normal threshold includes:

[0112] Based on the model of the air conditioning unit to be tested and its historical operating data, set the normal threshold range for different types of residuals and the normal threshold for comprehensive residuals.

[0113] Real-time monitoring of the trend of changes in the comprehensive residual; when the comprehensive residual exceeds the normal threshold for the first time, start timing and continuously collect operating status data, and recalculate the residual.

[0114] If the overall residual continuously exceeds the normal threshold within the preset duration, and at least two types of individual residuals simultaneously exceed the corresponding normal threshold, then the air conditioning unit under test is determined to have an abnormal refrigerant quantity.

[0115] Generate an anomaly detection report, which includes the abnormal unit number, the time of the anomaly, the residual data exceeding the standard, and the suspected cause of the anomaly, and send the report to the data center management terminal;

[0116] After determining that the air conditioning unit under test has an abnormal refrigerant quantity, the AI ​​robot is controlled to move to the refrigerant detection interface of the abnormal unit and start the refrigerant pressure sensor to collect refrigerant pressure data.

[0117] Input refrigerant pressure data into the digital twin to simulate the unit's operating status under different refrigerant replenishment amounts and predict the optimal refrigerant replenishment amount;

[0118] Based on the anomaly detection report and the optimal refrigerant replenishment prediction results, maintenance suggestions are generated, which include maintenance steps, required tools, and precautions.

[0119] Regularly control AI robots to re-inspect the repaired air conditioning units, collect operating status data and calculate residuals to confirm that the refrigerant volume has returned to normal.

[0120] In the above embodiments, dynamic threshold settings based on the model and historical data avoid the defects of fixed thresholds (such as the normal residual range of high-power units being wider than that of low-power units, and the normal threshold of old units being higher than that of new units), making the thresholds more in line with the actual operating characteristics of the units, reducing misjudgments or omissions caused by improper threshold settings. The continuous monitoring and multi-condition judgment mechanism requires that the comprehensive residual continuously exceeds the threshold and at least two individual residuals exceed the standard simultaneously, which can effectively filter instantaneous interference, ensure the rigor of anomaly judgment, reduce the misjudgment rate, and shorten the maintenance response time of abnormal units. By taking anomaly judgment as the starting point and linking with AI robots and digital twins, maintenance guidance and re-inspection verification functions are extended, and a full life cycle management system is built to overcome the defects of traditional detection and maintenance being disconnected and maintenance effects not being verified.

[0121] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for detecting abnormal refrigerant levels in air conditioning units based on digital twins and AI robots, characterized in that, Includes the following steps: Obtain basic information about the air conditioning unit to be tested in the target computer room, and construct a digital twin of the air conditioning unit to be tested; The AI ​​robot is controlled to carry the detection equipment and move autonomously within the target computer room, locate each air conditioning unit to be tested, and use the detection equipment to collect the operating status data of the air conditioning units to be tested; The collected operating status data is input into the preset analysis model to obtain the refrigerant flow status analysis results of the air conditioning unit under test; Retrieve ideal simulation data of the digital twin under the same operating conditions and standard refrigerant quantity as the air conditioner unit under test, calculate the residual between the actual operating status data of the air conditioner unit under test and the ideal simulation data; determine whether the residual continuously exceeds the preset normal threshold, if it exceeds the threshold, determine that the air conditioner unit under test has an abnormal refrigerant quantity; The calculation of the residual between the actual operating status data and the ideal simulation data of the air conditioning unit under test includes: Real-time operating data of the target computer room is collected by sensors, including ambient temperature, ambient humidity and the set operating parameters of the air conditioning unit; Input real-time operating data into the digital twin to simulate the operating status of the air conditioning unit under test under standard refrigerant quantity, and output ideal temperature distribution data, ideal return pipe temperature value, ideal frost simulation data and ideal operating audio data. The actual temperature distribution data of the air conditioning unit to be tested is compared with the ideal temperature distribution data point by point, and the temperature residual is calculated. Similarly, calculate the temperature residual of the return air pipe, the frost characteristic residual, and the audio characteristic residual respectively; Weighting coefficients are set based on the importance of each type of residual, and the comprehensive residual is obtained by weighted summation, which serves as the residual result between the actual operating data and the ideal simulation data of the air conditioning unit under test. Based on the operating years, maintenance records, and refrigerant type of the air conditioning unit under test, a residual correction model is constructed, and the residual correction coefficient is determined. The value range of the correction coefficient is [0.8, 1.2]. The composite residual and each individual residual are corrected using residual correction factors to obtain the corrected residual; Establish a residual-refrigerant shortage mapping relationship library. The mapping relationship library is constructed based on the refrigerant shortage experimental data under different models and operating conditions, and includes the correspondence between the corrected residual and the refrigerant shortage. Based on the corrected residual query mapping relationship library, the estimated refrigerant shortage of the air conditioning unit to be tested is determined; The estimated refrigerant shortage is input into the digital twin, and the unit's operating status after replenishing the estimated refrigerant shortage is simulated. The simulation residual between the simulated operating data and the ideal simulation data is calculated. If the simulation residual is less than the preset qualified threshold, the accuracy of the estimated refrigerant shortage is confirmed. If the simulation residual is greater than or equal to the preset qualified threshold, the estimated refrigerant shortage is adjusted and the simulation is repeated until the simulation residual meets the requirements. The final adjusted estimated shortage is taken as the refrigerant shortage determination result.

2. The method for detecting abnormal refrigerant levels in air conditioning units based on digital twins and AI robots as described in claim 1, characterized in that, The process of acquiring basic information about the air conditioning unit to be tested in the target computer room and constructing a digital twin of the air conditioning unit to be tested includes: The layout drawings of the target computer room and the equipment parameters of the air conditioning unit to be tested are retrieved through the computer room management system. The equipment parameters include the unit model, rated refrigerant capacity, evaporator specifications and return pipe size. Based on the layout drawings, a three-dimensional spatial model of the target computer room is established, and the coordinates of the unit are marked in the three-dimensional spatial model corresponding to the actual installation position of the air conditioning unit to be tested. Input the equipment parameters of the air conditioning unit to be tested into the three-dimensional space model to generate an initial digital model that includes the internal structure of the unit and the refrigerant circulation path; Historical operating data of the air conditioning unit under test under standard refrigerant volume and different operating conditions are obtained, and the parameters of the initial digital model are calibrated to obtain a digital twin that is precisely matched with the air conditioning unit under test.

3. The method for detecting abnormal refrigerant levels in air conditioning units based on digital twins and AI robots as described in claim 1, characterized in that, The AI ​​robot, carrying detection equipment, moves autonomously within the target computer room, locating each air conditioning unit to be tested, including: The three-dimensional spatial model of the target computer room and the unit coordinates of the air conditioning unit to be tested are imported into the navigation system of the AI ​​robot to generate an autonomous movement path. The autonomous movement path avoids equipment obstacles and personnel passages in the computer room. The AI ​​robot activates its LiDAR to scan the surrounding environment, acquires environmental point cloud data in real time, compares it with a preset 3D spatial model, and corrects its movement trajectory. When the AI ​​robot moves to the preset distance range of the air conditioning unit to be inspected, the camera is activated to capture images of the unit's appearance and extract its appearance features. The collected unit appearance features are matched with the appearance parameters of the air conditioning unit to be tested stored in the digital twin. After confirming the target unit, the AI ​​robot is controlled to adjust its posture so that the testing equipment is aligned with the testing area of ​​the unit.

4. The method for detecting abnormal refrigerant levels in air conditioning units based on digital twins and AI robots as described in claim 1, characterized in that, The process of collecting operating status data of the air conditioning unit under test using the detection equipment includes: The thermal imaging camera is activated to acquire thermal imaging images of the evaporator and return pipe of the air conditioning unit under test, and to obtain temperature distribution data on the surface of the evaporator and real-time temperature value of the return pipe. The evaporator surface is captured by a high-definition camera to identify the frost formation and record the area and distribution of the frost formation. Turn on the audio acquisition device to acquire the sound signal of the air conditioning unit under test during operation within a preset time period, filter out ambient noise, and extract the characteristic audio of the unit operation; The temperature distribution data, real-time temperature value of the return pipe, evaporator frost data, and characteristic audio are integrated to form the operating status data of the air conditioning unit under test.

5. The method for detecting abnormal refrigerant levels in air conditioning units based on digital twins and AI robots as described in claim 1, characterized in that, The step of inputting the collected operating status data into a preset analysis model to obtain the refrigerant flow status analysis results of the air conditioning unit under test includes: The thermal imaging images in the operating status data are preprocessed, and the temperature regions of the evaporator and return pipe are extracted by the image segmentation algorithm. The average temperature and temperature variance of the temperature regions are calculated. The evaporator frost image is input into a convolutional neural network model to identify the uniformity of frost. When the proportion of uneven frost areas exceeds a preset ratio, it is marked as a candidate for abnormal refrigerant flow. Perform Fourier transform on the characteristic audio to convert the time domain signal into a frequency domain signal, extract the frequency and amplitude features of the audio, and compare them with the preset refrigerant abnormal noise feature library. Based on the combined temperature analysis results, frost uniformity identification results, and audio comparison results, a weighted voting algorithm is used to obtain the refrigerant flow status analysis results of the air conditioning unit under test. The analysis results include normal, suspected abnormal, and abnormal.

6. The method for detecting abnormal refrigerant levels in air conditioning units based on digital twins and AI robots as described in claim 5, characterized in that, After obtaining the refrigerant flow state analysis results through a weighted voting algorithm, the following is also included: Obtain the historical refrigerant flow status analysis results of the air conditioning unit under test within the past month, and statistically analyze the frequency, duration and corresponding operating conditions of historical suspected anomalies and abnormal results. Construct a working condition-anomaly correlation model, input current working condition data and historical correlation data, and calculate the confidence level of the current refrigerant flow state analysis results; If the confidence level is lower than the preset confidence threshold, the AI ​​robot is controlled to adjust the detection equipment parameters, including the shooting angle of the thermal imaging camera, the sampling frequency of the audio acquisition device, and the focal length of the high-definition camera, and the operating status data is re-acquired. The newly collected operating status data is input into the preset analysis model for secondary analysis. The difference between the two analysis results is compared. If the difference is less than the preset difference threshold, the result with the higher risk level is taken as the final refrigerant flow status analysis result. If the difference is greater than or equal to the preset difference threshold, a third-party testing device is activated for verification, and the verification result is taken as the final analysis result.

7. The method for detecting abnormal refrigerant levels in air conditioning units based on digital twins and AI robots as described in claim 1, characterized in that, The step of determining whether the residual continuously exceeds a preset normal threshold includes: Based on the model of the air conditioning unit to be tested and its historical operating data, set the normal threshold range for different types of residuals and the normal threshold for comprehensive residuals. Real-time monitoring of the trend of changes in the comprehensive residual; when the comprehensive residual exceeds the normal threshold for the first time, start timing and continuously collect operating status data, and recalculate the residual. If the overall residual continuously exceeds the normal threshold within the preset duration, and at least two types of individual residuals simultaneously exceed the corresponding normal threshold, then the air conditioning unit under test is determined to have an abnormal refrigerant quantity. An anomaly detection report is generated, which includes the abnormal unit number, the time of the anomaly, the residual data exceeding the standard, and the suspected cause of the anomaly. The report is then sent to the data center management terminal.

8. The method for detecting abnormal refrigerant levels in air conditioning units based on digital twins and AI robots as described in claim 7, characterized in that, After determining that the air conditioning unit under test has an abnormal refrigerant level, the following steps are also included: Control the AI ​​robot to move to the refrigerant detection interface of the abnormal unit and activate the refrigerant pressure sensor to collect refrigerant pressure data; Input refrigerant pressure data into the digital twin to simulate the unit's operating status under different refrigerant replenishment amounts and predict the optimal refrigerant replenishment amount; Based on the anomaly detection report and the optimal refrigerant replenishment prediction results, maintenance suggestions are generated, which include maintenance steps, required tools, and precautions. Regularly control AI robots to re-inspect the repaired air conditioning units, collect operating status data and calculate residuals to confirm that the refrigerant volume has returned to normal.

Images