Method and device for intelligently diagnosing refrigerant leakage, storage medium and refrigerator

Abstract

The invention provides a method and device for intelligently diagnosing refrigerant leakage, a storage medium and a refrigerator, the method is suitable for networking refrigeration equipment, and the method comprises the following steps that multi-dimensional serial port data in the operation process of the refrigerator is collected; performing validity verification and encryption processing on the collected data, and uploading the data to a cloud big data platform; cleaning, missing value complementing and abnormal value processing are carried out on the uploaded data at the cloud; key features are extracted to construct a time sequence diagnosis fragment, the time sequence diagnosis fragment is analyzed based on a machine learning decision tree model, and whether the refrigerant leakage risk exists or not is judged; and when it is judged that the refrigerant leakage risk exists, a diagnosis result is generated and pushed to an operation and maintenance background and a user terminal. According to the invention, on the premise of not adding any sensor, early detection and accurate early warning of refrigerant leakage can be realized.

Description

Technical Field

[0001] This invention relates to the field of refrigerator fault diagnosis technology, and more specifically, to a method, apparatus, storage medium, and refrigerator for intelligent diagnosis of refrigerant leakage. Background Technology

[0002] Currently, most household refrigerators on the market rely on additional specialized sensor hardware, such as pressure sensors, gas concentration sensors, or infrared leak detectors, to detect refrigerant leaks. These methods determine the presence of a leak by monitoring changes in refrigerant line pressure or the concentration of refrigerant molecules in the surrounding environment. However, these traditional detection methods depend on hardware sensors, are costly, and have poor accuracy.

[0003] More importantly, traditional detection mechanisms are generally lagging behind. Usually, users only notice the abnormality and report the problem when a large amount of refrigerant has been lost and the refrigerator has obvious cooling failure. Repair personnel need to disassemble the machine on-site to confirm the location and extent of the leak. This "post-incident repair" model is not only slow to respond and has a long service cycle, but it also often misses the best time for intervention, which may lead to serious consequences such as compressor burnout and the scrapping of the entire machine, greatly affecting user experience and brand reputation. Summary of the Invention

[0004] The purpose of this invention is to provide a method, device, storage medium, and refrigerator for intelligent diagnosis of refrigerant leaks. It makes full use of the multi-dimensional serial port operation data output by the refrigerator controller, and performs cleaning, feature extraction, and machine learning modeling through a cloud big data platform to construct an XGBoost decision tree model that can identify typical refrigerant leak patterns. This enables early detection and accurate warning of refrigerant leaks without adding any sensors.

[0005] To achieve the above objectives, this invention provides a method, apparatus, storage medium, and refrigerator for intelligent diagnosis of refrigerant leaks. The technical solution of this invention is implemented as follows:

[0006] A method for intelligently diagnosing refrigerant leaks, applicable to networked refrigeration equipment, includes the following steps:

[0007] Collect multi-dimensional serial port data during the operation of refrigeration equipment;

[0008] The collected data is validated and encrypted before being uploaded to the cloud-based big data platform.

[0009] Uploaded data is cleaned, missing values ​​are filled, and outliers are handled in the cloud.

[0010] Key features are extracted to construct time-series diagnostic segments. These segments are then analyzed using a machine learning decision tree model to determine if there is a risk of refrigerant leakage.

[0011] When a risk of refrigerant leakage is detected, a diagnostic result is generated and pushed to the operation and maintenance backend and user terminals.

[0012] Furthermore, the multi-dimensional serial port data includes one or more of the following: ambient temperature, freezer temperature, refrigerator temperature, freezer defrosting temperature, refrigerator defrosting temperature, real-time power, compressor operating status, refrigerator set operating mode, freezer door open / close status, refrigerator door open / close status, compressor start / stop ratio, and compartment cooling status.

[0013] Furthermore, each of the time-series diagnostic segments contains sampled data over a continuous time period.

[0014] Furthermore, the machine learning decision tree model is an XGBoost model, which is trained using historical normal operation data and known refrigerant leak sample data to learn the typical operating mode change patterns before and after refrigerant leaks.

[0015] Furthermore, the logic for determining refrigerant leakage is as follows: whether the refrigerant quantity meets the standard is indirectly determined through operational characteristics, without the need to directly measure the refrigerant quantity.

[0016] Furthermore, the operating characteristics include: using linear regression to fit the slope of temperature and power changes to reflect the continuous changes in the equipment's cooling performance.

[0017] Furthermore, the operating characteristics also include: using the power standard deviation to quantify the fluctuations to reflect whether the equipment's cooling performance is stable.

[0018] Furthermore, the operating characteristics also include: using the difference in the number of state switching to reflect whether the device's cooling is frequently adjusted.

[0019] Furthermore, the diagnostic results include at least one of the following: risk level, abnormal indicators, and suspected leakage time.

[0020] Furthermore, the diagnostic results are simultaneously pushed to the user's APP or notified to the user via SMS, informing them of the risk level and temporary response suggestions; at the same time, the diagnostic results and key abnormal data are pushed to the manufacturer's operation and maintenance management backend for after-sales service personnel to view and formulate maintenance plans.

[0021] Furthermore, it also includes conducting feature importance analysis and performing multi-dimensional cross-validation and interference elimination.

[0022] A device for intelligently diagnosing refrigerant leaks in a refrigerator, employing the method described above for intelligent diagnosis, the device comprising:

[0023] The data reporting module is used to obtain multi-dimensional operating parameters from the networked refrigerator controller and perform digital signal encryption and verification processing.

[0024] The data transmission and storage module is used to upload data to the cloud server;

[0025] The intelligent diagnostic module includes a data processing module, a feature importance analysis module, a refrigerant leakage algorithm model, and a risk assessment conclusion module.

[0026] The early warning and feedback module is used to trigger the early warning mechanism and distribute the diagnostic results to the operation and maintenance management device and user terminal.

[0027] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.

[0028] A refrigerator employs the method described above for intelligent diagnosis of refrigerant leaks.

[0029] Compared with existing technologies, the intelligent method, device, storage medium, and refrigerator for diagnosing refrigerant leaks described in this invention have the following advantages:

[0030] 1. No new hardware required, reducing device costs. By fully utilizing the refrigerator's existing sensors and serial communication modules to collect operational data, there is no need to install additional pressure sensors or gas detection devices, avoiding increased hardware investment and production process modifications, thus significantly reducing manufacturing and maintenance costs.

[0031] 2. Data-driven early warning system. Through cloud-based big data analytics and the XGBoost machine learning model, feature extraction and pattern recognition are performed on multi-dimensional time-series data. This enables the identification of abnormal trends before minor refrigerant leaks cause significant refrigeration failures, overcoming the limitations of "post-incident repair" and achieving early detection and proactive intervention in faults.

[0032] 3. High diagnostic accuracy and strong anti-interference capability. Employing a multi-dimensional cross-validation mechanism combined with refrigerant quantity indirect inference logic, the system comprehensively judges the device's operating status, effectively eliminating misjudgments caused by non-fault factors such as environmental changes and frequent door openings, thus improving diagnostic accuracy and robustness.

[0033] 4. Remote real-time monitoring, supporting large-scale management. Data is encrypted and uploaded to the cloud platform via Wi-Fi, enabling remote, real-time, and continuous monitoring. This is suitable for centralized operation and maintenance management of massive numbers of networked devices, improving service response efficiency and promoting the development of home appliances towards intelligence and networking.

[0034] 5. Dual-end synchronous notification forms a closed-loop handling mechanism. Once a leakage risk is determined, the device automatically generates a diagnostic report containing the risk level, abnormal indicators, and suspected leakage time, and pushes it simultaneously to the operation and maintenance backend and user terminals (APP or SMS), realizing a closed-loop management of the entire process of "cloud diagnosis - user reminder - after-sales handling", enhancing user experience and brand trust.

[0035] 6. Highly scalable and with potential for general application. The proposed method is not only applicable to household refrigerators, but can also be extended to other refrigeration devices such as freezers, air conditioners, and commercial refrigeration equipment, providing a standardized technical path for health status assessment and intelligent operation and maintenance of various refrigeration equipment. Attached Figure Description

[0036] Figure 1 The flowchart of the XGboost core prediction model described in Embodiment 1 of the present invention is shown below;

[0037] Figure 2 This is a schematic diagram of the data reporting module and intelligent diagnosis module of the device described in Embodiment 2 of the present invention.

[0038] Explanation of reference numerals in the attached figures:

[0039] none. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the described embodiments are only some, not all, of the embodiments of this invention. The specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0041] With the development of IoT technology and big data analytics, an increasing number of smart refrigerators now possess WiFi connectivity and operational data upload capabilities, providing a practical foundation for remote fault diagnosis based on cloud-based algorithms. However, the entire diagnostic process heavily relies on human experience and lacks a standardized, data-driven analysis workflow. This makes it difficult to achieve large-scale, unified remote monitoring and intelligent management of equipment, and fails to meet the current development needs of smart homes for self-diagnosis and self-early warning capabilities.

[0042] This invention proposes a method, device, storage medium, and refrigerator for intelligent diagnosis of refrigerant leaks, aiming to solve the problems of traditional detection methods relying on hardware sensors, high cost, and poor accuracy.

[0043] Example 1

[0044] This embodiment provides a method for intelligent diagnosis of refrigerant leaks based on cloud-based machine learning, applicable to networked refrigerators. This method achieves early identification and simultaneous dual-end warning of refrigerant leaks through multi-dimensional serial port data acquisition, feature extraction, and decision tree model analysis, forming a "cloud-user" information linkage and closed-loop management system. The specific steps are as follows:

[0045] S1 collects multi-dimensional serial port data during the refrigerator's operation.

[0046] The S11 refrigerator's built-in sensors collect operating parameters in real time via a serial communication link, including ambient temperature, freezer temperature, refrigerator temperature, freezer defrosting temperature, refrigerator defrosting temperature, real-time power, compressor operating status, refrigerator set operating mode, freezer door open / close status, refrigerator door open / close status, compressor start / stop ratio, and compartment cooling status.

[0047] S12 converts the collected analog signal into a digital signal and transmits it to the refrigerator's mainboard for preliminary processing.

[0048] By utilizing the sensor outputs in the refrigerator's existing control device, there is no need to install additional dedicated detection hardware, reducing equipment costs; it enables continuous monitoring of all key parameters, providing a data foundation for subsequent accurate diagnosis.

[0049] S2 verifies the validity of the collected data, encrypts it, and then uploads it to the cloud-based big data platform.

[0050] S21, the motherboard verifies the validity of the received data, excluding outliers or missing data caused by transmission errors or sensor malfunctions. An initial edge filtering mechanism is included before data upload: if there are no significant changes in the operating mode before or after the current moment, or if preset trigger conditions are not met, upload is temporarily suspended to reduce communication load.

[0051] S22, encrypts the valid data and standardizes it into a unified format.

[0052] S23 uploads encrypted, standardized data to a cloud-based big data platform via a WiFi network.

[0053] Validation and encryption ensure data integrity and security, preventing invalid or forged data from interfering with diagnostic results; real-time remote uploading is achieved through wireless communication, supporting large-scale device access and centralized management.

[0054] S3, cloud-based data preprocessing, involves cleaning, filling in missing values, and handling outliers on uploaded data in the cloud.

[0055] S31 receives and uploads data in the cloud and stores it in a distributed database.

[0056] S32 performs a cleaning operation on the stored data to remove noise and obvious outliers.

[0057] S33, use interpolation or time series prediction algorithms to fill in missing values.

[0058] Preprocessing improves data quality and ensures stable and reliable input for subsequent modeling.

[0059] S4. Conduct feature importance analysis and perform multi-dimensional cross-validation and interference elimination.

[0060] S41. Conduct feature importance analysis to identify the variables that contribute the most to the diagnosis, such as compressor runtime and power matching.

[0061] S42 uses correlation analysis to verify the consistency between temperature changes and compressor frequency, power, and compartment cooling status, eliminating false alarms caused by a single parameter anomaly. This reduces false alarms caused by environmental fluctuations or temporary usage habits (such as frequent door opening), improving diagnostic accuracy and robustness.

[0062] S5 uses a refrigerant leakage algorithm model to assess the risk of refrigerant leakage.

[0063] S51, Feature Extraction and Construction.

[0064] By combining a fixed-time sliding window mechanism, data within a continuous time period is divided into multiple diagnostic segments. Each segment contains sampled data and its statistical characteristics over a continuous time period, including mean, variance, slope change rate, and periodic abrupt change points. Ambient temperature, freezing temperature, defrosting temperature, power, compressor status, compartment cooling status, and special modes are extracted from the time-series data uploaded from the refrigerator. A time-series information over a period of time is defined as a diagnostic segment. Organizing data into time-series segments facilitates the construction of analytical units with temporal context characteristics, enhancing the model's judgment capabilities.

[0065] From each diagnostic segment Extract key features from the data to construct a feature vector:

[0066]

[0067] in, This refers to the amount of refrigerant (g). For ambient temperature, This is the freezing temperature. This refers to the defrosting temperature of the freezer. For power, The compressor is in operation. The room is in cooling mode. This is a special mode.

[0068] S52, Model Building.

[0069] Using the XGBoost decision tree model As the core classifier, its training objective is:

[0070]

[0071] in, This indicates that it is normal. This indicates a risk of refrigerant leakage.

[0072] The XGBoost model is built upon the Gradient Boosting Decision Tree (GBDT) framework, employing an ensemble optimization strategy combining an additive model, gradient descent, and regularization. It is suitable for high-precision classification tasks with multi-dimensional feature inputs. Its complete process is as follows: Figure 1 As shown, it includes six key nodes: model initialization, weak learner construction, training completion, ensemble optimization, performance evaluation, and prediction execution.

[0073] The model is initialized using a CART tree as its basic structure, combined with a binary classification loss function (logloss) for initial modeling. The initial predicted value is usually set as the mean of the sample labels, serving as the starting point for subsequent iterative optimization.

[0074] In the construction of weak learners, a "weak learner," or a single decision tree, is generated in each iteration to fit the residuals (negative gradients) of the previous model. The splitting of the tree is based on maximizing information gain or minimizing the loss function, with the optimal split point selected through feature importance analysis. It supports both continuous and discrete features and automatically handles missing values.

[0075] Training stops when the preset number of iterations is reached, validation set performance no longer improves, or error converges. The final model parameters are output, including all subtree structures, leaf node scores, learning rate, etc., and can be used for inference deployment.

[0076] The ensemble optimization employs an additive model approach: for each newly added tree, its predictions are added to the overall output. Gradient descent is used to update the model parameters, causing the loss function to gradually converge. L1 / L2 regularization terms are introduced to prevent overfitting and control the complexity of the trees (such as the number of leaf nodes, depth, and leaf node weights).

[0077] Performance evaluation employs multiple metrics to assess model effectiveness: confusion matrix analysis for true positives, false positives, true negatives, and false negatives; accuracy, i.e., the overall percentage of correctly classified cases; AUC (Area Under ROC Curve) to measure the model's discriminative ability, particularly suitable for imbalanced class scenarios; and K-fold cross-validation to ensure model generalization ability. A key focus is on recall for "minor leaks" to avoid false negatives that could damage equipment.

[0078] During prediction, after new sample data is input, it is processed sequentially through each decision tree, with each tree outputting a score for a leaf node. All scores are summed to obtain the total prediction score; this score is then converted into a probability output using a sigmoid function. If the probability exceeds a set threshold (e.g., 0.7), it is determined that there is a risk of refrigerant leakage, triggering an early warning mechanism.

[0079] XGBoost effectively captures nonlinear combinations of features such as abnormal temperature, frequent compressor start-stop, and energy consumption fluctuations, enabling accurate identification of early micro-leakage. It integrates data from multiple sensors, including temperature, pressure, operating status, and door opening / closing frequency, and uses feature importance analysis to filter key indicators, improving model interpretability and robustness. Regularization mechanisms suppress noise interference, preventing misjudgments of leaks due to environmental changes (such as room temperature fluctuations). After training, the model can be deployed on a cloud-based large-scale model platform without requiring local hardware upgrades, adapting to the remote monitoring needs of large-scale refrigerator equipment.

[0080] S53, Determine the classification criteria and decision-making logic.

[0081] S531 sets the refrigerant quantity threshold leakage_threshold, and marks it as "leaking" when the inferred refrigerant quantity is less than or equal to the refrigerant quantity threshold.

[0082] When the refrigerant amount is less than or equal to the leakage threshold, it is marked as "leaking (1)"; otherwise, it is marked as "normal (0)". The refrigerant amount is indirectly assessed by the operating characteristics, which eliminates the need to directly measure the refrigerant amount and reduces the detection cost.

[0083] S532 uses historical data as the training set and future data as the test set, employing a "time series partitioning" strategy to prevent data leakage and ensure that the model has real predictive capabilities.

[0084] By employing a "time-series partitioning" approach, historical segments are used as the training set and future segments as the test set, thus avoiding the false high performance of "using future data to predict the past" and ensuring that the model can predict whether the future will be leaked.

[0085] The XGBoost model has strong nonlinear fitting capabilities and feature importance assessment functions, which can accurately identify typical refrigerant leakage patterns; the time-series partitioning method ensures the model's generalization performance in practical applications.

[0086] S533 uses linear regression to fit the slope of temperature and power changes to reflect the continuous changes in the equipment's cooling performance. The trend of temperature or power change over time can be calculated by linear regression. If the absolute value of the slope is small and continues to slow down, it indicates a decrease in cooling efficiency (such as insufficient refrigerant leading to weakened cooling capacity). An abnormal power slope may reflect an increase in compressor workload or frequent adjustments, indirectly indicating instability of the device.

[0087] S534 uses the power standard deviation to quantify fluctuations reflecting the stability of equipment cooling performance. This method quantifies fluctuations by calculating the standard deviation of power data over a period of time to reflect the stability of the device's operation. A larger standard deviation indicates more severe power fluctuations, indirectly reflecting frequent compressor start-stops or unstable loads due to insufficient refrigerant, thus indicating unstable cooling performance. This method is an indirect diagnostic approach based on time-series characteristics, requiring no additional sensors and reducing testing costs.

[0088] S535 uses the difference in state switching frequency to reflect whether the unit's cooling is frequently adjusted. It calculates the number of state switching times to detect abnormal behaviors such as frequent compressor adjustments and repeated temperature rises and falls in the compartments. By calculating the difference in changes in variables such as compressor state and compartment cooling state over time, it statistically analyzes the number of state switching times per unit time (e.g., start-stop frequency, cooling on / off frequency). An increased number of switching times indicates that the unit is frequently adjusting to maintain temperature, reflecting a decline in cooling performance and potential leakage risks.

[0089] Indirect assessment of operational characteristics avoids the high costs and technical difficulties associated with directly measuring refrigerant levels; through the fusion of multi-dimensional indirect features, it can comprehensively characterize the abnormal operation of the device caused by refrigerant leakage, thereby improving diagnostic sensitivity.

[0090] S54, Leakage Risk Assessment and Result Generation.

[0091] The final output is the diagnostic record for each segment:

[0092] All results are stored in structured files for subsequent user alerts and analysis. Structured diagnostic reports are provided to help operations and maintenance personnel quickly locate problems; clearly defined risk levels help in developing differentiated response strategies.

[0093] When the model outputs At that time, it was determined that there was a risk of refrigerant leakage in the refrigerator.

[0094] Automatically generate diagnostic results, including the leakage risk level (low / medium / high), key abnormal indicators (such as "compressor has been running continuously for more than 6 hours"), and the suspected leakage start time.

[0095] S6, dual-end information synchronization notification.

[0096] Reminders are sent to end users via pop-up windows in the user's app or SMS, informing them of the risk level and providing temporary response suggestions (such as "Do not open the refrigerator door for an extended period of time" or "Contact after-sales service for inspection"). Simultaneously, diagnostic results and key anomaly data are pushed to the manufacturer's maintenance management backend for after-sales service personnel to review and develop repair plans.

[0097] This establishes a closed-loop management process encompassing "cloud-based diagnostics - user notification - after-sales handling": once a high-risk leakage event is confirmed, the device automatically creates a work order and assigns it to the nearest after-sales service center, automating fault response. It achieves synchronization between the cloud and the user, improving service response efficiency; enhances user experience and sense of security; and promotes proactive home appliance health management.

[0098] S7, Report Generation and Closed-Loop Processing.

[0099] S71, after the diagnostic process is completed, the device automatically generates a "User Usage Report" and a "Refrigerator Health Report", which cover recent operating status, abnormal event records, energy-saving suggestions, etc.

[0100] S72, maintenance personnel rectify the faulty refrigerator based on the report, update the device status after completing the repair, and form a complete business loop.

[0101] Provide data support for after-sales rectification and improve service quality; accumulate long-term data for product optimization and user behavior research.

[0102] This embodiment breaks through the limitations of traditional "post-event detection" and can issue early warnings to users via APP or SMS at the early stage of minor refrigerant leaks. This allows users and maintenance personnel to respond in a timely manner and take intervention measures, effectively preventing the leak from spreading, avoiding refrigerator malfunctions, and improving equipment reliability and safety.

[0103] By integrating multi-dimensional serial port data and combining key indicators such as temperature change slope, power fluctuation characteristics, and state switching frequency, and using machine learning models such as decision trees and XGBoost for cross-validation and in-depth analysis, high-precision identification of refrigerant micro-leakage is achieved, promoting refrigerator fault diagnosis from "experience-based and extensive" to "intelligent and precise".

[0104] This intelligent diagnostic technology can be deployed on a large-scale cloud platform without the need for dedicated sensors or professional testing equipment and manual inspections. It significantly reduces hardware investment and maintenance costs, enabling low-cost, large-scale, and automated refrigerant leak monitoring, and helping home appliance services transform towards digitalization and AI-driven approaches.

[0105] Example 2

[0106] This embodiment discloses a device for intelligent diagnosis of refrigerant leaks in refrigerators. The device collects multi-dimensional serial port data during the refrigerator's operation and combines cloud computing and machine learning algorithms to achieve early identification and risk warning of minor refrigerant leaks. This breaks through the limitations of traditional "post-event detection" and improves the intelligence and accuracy of fault diagnosis.

[0107] The device includes: a data reporting module, a data transmission and storage module, an intelligent diagnosis module, and an early warning and feedback module. The data reporting module and the intelligent diagnosis module are as follows: Figure 2 As shown.

[0108] The data reporting module is located in the refrigerator's local control unit and is used to collect key parameters during the refrigerator's operation in real time, including but not limited to: ambient temperature, freezer temperature, refrigerator temperature, freezer defrosting temperature, refrigerator defrosting temperature, real-time power, compressor operating status, refrigerator set operating mode, freezer door open / close status, refrigerator door open / close status, compressor on / off ratio, and compartment cooling status. The data is transmitted to the local communication unit via serial communication and undergoes digital signal encryption and verification processing to ensure data integrity and security.

[0109] The data transmission and storage module is used to upload data to the cloud server. The local communication unit uploads encrypted data to the cloud server via WiFi or 4G / 5G networks. Security protocols are used during data transmission to ensure privacy and prevent tampering. A dedicated database in the cloud is used to receive, store, and manage historical and real-time operational data from multiple devices, providing data support for subsequent analysis.

[0110] The intelligent diagnostic module (cloud) is deployed on a large model platform in the cloud and includes the following sub-functional modules:

[0111] Data processing module: Cleans and normalizes the received raw data, and removes outliers and noise interference;

[0112] Feature Importance Analysis Module: Based on multi-dimensional cross-validation methods (such as correlation analysis, principal component analysis, mutual information, etc.), extract combined features that are strongly correlated with refrigerant leakage, such as abnormal temperature rise slope, frequent compressor start-stop, increased unit energy consumption, and extended defrosting cycle.

[0113] Refrigerant Leakage Algorithm Model: Machine learning models such as XGBoost, Random Forest, or Deep Neural Networks are used to train and establish a leakage detection model based on historical fault samples. The input is the aforementioned multi-dimensional features, and the output is the probability of leakage. A feedback learning mechanism is configured to receive manually confirmed diagnostic results as label data, and the model is periodically subjected to incremental training and performance optimization.

[0114] Risk assessment conclusion module: Based on the model output results, set the graded early warning thresholds (such as low risk, medium risk, high risk) and generate the final diagnostic conclusion.

[0115] The early warning and feedback module is used to trigger the early warning mechanism and distribute diagnostic results to the operation and maintenance management device and user terminal. When the diagnostic results indicate a potential leakage risk, the device automatically triggers the early warning mechanism: pushes a notification to the user's mobile APP or sends an SMS reminder, informing them of the possible refrigerant leakage problem and suggested measures; simultaneously, it uploads a detailed diagnostic report to the manufacturer's operation and maintenance backend, facilitating remote investigation or on-site service arrangements by after-sales personnel.

[0116] This device requires no additional sensors or specialized testing equipment. Utilizing data from existing built-in refrigerator sensors combined with cloud-based AI algorithms, it can intelligently identify and proactively warn of minor refrigerant leaks. This improves the timeliness and accuracy of fault prediction, reduces the cost and complexity of manual inspections, and drives the transformation of home appliances from passive repair to proactive health management.

[0117] This embodiment provides a low-cost, highly reliable, and scalable intelligent diagnostic device suitable for refrigerant leakage monitoring in various intelligent refrigerator products.

[0118] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. A method for intelligently diagnosing refrigerant leaks, characterized in that, The method is applicable to networked refrigeration equipment and includes the following steps: Collect multi-dimensional serial port data during the operation of refrigeration equipment; The collected data is validated and encrypted before being uploaded to the cloud-based big data platform. Uploaded data is cleaned, missing values ​​are filled, and outliers are handled in the cloud. Key features are extracted to construct time-series diagnostic segments. These segments are then analyzed using a machine learning decision tree model to determine if there is a risk of refrigerant leakage. When a risk of refrigerant leakage is detected, a diagnostic result is generated and pushed to the operation and maintenance backend and user terminals.

2. The method according to claim 1, characterized in that, The multi-dimensional serial port data includes one or more of the following: ambient temperature, freezer temperature, refrigerator temperature, freezer defrosting temperature, refrigerator defrosting temperature, real-time power, compressor operating status, refrigerator set operating mode, freezer door open / close status, refrigerator door open / close status, compressor start / stop ratio, and compartment cooling status.

3. The method according to claim 1, characterized in that, Each of the aforementioned time-series diagnostic segments contains sampled data over a continuous time period.

4. The method according to claim 1, characterized in that, The machine learning decision tree model is an XGBoost model, which is trained using historical normal operation data and known refrigerant leak sample data to learn the typical operating mode change patterns before and after refrigerant leaks.

5. The method according to claim 1, characterized in that, The logic for judging refrigerant leakage is as follows: whether the refrigerant quantity meets the standard is indirectly determined by the operating characteristics, without the need to directly measure the refrigerant quantity.

6. The method according to claim 5, characterized in that, The operational characteristics include: using linear regression to fit the slope of temperature and power changes to reflect the continuous changes in the equipment's cooling performance.

7. The method according to claim 5, characterized in that, The operational characteristics also include: using the power standard deviation to quantify fluctuations to reflect whether the equipment's cooling performance is stable.

8. The method according to claim 5, characterized in that, The operational characteristics also include: using the difference in the number of state switching to reflect whether the device's cooling is frequently adjusted.

9. The method according to claim 1, characterized in that, The diagnostic results include at least one of the following: risk level, abnormal indicators, and suspected leakage time.

10. The method according to claim 9, characterized in that, The diagnostic results are simultaneously pushed to the user's APP or notified to the user via SMS, informing them of the risk level and temporary response suggestions; at the same time, the diagnostic results and key abnormal data are pushed to the manufacturer's operation and maintenance management backend for after-sales service personnel to view and formulate maintenance plans.

11. The method according to claim 1, characterized in that, It also includes conducting feature importance analysis, performing multi-dimensional cross-validation, and eliminating interference.

12. A device for intelligently diagnosing refrigerant leaks in a refrigerator, characterized in that, The device for intelligent diagnosis of refrigerant leaks using the method described in any one of claims 1 to 11 includes: The data reporting module is used to obtain multi-dimensional operating parameters from the networked refrigerator controller and perform digital signal encryption and verification processing. The data transmission and storage module is used to upload data to the cloud server; The intelligent diagnostic module includes a data processing module, a feature importance analysis module, a refrigerant leakage algorithm model, and a risk assessment conclusion module. The early warning and feedback module is used to trigger the early warning mechanism and distribute the diagnostic results to the operation and maintenance management device and user terminal.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 11.

14. A refrigerator, characterized in that, Intelligent diagnosis of refrigerant leakage is performed using the method described in any one of claims 1 to 11.

Images