Fault prediction method, control device, electronic equipment and storage medium

Abstract

The invention provides a fault prediction method, a control device, electronic equipment and a storage medium, and is suitable for the technical field of refrigerators, the method comprises the following steps: obtaining operation data of the refrigerator, the operation data comprising acoustic data, vibration data, temperature data and current data; performing anomaly calculation based on the operation data and a preset isolated forest model, and obtaining an anomaly score; and performing local anomaly early warning based on the anomaly score and a preset anomaly threshold. According to the fault prediction method provided by the invention, the multi-dimensional operation data of the refrigerator is acquired, the anomaly calculation is performed on the multi-dimensional operation data by using the isolated forest model, the weak change signal is detected from multiple dimensions, the reliability of the calculation result is ensured, and finally, local early warning is performed according to the calculated anomaly score, so that the fault prediction accuracy is improved. Abnormality early warning is carried out at the early stage of the fault, and the situation that the refrigerator is shut down to be overhauled after the fault seriously affects operation of the refrigerator, and use is affected is avoided.

Description

Technical Field

[0001] This application belongs to the field of refrigerator technology, and particularly relates to a fault prediction method, control device, electronic device and storage medium. Background Technology

[0002] Existing refrigerator fault detection systems typically only issue a fault alert after the refrigerator has already malfunctioned, when the sensor readings significantly exceed the preset range, requiring the refrigerator to be shut down for repairs and affecting its use. Summary of the Invention

[0003] This application provides a fault prediction method, control device, electronic device, and storage medium to solve the problem that existing refrigerator fault detection systems typically only issue fault alerts after a serious fault has already occurred, requiring shutdown for repairs and affecting usability.

[0004] In a first aspect, embodiments of this application provide a fault prediction method applicable to refrigerators, the method comprising: The refrigerator's operating data is acquired, including acoustic data, vibration data, temperature data, and current data. Anomaly calculations are performed based on the aforementioned operational data and a preset isolated forest model, and anomaly scores are obtained. Local anomaly warnings are issued based on the anomaly score and a preset anomaly threshold.

[0005] Secondly, embodiments of this application also provide a fault prediction and control device, applicable to a refrigerator, the device comprising: The operation data acquisition module is configured to acquire the operation data of the refrigerator, including acoustic data, vibration data, temperature data and current data; The analysis module is configured to perform anomaly calculations based on the running data and a preset isolated forest model, and obtain anomaly scores. The early warning module is configured to provide local early warnings based on the anomaly score and a preset anomaly threshold.

[0006] Optionally, the anomaly calculation based on the running data and the preset isolated forest model includes: The running data is preprocessed to obtain processed running data; Feature extraction is performed on the processed running data to obtain a feature vector; The feature vector is input into the preset isolated forest model for anomaly calculation.

[0007] Optionally, inputting the feature vector into the preset isolated forest model for anomaly calculation includes: The feature vector is traversed through each isolated tree in the preset isolated forest model; Calculate the path length required for the feature vector to be isolated on each isolated tree; Calculate the mean path length on each isolated tree; The outlier score is determined based on the mean.

[0008] Optionally, it also includes: If the anomaly score is higher than the preset anomaly threshold, the feature vector is uploaded to the processing device for fault type analysis.

[0009] Thirdly, embodiments of this application also provide a fault prediction method, applicable to processing equipment, the method comprising: In response to the received feature vector, a fault type analysis is performed on the feature vector based on a preset fault diagnosis model; Based on the fault type obtained from the analysis, an early warning message is generated and sent to the user's mobile device; The feature vector is generated by the refrigerator performing anomaly calculations based on operating data and a preset isolated forest model to obtain an anomaly score. When the anomaly score is higher than a preset anomaly threshold, it is uploaded to the processing device. The feature vector is obtained by the refrigerator preprocessing the collected operating data and extracting features from the processed operating data. The operating data includes acoustic data, vibration data, temperature data, and current data.

[0010] Optionally, the fault type analysis of the feature vector based on the preset fault diagnosis model includes: Based on the classification boundary or decision function of the preset fault diagnosis model, the feature vector is mapped to the preset fault category space to obtain the mapped feature information. Based on the mapping feature information, a fault probability distribution vector is output, where each vector in the fault probability distribution vector represents a fault type and its fault probability.

[0011] Optionally, the preset fault diagnosis model is an integrated model of one or more of the following: support vector machine model, random forest or gradient boosting decision tree model, Naive Bayes classifier, artificial neural network model, convolutional neural network model or recurrent neural network model.

[0012] Fourthly, this application also provides a fault prediction and control device suitable for processing equipment. The device includes an analysis module and an early warning module. The analysis module is configured to perform fault type analysis on the feature vector based on a preset fault diagnosis model in response to a received feature vector. The early warning module is configured to generate early warning information based on the fault type obtained from the analysis and send it to the user's mobile device. The feature vector is obtained by the refrigerator performing anomaly calculation based on operating data and a preset isolated forest model, and then uploading it to the processing equipment when the anomaly score is higher than a preset anomaly threshold. The operating data includes acoustic data, vibration data, temperature data, and current data.

[0013] Fifthly, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the fault prediction method as described above.

[0014] Sixthly, embodiments of this application also provide a storage medium storing control instructions, which, when executed by a processor, implement the fault prediction method described above.

[0015] In a seventh aspect, embodiments of this application also provide a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations described in embodiments of this application.

[0016] The fault prediction method provided in this application addresses the issue that in the early stages of a fault, the impact on a single data point of the refrigerator is relatively weak, making it difficult to directly detect the fault state from a single data point. Therefore, by acquiring multi-dimensional operating data of the refrigerator and using an isolated forest model to perform anomaly calculations on the multi-dimensional operating data, the method enables the detection of weakly changing signals from multiple dimensions, ensuring the reliability of the calculation results. Finally, based on the calculated anomaly score, a local early warning is issued, enabling anomaly warnings to be issued in the early stages of a fault, avoiding the need for shutdown and repair only after the fault has severely affected the refrigerator's operation, thus avoiding disruption to its use. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] To gain a more complete understanding of this application and its beneficial effects, the following description will be provided in conjunction with the accompanying drawings. In the following description, the same reference numerals denote the same parts.

[0019] Figure 1 This is a flowchart illustrating a fault prediction method provided in an embodiment of this application.

[0020] Figure 2 This is a flowchart illustrating a fault prediction method provided in another embodiment of this application.

[0021] Figure 3 This is a flowchart illustrating a fault prediction method provided in another embodiment of this application.

[0022] Figure 4 This is a schematic diagram of the structure of a fault prediction and control device provided in an embodiment of this application.

[0023] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0024] Explanation of icon numbers: 301. Data Acquisition Module; 302. Analysis Module; 303. Early Warning Module; 400. Electronic device; 401. Memory; 402. Processor; 4011. Computer program. Detailed Implementation

[0025] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0026] In the description of the embodiments of this application, "module" and "processor" can include hardware, software, or a combination of both. A module can include hardware circuitry, various suitable sensors, communication ports, and memory, and may also include software components, such as program code, or a combination of software and hardware. A processor can be a central processing unit, a microprocessor, a digital signal processor, or any other suitable processor. The processor has data and / or signal processing capabilities. The processor can be implemented in software, in hardware, or a combination of both. Non-transitory computer-readable storage media includes any suitable medium capable of storing program code, such as magnetic disks, hard disks, optical disks, flash memory, read-only memory, random access memory, etc.

[0027] This application provides a fault prediction method, control device, electronic device, and storage medium to solve the problem that existing refrigerator fault detection systems usually only issue fault warnings after a serious fault has occurred in the refrigerator, requiring shutdown for repair and affecting its use. The following description is in conjunction with the accompanying drawings.

[0028] The fault prediction method provided in this application is applicable to refrigerators. Please refer to [link / reference]. Figure 1 The method includes the following steps: S101: Acquire the refrigerator's operating data, including acoustic data, vibration data, temperature data, and current data.

[0029] Refrigerator operating data refers to the collection of all digital information that a refrigerator collects, generates, and records in real time throughout its operating lifecycle, using its built-in sensors, controllers, and intelligent modules. This data reflects the refrigerator's status, performance, environment, and usage.

[0030] Acoustic data refers to the sound signals generated by the refrigerator during operation, such as the hum of the compressor and the sound of refrigerant flow, which can be acquired using a microphone sensor (such as a MEMS microphone). Vibration data can be acquired using an accelerometer. Temperature data can be acquired using a temperature sensor. Current data can be acquired using a current transformer or a Hall effect current sensor.

[0031] S102: Perform anomaly calculation based on runtime data and a preset isolated forest model, and obtain anomaly scores.

[0032] The preset isolated forest model is an offline model built on the isolated forest algorithm. It is obtained by training the offline model with a large amount of training data (historical operating data of the refrigerator). Specifically, its main process includes: data collection, preprocessing and feature extraction, model training, model validation and threshold determination.

[0033] For example, data collection includes collecting four types of time-series data—acoustic, vibration, temperature, and current—from a batch of normally operating (preferably brand new or in good condition) refrigerators over a long period of time. The amount of data should be large enough to cover different operating modes of the refrigerators (such as compressor start-up, operation, and shutdown, defrosting cycles, fan speed changes, etc.).

[0034] Preprocessing and feature extraction include cleaning and alignment (handling missing values, removing obvious noise, and ensuring that the timestamps of multi-sensor data are aligned), feature extraction (raw time series data cannot be directly used for the model, and meaningful features need to be extracted for each data window (e.g., every 5 minutes or each compressor run cycle)), and feature standardization (scaling all features to a similar numerical range (e.g., using Z-score standardization) to prevent differences in units from affecting the model).

[0035] Model training involves using the feature matrix of all preprocessed normal data as the training set, allowing the offline model to learn the "multi-dimensional feature space distribution pattern of normal state". The model constructs multiple isolated trees, which define the "normal contour" of the data. For example, a feature such as a relevant feature of vibration data is randomly selected, and a cut point is randomly selected within the current data range (such as a random number between the minimum and maximum vibration values). Based on this cut point, the data is divided into two branches (left branch: vibration value ≤ cut point; right branch: vibration value > cut point). The above process is recursively repeated on each branch until either of the following conditions is met: the data point is isolated (each leaf node has only one data point), or the tree reaches the preset maximum depth (max_depth). The above process is repeated to construct a large number of isolated trees. Each tree, due to the random selection of features and cut points, provides a "division perspective" of the data space from different angles. The "normal contour" of the data is defined through these trees.

[0036] Model validation and threshold determination involve inputting a portion of normal data (not used in training) into the trained model, calculating the anomaly score for each sample, and then using these anomaly scores to form a distribution (usually close to a normal distribution) to obtain a "preset anomaly threshold". The trained model can be tested using known, simulated fault data (such as fan jamming or compressor wear) to see if its anomaly score significantly exceeds the threshold, thus verifying the model's sensitivity.

[0037] S103: Provide local anomaly warnings based on anomaly scores and preset anomaly thresholds. For example, the anomaly thresholds can be 0.6, 0.678, 0.7, 0.8, 0.9, etc., and the specific data can be further determined as needed.

[0038] The method of local anomaly warning is not further limited here. For example, it can be a visual prompt, such as a display screen / control panel warning: display of dedicated fault codes (such as "E1", "F2", etc.), text prompts (such as "compressor malfunction", "fan malfunction"), flashing icons (warning triangle, exclamation mark icon), LED indicator color change (green light → red light, or breathing flashing mode), ambient light feedback (the refrigerator's internal lighting changes color (such as normal white light, abnormal red light)), and door-side ambient light strip color change warning; or it can be an auditory prompt, such as a buzzer / speaker alarm: intermittent buzzing (such as "beep beep beep" three times followed by a pause), continuous long beep (emergency fault), voice broadcast (such as "an operational abnormality has been detected, please contact after-sales service"), and melodic alarm sound (different melodies correspond to different faults); or it can be a tactile prompt: door vibration feedback (slight vibration when the user touches the door handle), smart knob vibration (the temperature control knob provides tactile feedback), etc.

[0039] The fault prediction method provided in this application addresses the issue that in the early stages of a fault, the impact on a single data point of the refrigerator is relatively weak, making it difficult to directly detect the fault state from a single data point. Therefore, by acquiring multi-dimensional operating data of the refrigerator and using an isolated forest model to perform anomaly calculations on the multi-dimensional operating data, the method enables the detection of weakly changing signals from multiple dimensions, ensuring the reliability of the calculation results. Finally, based on the calculated anomaly score, a local early warning is issued, enabling anomaly warnings to be issued in the early stages of a fault, avoiding the need for shutdown and repair only after the fault has severely affected the refrigerator's operation, thus avoiding disruption to its use.

[0040] Optionally, anomaly calculation is performed based on the running data and a preset isolated forest model, including: preprocessing the running data to obtain processed running data; extracting features from the processed running data to obtain feature vectors; and inputting the feature vectors into the preset isolated forest model for anomaly calculation.

[0041] Preprocessing can include data cleaning, filtering and noise reduction, timestamp alignment, and standardization.

[0042] For acoustic data, its root mean square characteristics in the time domain, peak-to-peak characteristics, peak frequency characteristics in the frequency domain, and harmonic characteristics can be extracted; for vibration data, its vibration amplitude characteristics in the time domain, waveform distortion factor characteristics, frequency band division characteristics in the frequency domain, and resonance offset characteristics can be extracted; for temperature data, its temperature value characteristics and temperature fluctuation amplitude characteristics can be extracted; for current data, its effective current value characteristics, peak-to-peak characteristics, and harmonic characteristics can be extracted.

[0043] A feature vector is a multidimensional numerical array composed of multiple extracted features.

[0044] By preprocessing operational data to eliminate sensor noise, standardize data formats, and handle missing values, data quality and consistency are ensured, preparing for feature extraction. Feature extraction is then performed on the processed operational data to achieve information condensation and dimensionality reduction. For example, vibration waveforms lasting several seconds can be condensed into a few key values ​​(such as peak value and frequency components), significantly reducing the data volume. Furthermore, multi-dimensional features are integrated, such as combining the correlation features of sound, vibration, temperature, and electricity to form a "comprehensive profile" of the equipment status. Finally, the feature vectors are input into a pre-defined isolated forest model for anomaly calculation. Without prior knowledge of all fault types, new anomalies can be detected by calculating whether the feature vectors deviate from the "normal pattern." Moreover, before a fault fully manifests, early, minute feature changes can capture impending faults, enabling predictive maintenance. The entire process from data input to output score can be completed in milliseconds, achieving truly automated real-time monitoring without human intervention.

[0045] Optionally, the feature vector is input into a preset isolated forest model for anomaly calculation, including: traversing each isolated tree in the preset isolated forest model with the feature vector; calculating the path length required for the feature vector to be isolated on each isolated tree; calculating the mean of the path length on each isolated tree; and determining the anomaly score based on the mean.

[0046] For example, the following formula can be used to determine outlier scores based on the mean: ; ; Where s is the anomaly score, x is the data point to be detected (i.e., the extracted feature vector), n is the number of training samples used to construct the isolation forest, h(x) is the path length of data point x in a certain isolation tree, E(h(x)) is the average path length of data point x in all isolation trees, c(n) is the standardization coefficient for a given number of samples n, which is the expected value of the path length and is used to standardize E(h(x)), and H(n-1) is the harmonic number, which can be approximated by the natural logarithm. H(n-1)≈ln(n-1)+γ (γ is the Euler-Marcheroni constant, approximately equal to 0.5772156649).

[0047] For example, suppose four standardized features are extracted from the refrigerator data to represent the state of a certain time window: X = [feature 1: 0.2, feature 2: 1.1, feature 3: -0.5, feature 4: 0.8]; suppose the isolated forest consists of 3 isolated trees (iTree) (the actual number of isolated trees is usually much greater than 3, but for the sake of illustration, only 3 trees are used as an example here), suppose the subsample size n = 6 during training, and the corresponding standardization coefficient c(6) ≈ 3.67 (calculated by the formula c(n) = 2H(n-1) - 2(n-1) / n, where H(5) ≈ 2.28).

[0048] First, assume the decision rule for isolated tree 1 is as follows: Root node: Determine if feature 2 is less than or equal to 0.7. If yes, proceed to the left child node; otherwise, proceed to the right child node. If the right child node is reached, continue to check if feature 3 is less than or equal to 0.1. If it is, go to the left child node (depth=2); otherwise, go to the right child node (depth=2). If the left child node is reached (from the root node), continue to check if feature 1 is less than or equal to 0.5. If yes, go to the left child node (depth=2); otherwise, go to the right child node (depth=2).

[0049] Suppose the decision rule for isolated tree 2 is as follows: Root node: Determine if feature 4 is less than or equal to 0.3. If yes, go to the left child node (depth=1); otherwise, go to the right child node (depth=1). If the right child node is reached, continue to check if feature 1 is less than or equal to -0.2. If yes, go to the left child node (depth=2); otherwise, go to the right child node (depth=2). Suppose the decision rule for isolated tree 3 is as follows: Root node: Determine if feature 3 is less than or equal to 1.0. If yes, go to the left child node; otherwise, go to the right child node (depth = 1). If the right child node is reached, continue to check if feature 2 is less than or equal to 0.9. If it is, go to the left child node; otherwise, go to the right child node (depth = 2). If the left child node is reached (depth=2), continue to check if feature 4 is less than or equal to 0.6. If yes, go to the left child node (depth=3); otherwise, go to the right child node (depth=3).

[0050] Secondly, calculate the path length required for the feature vector to be isolated on each tree, for the feature vector X=[0.2,1.1,-0.5,0.8]: The path on isolated tree 1 is as follows: Root node: Is feature 2 less than or equal to 0.7, that is, is 1.1 less than or equal to 0.7? The answer is no, so it enters the right child node (passing 1 edge). Current node: Is feature 3 less than or equal to 0.1, that is, is -0.5 less than or equal to 0.1? The answer is yes, so it reaches the left leaf node (passing 1 more edge). Therefore, the total path length h1(X) = 2.

[0051] The path on isolated tree 2 is as follows: Root node: Is feature 4 less than or equal to 0.3, that is, is 0.8 less than or equal to 0.3? The answer is no, so it enters the right child node (passing 1 edge). Current node: Is feature 1 less than or equal to -0.2, that is, is 0.2 less than or equal to -0.2? The answer is no, so it reaches the right leaf node (passing 1 more edge). Therefore, the total path length h2(X) = 2.

[0052] The path on isolated tree 3 is as follows: Root node: Is feature 3 less than or equal to 1.0, that is, is -0.5 less than or equal to 1.0? The answer is yes, so it enters the left child node (passing 1 edge). Current node: Is feature 2 less than or equal to 0.9, that is, is 1.1 less than or equal to 0.9? The answer is no, so it reaches the right leaf node (passing 1 more edge). Therefore, the total path length h3(X) = 2.

[0053] Next, calculate the mean path length on each tree: E(h(X)) =[ h1(X)+ h1(X)+ h1(X)] / 3=(2+2+2) / 3=2 Finally, outlier scores were determined based on the mean (where n=6, c(n)=3.67): S=0.685.

[0054] The final anomaly score is 0.685. By comparing this anomaly score with the preset anomaly threshold, it can be determined whether the refrigerator is currently in an abnormal state. For example, if the preset anomaly threshold is 0.6, it indicates that the refrigerator is currently in an abnormal state and a local anomaly warning can be issued.

[0055] Optionally, it also includes: if the anomaly score is higher than a preset anomaly threshold, uploading the feature vector to the processing device for fault type analysis. In some examples, the processing device can be a cloud-based device.

[0056] Since the refrigerator's default isolated forest model is only used to calculate whether the current state is abnormal, but cannot directly determine the specific fault type, when the abnormality score is higher than the default abnormality threshold, the feature vector is uploaded to the processing device. The processing device analyzes the feature vector to obtain the specific fault type for subsequent adjustment work.

[0057] Optionally, after the feature vector is uploaded to the processing device, the processing device performs fault type analysis on the feature vector based on a preset fault diagnosis model, generates early warning information based on the fault type obtained from the analysis, and sends it to the user's mobile device. The fault type analysis of the feature vector based on the preset fault diagnosis model includes: the processing device maps the feature vector to the preset fault category space based on the classification boundary or decision function of the preset fault diagnosis model to obtain the mapped feature information; the processing device outputs the fault probability distribution vector according to the mapped feature information, where each vector in the fault probability distribution vector represents a fault type and its fault probability.

[0058] The preset fault diagnosis model is an integrated model of one or more of the following: support vector machine model, random forest or gradient boosting decision tree model, Naive Bayes classifier, artificial neural network model, convolutional neural network model or recurrent neural network model.

[0059] Optionally, when the abnormal score is less than or equal to the preset abnormal threshold, the refrigerator's operating data can be retrieved again.

[0060] Optionally, when the anomaly score exceeds a preset anomaly threshold, the preset isolated forest model is updated using the current running data and the anomaly score to further optimize the model's computational power.

[0061] In summary, the fault prediction method for refrigerators provided above enables the detection of minor, progressive faults locally on the refrigerator using a preset isolated forest model and real-time refrigerator operating data, thus preventing serious faults that would lead to downtime for repairs. This method achieves rapid fault detection without relying on the cloud.

[0062] This application also provides a fault prediction method, applicable to processing devices such as the cloud. Please refer to [link to relevant documentation]. Figure 2 The method includes the following steps: S201: In response to the received feature vector, perform fault type analysis on the feature vector based on the preset fault diagnosis model.

[0063] The feature vector is generated by the refrigerator calculating anomalies based on operational data and a preset isolated forest model, and then uploading it to the processing device when the anomaly score exceeds a preset anomaly threshold. The feature vector is obtained by the refrigerator preprocessing the collected operational data and extracting features from the processed data. The operational data includes acoustic data, vibration data, temperature data, and current data. Fault types include, but are not limited to, compressor faults (such as start-up failures and abnormal operation), refrigerant leaks, capillary tube / expansion valve blockages, evaporator faults, condenser faults, circulation system faults, electrical control system faults, and defrosting system faults.

[0064] S202: Generate early warning information based on the fault type obtained from the analysis and send it to the user's mobile device. The mobile device may include, but is not limited to, mobile phones, tablets, and other devices.

[0065] The feature vectors are processed by the processing equipment to obtain the specific fault type, thereby saving the refrigerator's local computing power and reducing the refrigerator's computational burden. At the same time, the refrigerator directly uploads the processed feature vectors to the processing equipment, avoiding the processing equipment from processing the running data again until it obtains the feature vectors, thus saving computing resources on the processing equipment side. Finally, the fault type is analyzed by the processing equipment and sent directly to the user's mobile device, forming a complete service loop. This realizes the transformation from passive repair to proactive care, improving the product's service capabilities and competitiveness.

[0066] Optionally, fault type analysis is performed on the feature vector based on a preset fault diagnosis model, including: mapping the feature vector to a preset fault category space based on the classification boundary or decision function of the preset fault diagnosis model to obtain mapped feature information; and outputting a fault probability distribution vector based on the mapped feature information, where each vector in the fault probability distribution vector represents a fault type and its fault probability.

[0067] Among them, the mapping feature information is a new representation after transforming the original feature vector into a preset fault category space. It is not a simple label, but a multi-dimensional coordinate rich in diagnostic semantics.

[0068] The classification boundary is a judgment rule. When a new feature vector appears, the classification boundary can be used to determine which side of the boundary the feature vector falls on, thereby determining the fault type.

[0069] A decision function is a computational rule; it's a mathematical function that calculates the "membership degree" of a feature vector belonging to a certain fault category. It's not a simple "yes / no" judgment, but rather a calculation of "to what extent," such as a linear decision function (f(x) = w1·feature1 + w2·feature2 + ... + w...). n ·feature n + b), a distance-based decision function (f_k(x) = -distance(x, the center point of fault category k)), or a probabilistic decision function (such as Softmax) (P(fault k|x)=exp(z_k) / ∑exp(z_i), where z_k is the evidence score for the fault category).

[0070] For example, the original feature vector is x = [Vibration high-frequency energy: 0.78, vibration dominant frequency offset: 0.65, current harmonics: 0.42, peak starting current: 0.30, steady-state current: 0.25, cooling rate: 0.85, temperature fluctuation: 0.15, door opening / closing frequency: 0.20, ambient temperature: 0.35, compressor operating percentage: 0.40]. The original scores calculated using the decision function are: f1(x) = 2.85 (compressor mechanical wear); f2(x) = 1.20 (fan bearing damage); f3(x) = 0.45 (capillary blockage); f4(x) = 0.30 (door seal leakage); f5(x) = 0.75 (capacitor aging); f0(x) = -1.50 (normal state) (baseline). Using Softmax, the probability distribution is: e²· 85 =17.29 (compressor); e¹·² 0 =3.32 (fan); e 0 · 45 =1.57 (capillary); e 0 ·³ 0 =1.35 (door seal); e 0 · 75 =2.12 (capacitance); e - ¹· 50=0.22 (normal); Total = 17.29 + 3.32 + 1.57 + 1.35 + 2.12 + 0.22 = 26.87. That is, P (compressor wear) = 17.29 / 26.87 = 64.3%; P (fan damage) = 3.32 / 26.87 = 12.4%; P (capillary tube blockage) = 1.57 / 26.87 = 5.8%; P (door seal leakage) = 1.35 / 26.87 = 5.0%; P (capacitor aging) = 2.12 / 26.87 = 7.9%; P (normal state) = 0.22 / 26.87 = 0.8%.

[0071] Therefore, the final probability distribution vector is the fault probability distribution vector = [ {"Fault Type": "Compressor Mechanical Wear", "Probability": 64.3%}, {"Fault Type": "Fan bearing damaged", "Probability": 12.4%}, {"Fault Type": "Capacitor Aging", "Probability": 7.9%}, {"Fault Type": "Capillary Blockage", "Probability": 5.8%}, {"Fault Type": "Door Seal Leakage","Probability": 5.0%}, {"Fault Type": "Normal State","Probability": 0.8%} ].

[0072] Optionally, the preset fault diagnosis model is an integrated model of one or more of the following: support vector machine model, random forest or gradient boosting decision tree model, Naive Bayes classifier, artificial neural network model, convolutional neural network model or recurrent neural network model.

[0073] Optionally, please refer to Figure 3 The processing equipment can also calculate a health score based on feature vectors, thereby assessing the refrigerator's health status. Users can intuitively see the refrigerator's current health status based on the health status assessment, so that they can perform corresponding maintenance. If the current health status is good, appropriate measures can be taken immediately to deal with the current minor faults, or the user can deal with the minor faults when they have free time. If the current health status is poor, a repair technician can be called to perform a comprehensive inspection immediately.

[0074] For example, the processing device can directly calculate the health score based on the received anomaly score, or it can calculate the health score based on the obtained fault probability distribution vector. Further explanation is not provided here.

[0075] Optionally, the processing device can perform graded warning processing based on the level of the warning information. For example, when the warning processing is at level one, the warning information is sent to the user's mobile device to remind the user to perform maintenance; when the warning processing is at level two, the warning information is sent to the user's mobile device and the after-sales platform; when the warning processing is at level three, the warning information is sent to the user's mobile device and the after-sales platform and an after-sales work order is generated, so that the manufacturer can track and handle the corresponding problems in a timely manner and update the solutions to the corresponding problems.

[0076] Optionally, after generating the warning information, the method further includes: using the warning information and corresponding operational data to update the preset fault diagnosis model, thereby improving the model's sustainable learning ability.

[0077] This application also provides a fault prediction and control device suitable for refrigerators. Please refer to [link / reference]. Figure 4 The device includes an operation data acquisition module 301, an analysis module 302, and an early warning module 303. The operation data acquisition module 301 is configured to acquire the refrigerator's operation data, including acoustic data, vibration data, temperature data, and current data. The analysis module 302 is configured to perform anomaly calculations based on the operation data and a preset isolated forest model, and obtain anomaly scores. The early warning module 303 is configured to provide local anomaly warnings based on the anomaly scores and preset anomaly thresholds.

[0078] The refrigerator uses this fault prediction and control device to acquire multi-dimensional operating data, and uses an isolated forest model to perform anomaly calculations on the multi-dimensional operating data. This enables the detection of subtle changes in signals from multiple dimensions, ensuring the reliability of the calculation results. Finally, it provides local early warnings based on the calculated anomaly scores, enabling early warnings of anomalies at an early stage of a fault, avoiding the need for shutdown and repair only after the fault has severely affected the refrigerator's operation.

[0079] This application embodiment also provides a fault prediction and control device suitable for processing equipment. The device includes an analysis module and an early warning module. The analysis module is configured to perform fault type analysis on the feature vector based on a preset fault diagnosis model in response to a received feature vector. The early warning module is configured to generate early warning information based on the fault type obtained from the analysis and send it to the user's mobile device. The feature vector is obtained by the refrigerator performing anomaly calculation based on operating data and a preset isolated forest model, and then uploading it to the processing equipment when the anomaly score is higher than a preset anomaly threshold. The operating data includes acoustic data, vibration data, temperature data, and current data.

[0080] Through this fault prediction and control device, the processing equipment can quickly identify fault types and provide early warning responses, saving local computing power and reducing the computational burden on the refrigerator. At the same time, the refrigerator directly uploads the processed feature vectors to the processing equipment, avoiding the processing equipment from processing the running data again until it obtains the feature vectors, thus saving computing resources on the processing equipment side. Finally, the processing equipment performs fault type analysis and sends the results directly to the user's mobile device, forming a complete service loop. This realizes a shift from passive repair to proactive care, improving the product's service capabilities and competitiveness.

[0081] This application also provides an electronic device 400, please refer to... Figure 5 It includes a memory 401, a processor 402, and a computer program 4011 stored in the memory 401 and executable on the processor 402. When the processor 402 executes the computer program 4011, it implements the fault prediction method as described above.

[0082] The method includes the following steps: S101: Acquire the refrigerator's operating data, including acoustic data, vibration data, temperature data, and current data. S102: Perform anomaly calculation based on the operating data and a preset isolated forest model, and obtain anomaly scores. S103: Provide local anomaly warnings based on the anomaly scores and preset anomaly thresholds.

[0083] Alternatively, the method includes the following steps: S201: In response to the received feature vector, perform fault type analysis on the feature vector based on a preset fault diagnosis model. The feature vector is obtained by the refrigerator calculating anomalies based on operating data and a preset isolated forest model, and uploading it to the processing device when the anomaly score exceeds a preset anomaly threshold. The feature vector is obtained by the refrigerator preprocessing the collected operating data and extracting features from the processed operating data, which includes acoustic data, vibration data, temperature data, and current data. S202: Generate early warning information based on the analyzed fault type and send it to the user's mobile device.

[0084] This application embodiment also provides a storage medium that stores control instructions, which, when executed by a processor, implement the fault prediction method described above.

[0085] The method includes the following steps: S101: Acquire the refrigerator's operating data, including acoustic data, vibration data, temperature data, and current data. S102: Perform anomaly calculation based on the operating data and a preset isolated forest model, and obtain anomaly scores. S103: Provide local anomaly warnings based on the anomaly scores and preset anomaly thresholds.

[0086] Alternatively, the method includes the following steps: S201: In response to the received feature vector, perform fault type analysis on the feature vector based on a preset fault diagnosis model. The feature vector is obtained by the refrigerator calculating anomalies based on operating data and a preset isolated forest model, and uploading it to the processing device when the anomaly score exceeds a preset anomaly threshold. The feature vector is obtained by the refrigerator preprocessing the collected operating data and extracting features from the processed operating data, which includes acoustic data, vibration data, temperature data, and current data. S202: Generate early warning information based on the analyzed fault type and send it to the user's mobile device.

[0087] This application also provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations described in this application.

[0088] For example, a computer program can be divided into one or more modules / units, which are stored in memory and executed by a processor to perform the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in an electronic device.

[0089] Electronic devices can be desktop computers, laptops, handheld computers, and cloud servers, among other electronic devices. Electronic devices may include, but are not limited to, processors and memory. For example, electronic devices may also include input / output devices, network access devices, buses, etc.

[0090] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0091] In the embodiments provided by this invention, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. Multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0092] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units.

[0093] If integrated modules / units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program may include computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. Computer-readable media may include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0094] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0095] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more features.

[0096] The fault prediction method, control device, electronic device, and storage medium provided in the embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A fault prediction method, characterized in that, Applicable to refrigerators, the method includes: The refrigerator's operating data is acquired, including acoustic data, vibration data, temperature data, and current data. Anomaly calculations are performed based on the aforementioned operational data and a preset isolated forest model, and anomaly scores are obtained. Local anomaly warnings are issued based on the anomaly score and a preset anomaly threshold.

2. The fault prediction method according to claim 1, characterized in that, The anomaly calculation based on the operational data and the preset isolated forest model includes: The running data is preprocessed to obtain processed running data; Feature extraction is performed on the processed running data to obtain a feature vector; The feature vector is input into the preset isolated forest model for anomaly calculation.

3. The fault prediction method according to claim 2, characterized in that, The step of inputting the feature vector into the preset isolated forest model for anomaly calculation includes: The feature vector is traversed through each isolated tree in the preset isolated forest model; Calculate the path length required for the feature vector to be isolated on each isolated tree; Calculate the mean path length on each isolated tree; The outlier score is determined based on the mean.

4. The fault prediction method according to claim 2, characterized in that, Also includes: If the anomaly score is higher than the preset anomaly threshold, the feature vector is uploaded to the processing device for fault type analysis.

5. A fault prediction method, characterized in that, Suitable for processing equipment, the method includes: In response to the received feature vector, the feature vector is analyzed for fault type based on a preset fault diagnosis model; Based on the fault type obtained from the analysis, an early warning message is generated and sent to the user's mobile device; The feature vector is generated by the refrigerator performing anomaly calculations based on operating data and a preset isolated forest model, and then uploading the anomaly score to the processing device when the anomaly score exceeds a preset anomaly threshold. The operating data includes acoustic data, vibration data, temperature data, and current data.

6. The fault prediction method according to claim 5, characterized in that, The fault type analysis of the feature vector based on the preset fault diagnosis model includes: Based on the classification boundary or decision function of the preset fault diagnosis model, the feature vector is mapped to the preset fault category space to obtain the mapped feature information. Based on the mapping feature information, a fault probability distribution vector is output, where each vector in the fault probability distribution vector represents a fault type and its fault probability.

7. The fault prediction method according to claim 5, characterized in that, The preset fault diagnosis model is an integrated model of one or more of the following: support vector machine model, random forest or gradient boosting decision tree model, Naive Bayes classifier, artificial neural network model, convolutional neural network model or recurrent neural network model.

8. A fault prediction and control device, characterized in that, Suitable for refrigerators, the device includes: The operation data acquisition module is configured to acquire the operation data of the refrigerator, including acoustic data, vibration data, temperature data and current data; The analysis module is configured to perform anomaly calculations based on the running data and a preset isolated forest model, and obtain anomaly scores. The early warning module is configured to provide local early warnings based on the anomaly score and a preset anomaly threshold.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the fault prediction method as described in any one of claims 1-4, or implements the fault prediction method as described in any one of claims 5-7.

10. A storage medium, characterized in that, The storage medium stores control instructions, which, when executed by a processor, implement the fault prediction method as described in any one of claims 1-7.

Images