Fault prediction method and device for fresh air conditioning system

Abstract

The invention discloses a fault prediction method and device for a fresh air conditioning system. The method comprises the following steps that multi-source sensing parameters of the fresh air conditioning system are collected through a multi-source sensor; compressing data to form a standardized data packet; decompressing the data packet and extracting key features so as to combine a sensor topological graph, output a fault probability matrix and a space-time feature vector based on a space-time diagram convolutional network, and output an influence coefficient of external environment parameters on system performance based on an environment coupling model; jointly inputting the fault probability matrix and historical environment data into a Bayesian decision tree model so as to output confidence coefficients of various faults; moreover, inputting the spatio-temporal feature vector into a life prediction model to output residual life probability distribution; and carrying out visual interface display based on the confidence and residual life probability distribution of each type of fault, and dynamically adjusting an alarm threshold.

Description

Technical Field

[0001] The present invention relates to a fault prediction method for a fresh air air conditioning system and also to a corresponding fault prediction device, belonging to the technical field of air conditioning. Background Art

[0002] Current fresh air air conditioning systems generally rely on the traditional maintenance method of "periodic manual inspections," which lacks real-time, continuous performance monitoring. Without constant visibility into equipment status, maintenance personnel are forced to conduct on-site inspections at fixed intervals, making it difficult to identify potential problems in a timely manner.

[0003] In this mode, system alarms primarily rely on a "single-sensor threshold alarm" mechanism: a fixed threshold is set for the vibration or current sensor, and an alarm is triggered once the real-time value exceeds this threshold. This solution offers advantages in low hardware cost and simple implementation, but also significant disadvantages: environmental interference (such as pressure fluctuations caused by changes in damper opening) can easily lead to misjudgments, resulting in a high false alarm rate. Furthermore, it only issues an alarm when the value exceeds the threshold and cannot predict faults in advance.

[0004] In addition, the current "regular manual inspections" can only rely on the experience of operation and maintenance personnel to conduct visual or auditory inspections of equipment. Although it can detect obvious visible damage, it cannot identify early, hidden performance degradation, and it is difficult to quantitatively assess the health status of the equipment. On the other hand, some fresh air air conditioning systems have attempted to introduce "simple data analysis" methods, that is, statistics based on historical operating data to determine whether there are any abnormalities. However, such methods often only analyze a single data source in isolation and lack the ability to coordinate the processing of multiple sources such as vibration, current, and environment. This leads to one-sided analysis results and makes it difficult to fully and accurately reflect the true health level of the equipment.

[0005] Due to the compounding effects of these O&M difficulties, "post-event repairs and system downtime for component replacement" have become the norm. Bearings aren't detected until they're worn to the point of locking, filters are clogged until airflow is halved, and motors aren't shut down until they overheat and burn out from dust accumulation. Each sudden failure not only results in high emergency repair costs but also causes extended downtime for the fresh air conditioning system, directly impacting indoor air quality and user health. Therefore, shifting from "reactive repairs" to "proactive prevention"—using real-time, data-driven fault prediction technology to proactively identify and quantify the remaining lifespan of equipment at the first sign of performance degradation—has become an essential option for reducing O&M costs, shortening downtime, and ensuring continuous and healthy system operation. Summary of the Invention

[0006] The primary technical problem to be solved by the present invention is to provide a fault prediction method for a fresh air air conditioning system.

[0007] Another technical problem to be solved by the present invention is to provide a fault prediction device for a fresh air air conditioning system.

[0008] In order to achieve the above technical objectives, the present invention adopts the following technical solutions: According to a first aspect of an embodiment of the present invention, a fault prediction method for a fresh air air conditioning system is provided, comprising the following steps: The vibration signal, current signal and external environmental parameters of the fresh air air conditioning system are collected through multi-source sensors; Performing frequency domain sparsification on the vibration signal to extract vibration sparsity features; performing harmonic analysis on the current signal to extract current compression features; performing vector normalization on the external environmental parameters to extract an environmental state vector; Encapsulating, encrypting, and status-marking the vibration sparsity features, current compression features, and environmental state vectors to form a standardized data packet; The standardized data packet is decrypted and key features are extracted. Combined with the sensor topology, the working modes of the vibration and current signals in spatial distribution and time series are captured based on a spatiotemporal graph convolutional network, thereby outputting a fault probability matrix and spatiotemporal feature vectors. Furthermore, the impact of external environmental parameters on system performance is analyzed based on an environmental coupling model to output an impact coefficient. Inputting the failure probability matrix into a Bayesian decision tree model and performing risk assessment in combination with historical environmental data to output confidence levels for various types of failures; and inputting the spatiotemporal feature vector into a life prediction model to output a remaining life probability distribution; Based on the confidence level and remaining life probability distribution of the various types of faults, the system operating status, fault trends and maintenance plans are displayed through a three-dimensional graphical interface; and based on the impact coefficient, the alarm threshold of the fresh air air conditioning system is dynamically adjusted.

[0009] Preferably, the fault probability matrix and spatiotemporal feature vector are output based on the spatiotemporal graph convolutional network, specifically including: The vibration sparsity features, current compression features, and sensor topology map are input into a spatiotemporal graph convolutional network; wherein the sensor topology map is used to describe the physical position relationship, signal transmission path, and delay characteristics of each sensor in the system; The correlation between bearing vibration and impeller vibration is captured through spatial convolution, and the continuous upward trend of harmonic amplitude is analyzed through temporal convolution to output the probability of bearing wear, blade dust accumulation, filter clogging, and other fault probabilities. In addition, the sensor topology provides a physical correlation basis for the spatiotemporal graph convolution network, thereby outputting a spatiotemporal feature vector. The bearing wear probability, blade dust accumulation probability, filter blockage probability and other fault probabilities together form a fault probability matrix.

[0010] Preferably, the confidence levels of the various types of faults include: confidence level of bearing wear probability, confidence level of blade dust accumulation probability, confidence level of filter blockage probability, and confidence level of other fault probabilities.

[0011] Preferably, the life prediction model is constructed by the following steps: Collecting multi-source data and preprocessing it; wherein the multi-source data includes at least vibration signals, current waveforms and environmental parameters; Perform spatial graph convolution and temporal convolution on the preprocessed data to output spatiotemporal feature vectors; Inputting the spatiotemporal feature vector into an initial physical information neural network model to output remaining life distribution parameters; wherein the remaining life distribution parameters include at least: a controlled attenuation degree, a Weibull shape parameter, and a reference life; Inversely analyze physical parameters based on historical data and generate parameter labels to form multiple training samples, which together constitute the training set; The initial physical information neural network model is trained based on the training set to ultimately train and form the life prediction model.

[0012] Preferably, the alarm threshold of the fresh air air conditioning system is dynamically adjusted by the following formula: in, Indicates the alarm threshold at time t; T base Indicates the initial alarm threshold; r indicates the maximum adjustment range; Indicates a smooth transition; represents the influence coefficient at time t; k represents the sensitivity coefficient of the environmental change rate; s represents the normalization factor, which is a fixed parameter; Indicates the rate of environmental change.

[0013] Preferably, the fault prediction method further includes: generating a system maintenance instruction based on the remaining life probability distribution and the dynamically adjusted alarm threshold; The system maintenance instructions include at least no action, cleaning or replacement.

[0014] According to a second aspect of an embodiment of the present invention, a fault prediction device for a fresh air air conditioning system is provided, comprising: The perception layer is used to collect vibration signals, current signals, and external environmental parameters of the fresh air air conditioning system through multi-source sensors; and perform frequency domain sparsification on the vibration signals to extract vibration sparsity features; perform harmonic analysis on the current signals to extract current compression features; and perform vector normalization on the external environmental parameters to extract the environmental state vector; A transmission layer connected to the perception layer for encapsulating, encrypting, and status-marking the vibration sparsity features, current compression features, and environmental state vectors to form a standardized data packet; An analysis layer, connected to the transport layer, for decrypting the standardized data packets and extracting key features, and performing data analysis in combination with the sensor topology map, thereby outputting a fault probability matrix, a spatiotemporal feature vector, and an influence coefficient of external environmental parameters on system performance; A decision layer, connected to the analysis layer, makes decisions based on the fault probability matrix, spatiotemporal eigenvectors, and influence coefficients, and outputs confidence levels and remaining life probability distributions for various types of faults; The application layer is connected to the decision layer to display the system operation status, fault trends and maintenance plans through a three-dimensional graphical interface based on the confidence level of the various types of faults and the remaining life probability distribution; and the application layer is also connected to the analysis layer to dynamically adjust the alarm threshold of the fresh air air conditioning system based on the impact coefficient.

[0015] Preferably, the analysis layer includes: The fault analysis module is pre-configured with a spatiotemporal graph convolutional network to receive the key features and sensor topology, capture the working modes of the vibration signal and current signal in spatial distribution and time series, and output a fault probability matrix; The performance impact module is preset with an environmental coupling model to receive the key features and analyze the impact of external environmental parameters on system performance, thereby outputting an impact coefficient.

[0016] Preferably, the decision-making layer includes: A risk assessment module is pre-configured with a Bayesian decision tree model to receive the fault probability matrix and perform risk assessment in combination with historical environmental data, thereby outputting the confidence level of each type of fault; The life prediction module is preset with a life prediction model to receive the spatiotemporal feature vector and output the remaining life probability distribution through the life prediction model.

[0017] Preferably, the application layer includes: A visualization module is connected to the decision-making layer to display the system operation status, fault trends and maintenance plans through a three-dimensional graphical interface based on the confidence level of each type of fault and the probability distribution of remaining life; A threshold adjustment module is connected to the analysis layer to dynamically adjust the alarm threshold of the fresh air air conditioning system based on the influence coefficient.

[0018] Compared with the prior art, the present invention has the following technical effects: (1) Through data collection, feature extraction, state analysis, risk prediction and control feedback, a closed-loop management is formed to identify the performance degradation of the fresh air air conditioning system at an early stage and predict the remaining life of the fresh air air conditioning system, providing guidance for the maintenance of the fresh air air conditioning system and avoiding the blindness of system maintenance.

[0019] (2) Multi-source physical signals of the fresh air air conditioning system during operation are collected through multi-source sensors to perform fault analysis from multiple dimensions, thus achieving collaborative prediction of multiple types of faults (such as bearing wear, blade dust accumulation, filter blockage, etc.).

[0020] (3) The impact of external environmental parameters on system performance is analyzed through the environmental coupling model, and the alarm threshold of the fresh air air conditioning system is dynamically adjusted according to the influence coefficient to avoid system false alarms or missed alarms caused by environmental interference. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A flowchart of a fault prediction method for a fresh air air conditioning system provided by the first embodiment of the present invention; Figure 2 A structural diagram of a fault prediction device for a fresh air air conditioning system provided by a second embodiment of the present invention; Figure 3 This is a structural diagram of the perception layer in the second embodiment of the present invention; Figure 4 This is a structural diagram of the analysis layer in the second embodiment of the present invention; Figure 5 This is a structural diagram of the decision layer in the second embodiment of the present invention; Figure 6 This is a structural diagram of the application layer in the second embodiment of the present invention; Figure 7 This is a structural diagram of a fault prediction device for a fresh air air conditioning system provided in the third embodiment of the present invention. DETAILED DESCRIPTION

[0022] The technical content of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] The embodiment of the present invention provides a fault prediction method and system for fresh air air conditioning system. The whole system consists of five layers: perception layer, transmission layer, analysis layer, decision layer and application layer (such as Figure 2 , to jointly implement this fault prediction method. A closed-loop management system is formed between each layer through data collection, feature extraction, status analysis, risk prediction, and control feedback. This system is used to perform performance testing and fault prediction on the fresh air air conditioning system, thereby achieving intelligent monitoring of the equipment's operating status and providing support for operation and maintenance.

[0024] First embodiment like Figure 1 As shown, the first embodiment of the present invention provides a fault prediction method for a fresh air air conditioning system, which specifically includes the following steps: S10: Obtain multi-source perception parameters.

[0025] In this embodiment, multi-source physical signals of the fresh air air conditioning system during operation are collected by multi-source sensors and preliminarily processed to obtain multi-source perception parameters.

[0026] Specifically, the multi-source perception parameters include at least: (1) The vibration signals of the fresh air conditioning system are collected through a MEMS piezoelectric array, and the frequency domain sparseness of the vibration signals is performed to extract the vibration sparse features. This feature can be used to identify the early abnormal vibration characteristics of key equipment components (such as bearings and gears).

[0027] (2) The current signal of the fresh air air conditioning system is collected through an intelligent current sensor, and harmonic analysis is performed on the current signal to extract the current compression feature, which is used to detect the motor operating status, load changes and electrical fault precursors.

[0028] (3) Environmental parameters, including PM2.5 concentration, TVOC (total volatile organic compounds), temperature, humidity, etc., are collected in real time through environmental sensors. Vector normalization is performed on the external environmental parameters to extract the environmental state vector, thereby providing basic data for subsequent environmental coupling analysis.

[0029] It is understandable that the vibration sparsity feature, the current compression feature and the environmental state vector together constitute the multi-source sensing parameter. In other embodiments, other types of feature parameters may be added as needed.

[0030] S20: Data is compressed into a standardized data package and uploaded to the cloud or local server.

[0031] After acquiring the multi-source sensing parameters in step S10, they are encrypted, compressed, and remotely transmitted via the edge gateway. The LoRaWAN / 5G edge gateway supports both LoRaWAN and high-speed 5G dual-mode communications, boasting over 80% data compression capabilities. This ensures efficient and stable data upload to the cloud or local server even under limited bandwidth conditions.

[0032] S30: Decompress the data packet and extract key features for fault analysis and performance impact analysis.

[0033] After uploading the data packet to the cloud or local server in step S20, the data packet is downloaded from the cloud or local server and decompressed to extract key features. It is understood that the key features correspond to the multi-source perception parameters mentioned above.

[0034] Specifically, the data analysis process includes the following two parts: S31: Fault analysis.

[0035] After extracting key features, the sensor topology map is combined with a spatiotemporal graph convolutional network to capture the spatial distribution and temporal sequence of vibration and current signals, thereby outputting a fault probability matrix and spatiotemporal feature vector. The sensor topology map describes the physical location relationships, signal transmission paths, and delay characteristics of each sensor in the system.

[0036] Specifically, the vibration sparse features, current compression features and sensor topology map are first input into the spatiotemporal graph convolutional network.

[0037] Then, spatial convolution is used to capture the correlation between bearing vibration and impeller vibration, and temporal convolution is used to analyze the rising trend of harmonic amplitudes to output the probability of bearing wear, blade dust accumulation, filter clogging, and other fault probabilities. These probabilities together form a fault probability matrix.

[0038] Finally, the sensor topology map provides a physical association basis for the spatiotemporal graph convolutional network, thereby outputting a spatiotemporal feature vector.

[0039] In this embodiment, the spatiotemporal feature vector (128-dimensional floating-point array) is a fused feature representation extracted from multimodal data by the spatiotemporal graph convolutional network (STGCN), and includes the following core dimensions: The prediction of lifespan using spatiotemporal eigenvectors is achieved through a physical information neural network (PINN) and is completed in the following four steps: Step 1: Feature mapping to physical parameters. Input: spatiotemporal feature vector, Output: performance degradation coefficient (h -1 )α, Weibull distribution shape parameter k, benchmark characteristic life (h)λ.

[0040] Step 2: Solve the performance degradation equation.

[0041] Step 3: Generate the remaining life distribution.

[0042] Step 4: Failure probability prediction.

[0043] For example, the input feature vector is as follows: Step 1: Calculation of physical parameters: ; Step 2: Vibration Correction: ; Step 3: Lifetime distribution: ; Step 4: Failure Probability: .

[0044] S32: Performance impact analysis.

[0045] In this embodiment, an environmental coupling model is pre-built. After receiving the key feature, the environmental state vector is input into the environmental coupling model to analyze the impact of external environmental parameters on system performance and output an impact coefficient.

[0046] Among them, a training set is formed based on historical environmental parameters and system performance parameters, and the training set is used to train the model to form the environmental coupling model, so as to analyze the impact of external environmental parameters on system performance.

[0047] It can be understood that the output result after the fault analysis in step S31 is the null eigenvector of the fault probability matrix, and the output result after the performance impact analysis in step S32 is the impact coefficient.

[0048] S40: Conduct risk assessment and lifespan prediction based on the results of data analysis.

[0049] Specifically, step S40 includes the following two parts: risk assessment (corresponding to S41) and life prediction (corresponding to S42).

[0050] S41: Risk assessment.

[0051] In this embodiment, the fault probability matrix is ​​input into the Bayesian decision tree model, and risk assessment is performed in combination with historical environmental data to output the confidence level of each type of fault.

[0052] Specifically, in this embodiment, the Bayesian decision tree model comprises: Model architecture: [Failure probability matrix] + [Historical environmental data] --> [Bayesian network layer] --> [Decision tree classifier] --> [Confidence output] --> [Risk assessment].

[0053] The core components are shown in the following table: Risk assessment process input data: failure probability matrix O, current environment vector , Historical Environmental Database , Failure loss cost table .

[0054] Risk assessment process: environmental similarity matching (output similar environment set) → prior probability calculation → likelihood function modeling → Bayesian posterior calculation (output normalized posterior probability vector) → decision tree confidence grading → risk value assessment → confidence output.

[0055] The output confidence information includes: fault type, confidence, risk level, risk value and corresponding decision.

[0056] The confidence levels of various fault types include: bearing wear probability confidence level, blade dust accumulation probability confidence level, filter blockage probability confidence level, and other fault probability confidence levels; and the confidence level ranges from 0 to 1.

[0057] S42: Lifespan Prediction In this embodiment, the spatiotemporal feature vector is input into the life prediction model to output the remaining life probability distribution.

[0058] The life prediction model is constructed through the following steps: Step 1: Multi-source data collection and preprocessing.

[0059] Data sources: vibration signal, current waveform, environmental parameters; Preprocessing process: [Raw data] --> [Signal alignment] --> [Frequency domain analysis] --> [Physical normalization] Step 2: Spatiotemporal feature vector generation.

[0060] Input: preprocessed multidimensional feature matrix, Output: Spatiotemporal eigenvectors (example output: [0.32, -0.15, 0.08, ..., 0.21] bearing coupling ↑32%, THD change rate ↓15%, high-frequency energy ↑8%).

[0061] Step 3: Physical constraint life model construction.

[0062] Model Architecture: Physical Information Neural Network (PINN) Input: spatiotemporal feature vector Output: Remaining life distribution parameters (Weibull distribution k, λ) Core constraint physics equations: (Air volume attenuation equation) Feature → Physical Parameter Mapping: α: Controls the decay rate (positively correlated with vibration energy) k: Weibull shape parameter (k < 1 indicates early failure) : Reference life (determined by material properties) Step 4: Model training Inverse analysis of physical parameters based on historical data , to generate labels, form multiple training samples, to together constitute a training set, which is used for model training, and finally train to form the life prediction model.

[0063] S50: Visual interface display and dynamic adjustment of alarm thresholds.

[0064] It is understandable that after obtaining the confidence level and remaining life probability distribution of various types of faults through step S40, data processing can be combined with relevant modules to display the system operation status, fault trends and maintenance plans through a three-dimensional graphical interface.

[0065] Furthermore, based on the influence coefficient output in step S30, the alarm threshold of the fresh air air conditioning system is dynamically adjusted by the following formula: in, Indicates the alarm threshold at time t; T base Indicates the initial alarm threshold; r indicates the maximum adjustment range, ranging from 0 to 0.2, and the system default is 0.05; Indicates a smooth transition; represents the influence coefficient at time t; k represents the sensitivity coefficient of the environmental change rate; s represents the normalization factor, which is a fixed parameter; Indicates the rate of environmental change.

[0066] In addition, in this embodiment, a system maintenance instruction is generated based on the remaining life probability distribution and the dynamically adjusted alarm threshold. The system maintenance instruction includes at least no action, cleaning, or replacement, thereby providing maintenance guidance to system maintenance personnel.

[0067] Second embodiment like Figure 2 As shown, based on the above-mentioned first embodiment, the second embodiment of the present invention provides a fault prediction device for a fresh air air conditioning system, including a perception layer 1, a transmission layer 2, an analysis layer 3, a decision layer 4 and an application layer 5.

[0068] Specifically, the perception layer 1 is responsible for collecting multi-source physical signals during the operation of the equipment and performing preliminary processing. Figure 3As shown, the perception layer 1 includes the following three sensor modules: a MEMS piezoelectric array 11, which outputs sparse vibration spectrum data for identifying early abnormal vibration characteristics of key equipment components (such as bearings and gears). An intelligent current sensor 12 extracts compressed harmonic data to detect motor operating status, load changes, and electrical fault precursors. Multi-source environmental sensors 13 collect real-time environmental parameters, including PM2.5 concentration, total volatile organic compounds (TVOC), temperature, and humidity, providing basic data for subsequent environmental coupling analysis. This information is aggregated and outputted by the perception layer, which then passes to the next level.

[0069] The transport layer 2 receives data from the perception layer 1 and encrypts, compresses, and transmits it remotely via an edge gateway. Once the standardized data packet is transmitted, the data enters the analysis layer 3 for modeling and feature extraction.

[0070] The analysis layer 3 performs deep learning modeling and status analysis on the uploaded data. Figure 4 As shown, the analysis layer 3 includes a fault analysis module 31 and a performance impact module 32. The fault analysis module 31 is pre-configured with a spatiotemporal graph convolutional network (SGN), a deep neural network based on a graph structure. It is used to capture the complex patterns of spatial distribution and time series of signals such as equipment vibration and current, output a fault probability matrix, and assess the health status of different components. The performance impact module 32 is pre-configured with an environmental coupling model. It dynamically adjusts model parameters based on environmental data transmitted from the perception layer and outputs an impact coefficient that reflects the impact of environmental changes on equipment performance, thereby improving prediction accuracy. The analysis results are transmitted as input to the decision-making layer for further risk assessment and lifespan prediction.

[0071] The decision layer 4 makes a comprehensive judgment based on the results of the analysis layer 3 and generates specific operation and maintenance strategies, such as Figure 5 As shown, the decision layer 4 includes a risk assessment module 41 and a life prediction module 42. Specifically, the risk assessment module 41 is pre-configured with a Bayesian decision tree model. Based on the Bayesian inference algorithm, it combines historical data with the current status to calculate the confidence level of each type of fault occurrence and assist in formulating a priority response strategy. The life prediction module 42 is pre-configured with a life prediction model (physical information neural network). This neural network, constrained by physical laws, predicts the remaining useful life of the equipment and outputs it as a probability density function to guide the scheduling of preventive maintenance plans. The final decision result is sent to the application layer 5 for specific operations.

[0072] The application layer 5 is the interface for the system to interact with the user or the control system, and mainly includes a visualization module 51 and a threshold adjustment module 52 (e.g. Figure 6Specifically, the visualization module 51 is connected to the decision layer 4 to display the system's operating status, fault trends, and maintenance plans through a three-dimensional graphical interface based on the confidence level and remaining life probability distribution of various fault types. The threshold adjustment module 52 is connected to the analysis layer 3 to automatically adjust the alarm threshold based on the impact coefficient and real-time operating conditions and environmental changes, avoiding false alarms and missed alarms, thereby improving the system's stability and adaptability.

[0073] It can be understood that the functions and connection relationships of the various module units in this embodiment are only a specific implementation method for realizing the fault prediction method in the above-mentioned first embodiment. In other embodiments, the functions and connection relationships of the various module units can also be adaptively adjusted as needed to realize the above-mentioned fault prediction method.

[0074] Third embodiment like Figure 7 As shown, based on the above-mentioned fault prediction method for a fresh air air conditioning system, a third embodiment of the present invention further provides a fault prediction device for a fresh air air conditioning system. The fault prediction device includes one or more processors and a memory. The memory is coupled to the processor and is configured to store one or more programs. When the programs are executed by the processor, the processor implements the fault prediction method for a fresh air air conditioning system described in the above-mentioned embodiment.

[0075] The processor is used to control the overall operation of the fault prediction device to complete all or part of the steps of the above-mentioned fault prediction method for fresh air air conditioning systems. The processor can be a central processing unit (CPU), a graphics processing unit (GPU), a field programmable gate array (FPGA), an application-specific integrated circuit (ASIC), a digital signal processing (DSP) chip, etc. The memory is used to store various types of data to support the operation of the fault prediction device. This data may include, for example, instructions for any application or method operating on the fault prediction device, as well as application-related data. The memory can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, etc.

[0076] In an exemplary embodiment, the fault prediction device can be implemented by a computer chip or entity, or by a product with a certain function, to execute the above-mentioned fault prediction method for fresh air air conditioning systems and achieve the same technical effect as the above-mentioned method. A typical embodiment is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, an in-vehicle human-computer interaction device, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0077] In another exemplary embodiment, the present invention further provides a computer-readable storage medium comprising program instructions, which, when executed by a processor, implement the steps of the fault prediction method for a fresh air air conditioning system described in any of the aforementioned embodiments. For example, the computer-readable storage medium may be the aforementioned memory comprising the program instructions, which may be executed by a processor of a fault prediction device to perform the aforementioned fault prediction method for a fresh air air conditioning system and achieve the same technical effects as the aforementioned method.

[0078] It should be noted that the above embodiments are merely examples, and the technical solutions of the various embodiments can be combined and are all within the protection scope of the present invention.

[0079] Furthermore, 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 the technical features being referred to. Thus, a feature identified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0080] The above describes in detail the fault prediction method and apparatus for a fresh air air conditioning system provided by the present invention. For those skilled in the art, any obvious modification thereof without departing from the essence of the present invention would constitute an infringement of the present invention's patent rights and would incur corresponding legal liability.

Claims

1. A fault prediction method for fresh air air conditioning system, characterized in that The steps include: The vibration signal, current signal and external environmental parameters of the fresh air air conditioning system are collected through multi-source sensors; Performing frequency domain sparsification on the vibration signal to extract vibration sparse features; performing harmonic analysis on the current signal to extract current compression features; performing vector normalization on the external environmental parameters to extract an environmental state vector; Encapsulating, encrypting, and status-marking the vibration sparsity features, current compression features, and environmental state vectors to form a standardized data packet; The standardized data packet is decrypted and key features are extracted. Combined with the sensor topology, the working modes of the vibration and current signals in spatial distribution and time series are captured based on a spatiotemporal graph convolutional network, thereby outputting a fault probability matrix and spatiotemporal feature vectors. Furthermore, the impact of external environmental parameters on system performance is analyzed based on an environmental coupling model to output an impact coefficient. Inputting the failure probability matrix into a Bayesian decision tree model and performing risk assessment in combination with historical environmental data to output confidence levels for various types of failures; and inputting the spatiotemporal feature vector into a life prediction model to output a remaining life probability distribution; Based on the confidence level and remaining life probability distribution of the various types of faults, the system operating status, fault trends and maintenance plans are displayed through a three-dimensional graphical interface; and based on the impact coefficient, the alarm threshold of the fresh air air conditioning system is dynamically adjusted.

2. The fault prediction method according to claim 1, characterized in that Based on the spatiotemporal graph convolutional network, the fault probability matrix and spatiotemporal feature vector are output, including: The vibration sparsity features, current compression features, and sensor topology map are input into a spatiotemporal graph convolutional network; wherein the sensor topology map is used to describe the physical position relationship, signal transmission path, and delay characteristics of each sensor in the system; The correlation between bearing vibration and impeller vibration is captured through spatial convolution, and the continuous upward trend of harmonic amplitude is analyzed through temporal convolution to output the probability of bearing wear, blade dust accumulation, filter clogging, and other fault probabilities. In addition, the sensor topology provides a physical correlation basis for the spatiotemporal graph convolution network, thereby outputting a spatiotemporal feature vector. The bearing wear probability, blade dust accumulation probability, filter blockage probability and other fault probabilities together form a fault probability matrix.

3. The fault prediction method according to claim 2, wherein: The confidence levels of various types of faults include: bearing wear probability confidence level, blade dust accumulation probability confidence level, filter blockage probability confidence level, and other fault probability confidence levels.

4. The fault prediction method according to claim 2, characterized in that The life prediction model is constructed by the following steps: Collecting multi-source data and preprocessing it; wherein the multi-source data includes at least vibration signals, current waveforms and environmental parameters; Perform spatial graph convolution and temporal convolution on the preprocessed data to output spatiotemporal feature vectors; Inputting the spatiotemporal feature vector into an initial physical information neural network model to output remaining life distribution parameters; wherein the remaining life distribution parameters include at least: a controlled attenuation degree, a Weibull shape parameter, and a reference life; Inversely analyze physical parameters based on historical data and generate parameter labels to form multiple training samples, which together constitute the training set; The initial physical information neural network model is trained based on the training set to ultimately train and form the life prediction model.

5. The fault prediction method according to claim 1, characterized in that The alarm threshold of the fresh air air conditioning system is dynamically adjusted by the following formula: in, Indicates the alarm threshold at time t; T base Indicates the initial alarm threshold; r indicates the maximum adjustment range; Indicates a smooth transition; represents the influence coefficient at time t; k represents the sensitivity coefficient of the environmental change rate; s represents the normalization factor, which is a fixed parameter; Indicates the rate of environmental change.

6. The fault prediction method according to claim 1, characterized in that Also includes: generating a system maintenance instruction based on the remaining life probability distribution and the dynamically adjusted alarm threshold; The system maintenance instructions include at least no action, cleaning or replacement.

7. A fault prediction device for fresh air air conditioning system, characterized in that include: The perception layer is used to collect vibration signals, current signals, and external environmental parameters of the fresh air air conditioning system through multi-source sensors; and perform frequency domain sparsification on the vibration signals to extract vibration sparse features; performing harmonic analysis on the current signal to extract current compression features; performing vector normalization on the external environmental parameters to extract an environmental state vector; A transmission layer connected to the perception layer for encapsulating, encrypting, and status-marking the vibration sparsity features, current compression features, and environmental state vectors to form a standardized data packet; An analysis layer, connected to the transport layer, for decrypting the standardized data packets and extracting key features, and performing data analysis in combination with the sensor topology map, thereby outputting a fault probability matrix, a spatiotemporal feature vector, and an influence coefficient of external environmental parameters on system performance; A decision layer, connected to the analysis layer, makes decisions based on the fault probability matrix, spatiotemporal eigenvectors, and influence coefficients, and outputs confidence levels and remaining life probability distributions for various types of faults; The application layer is connected to the decision layer to display the system operation status, fault trends and maintenance plans through a three-dimensional graphical interface based on the confidence level of the various types of faults and the remaining life probability distribution; and the application layer is also connected to the analysis layer to dynamically adjust the alarm threshold of the fresh air air conditioning system based on the impact coefficient.

8. The fault prediction device according to claim 7, characterized in that The analysis layer includes: The fault analysis module is pre-configured with a spatiotemporal graph convolutional network to receive the key features and sensor topology, capture the working modes of the vibration signal and current signal in spatial distribution and time series, and output a fault probability matrix; The performance impact module is preset with an environmental coupling model to receive the key features and analyze the impact of external environmental parameters on system performance, thereby outputting an impact coefficient.

9. The fault prediction device according to claim 8, characterized in that The decision-making layer includes: A risk assessment module is pre-configured with a Bayesian decision tree model to receive the fault probability matrix and perform risk assessment in combination with historical environmental data, thereby outputting the confidence level of each type of fault; The life prediction module is preset with a life prediction model to receive the spatiotemporal feature vector and output the remaining life probability distribution through the life prediction model.

10. The fault prediction device according to claim 9, characterized in that The application layer includes: A visualization module is connected to the decision-making layer to display the system operation status, fault trends and maintenance plans through a three-dimensional graphical interface based on the confidence level of each type of fault and the probability distribution of remaining life; A threshold adjustment module is connected to the analysis layer to dynamically adjust the alarm threshold of the fresh air air conditioning system based on the influence coefficient.

Images