Equipment fault probability prediction method and device based on time sequence state data of Internet of Things

Abstract

The invention is suitable for the technical field of intelligent transportation, and provides an equipment fault probability prediction method and device based on time sequence state data of the Internet of Things. In the embodiment, for electromechanical equipment needing to be monitored on an expressway, discrete state data obtained by statistical analysis of discrete state data such as equipment switch and communication state and continuous physical quantity data obtained by time sequence analysis of continuous physical quantities such as equipment temperature, voltage and current are firstly collected, and then feature extraction is carried out on the discrete state data; the method comprises the steps of generating a unified preliminary feature vector, then extracting a fault risk feature from the preliminary feature vector, finally analyzing the risk feature through a causal attention mechanism, and outputting fault early warning information. The problems that existing high-speed electromechanical equipment is lagged in fault discovery, high in maintenance cost and low in fault response speed are solved. Real-time analysis and accurate prediction of the fault probability of the key equipment of the expressway are realized, so that the safety and efficiency of expressway operation are improved.

Description

Technical Field

[0001] This application relates to the field of intelligent transportation technology, and in particular to a method and apparatus for predicting the probability of equipment failure based on Internet of Things time-series status data. Background Technology

[0002] In today's rapidly developing transportation system, highways serve as the main arteries of the national economy, and their safe and efficient operation is directly related to the stability and development of the social economy. With the continuous expansion of the highway network and the continuous improvement of its intelligent level, highway electromechanical equipment, including but not limited to information boards, cameras, and environmental monitoring sensors, has become an indispensable infrastructure for ensuring road safety and improving traffic efficiency.

[0003] However, with the increasing variety of equipment and the complexity of systems, traditional manual periodic inspections and reactive maintenance are no longer sufficient to meet the stringent requirements of modern traffic management for efficiency, accuracy, and foresight. Traditional operation and maintenance methods suffer from slow response times, delayed fault detection, and high maintenance costs; minor faults can easily escalate into serious accidents, threatening traffic safety. Therefore, how to achieve real-time monitoring, fault probability prediction, and intelligent operation and maintenance of electromechanical equipment has become a critical issue that urgently needs to be addressed in the field of traffic equipment operation and maintenance.

[0004] Meanwhile, the rapid development of emerging technologies such as the Internet of Things (IoT), big data, cloud computing, and artificial intelligence has provided unprecedented opportunities for the intelligent operation and maintenance of highway electromechanical equipment. However, in the process of equipment use, there are currently problems such as delayed fault detection and high maintenance costs. By deploying a large number of sensors and network devices, it is possible to achieve comprehensive and all-weather monitoring of equipment operation status and collect massive amounts of real-time IoT status data. However, how to extract valuable information from complex, multi-source, heterogeneous IoT time-series status data, identify potential fault risks, and achieve accurate prediction is a major challenge. The heterogeneity of data, time-series dependencies, and the complexity of fault modes all place extremely high demands on data processing and analysis capabilities. Summary of the Invention

[0005] In view of this, this application provides a method and apparatus for predicting the probability of equipment failure based on IoT time-series status data, so as to realize real-time analysis and accurate prediction of the failure probability of key equipment on highways, thereby improving the safety and efficiency of highway operation.

[0006] The first aspect of this application provides a method for predicting the probability of device failure based on Internet of Things (IoT) time-series status data, applied to an intelligent transportation platform, the method comprising: The discrete state data and continuous physical quantity data of the target electromechanical equipment are acquired, and a continuous statistical time series feature vector is determined through the discrete state data, and a continuous analytical time series feature vector is determined through the continuous physical quantity data. The preliminary feature vector of the target electromechanical equipment is determined by the statistical time series feature vector and the analytical time series feature vector. Then, the preliminary feature vector is extracted based on the spatiotemporal dependence to obtain the core state feature vector containing fault risk characteristics. The core state feature vector is analyzed by a causal attention mechanism to obtain a risk analysis vector. The risk analysis vector is then input into a classifier network to obtain the predicted fault type of the target electromechanical equipment, and a warning message is generated based on the fault type.

[0007] Optionally, after obtaining the risk resolution vector, the method further includes: The risk analysis vector is input into a regression network to predict the shape parameters of the Weibull distribution. and scale parameters ; Through the shape parameters and scale parameters Generating the failure probability time function Then through the fault probability time function The warning information is generated based on the fault type, where T is the current time. The remaining running time from the start until the probability of failure exceeds the set threshold.

[0008] Optionally, after obtaining the risk resolution vector, the method further includes: The risk analysis vector is input into a regression network to predict the shape parameters of the Weibull distribution. and scale parameters ; Through the shape parameters and scale parameters Determine the estimated remaining lifespan of the target electromechanical equipment. The calculation formula is: Then, warning information is generated based on the estimated remaining lifespan and the fault type, wherein, This is the Gamma function.

[0009] Optionally, generate a time function for the failure probability. Subsequently, the method further includes: Through the shape parameters and scale parameters Determine the estimated remaining lifespan of the target electromechanical equipment. The calculation formula is: Then, warning information is generated using the failure probability time function, the estimated remaining lifespan, and the failure type, wherein, This is the Gamma function.

[0010] Optionally, determining the continuous statistical time-series feature vector using the discrete state data includes: The discrete state data is analyzed by a sliding window, and the frequency and persistence characteristics of abnormal events of each state type are calculated in real time. The number of abnormalities in the sliding window of each state type in the past N time units is calculated, and the immediate impact and duration of each abnormality are determined. The number of abnormalities, the immediate impact and duration of each abnormality are then determined as the statistical time series feature vector.

[0011] Optionally, generating early warning information using the failure probability-time function, estimated remaining lifetime, and failure type includes: The target fault type with the highest probability and its corresponding confidence level are determined from the fault types, and the core state feature vector is associated with a predefined high-order fault combination pattern dictionary to generate a textual description representing the fault mode. The text description, the failure probability time function, the estimated remaining lifespan, the target failure type, and the corresponding confidence level are combined into a warning message and sent.

[0012] A second aspect of this application provides a device for predicting device failure probability based on Internet of Things (IoT) time-series status data, applied to an intelligent transportation platform, the device comprising: The data acquisition unit is used to acquire discrete state data and continuous physical quantity data of the target electromechanical equipment, and to determine continuous statistical time series feature vectors through the discrete state data and continuous analytical time series feature vectors through the continuous physical quantity data. The feature extraction unit is used to determine the preliminary feature vector of the target electromechanical equipment through the statistical time series feature vector and the analytical time series feature vector, and then extract the preliminary feature vector based on the spatiotemporal dependence to obtain the core state feature vector containing fault risk features. The early warning information generation unit is used to parse the core state feature vector through a causal attention mechanism to obtain a risk analysis vector, and then input the risk analysis vector into a classifier network to obtain the predicted fault type of the target electromechanical equipment, and generate early warning information based on the fault type.

[0013] Optionally, after the early warning information generation unit obtains the risk analysis vector, the device further includes: The second early warning information generation unit is used to input the risk analysis vector into the regression network to predict the shape parameters of the Weibull distribution. and scale parameters ; Through the shape parameters and scale parameters Generating the failure probability time function Then through the fault probability time function The warning information is generated based on the fault type, where T is the current time. The remaining running time from the start until the probability of failure exceeds the set threshold.

[0014] Optionally, after the early warning information generation unit obtains the risk analysis vector, the device further includes: The third early warning information generation unit is used to input the risk analysis vector into the regression network to predict the shape parameters of the Weibull distribution. and scale parameters ; Through the shape parameters and scale parameters Determine the estimated remaining lifespan of the target electromechanical equipment. The calculation formula is: Then, warning information is generated based on the estimated remaining lifespan and the fault type, wherein, This is the Gamma function.

[0015] Optionally, the second early warning information generation unit generates a fault probability time function. Subsequently, the device further includes: The fourth early warning information generation unit is used to generate early warning information through the shape parameters. and scale parameters Determine the estimated remaining lifespan of the target electromechanical equipment. The calculation formula is: Then, warning information is generated using the failure probability time function, the estimated remaining lifespan, and the failure type, wherein, This is the Gamma function.

[0016] In the embodiments provided in this application, for electromechanical equipment that needs to be monitored on highways, discrete state data such as equipment switching and communication status are statistically analyzed, and continuous physical quantity data such as equipment temperature, voltage, and current are time-series analyzed. Then, features are extracted to generate a unified preliminary feature vector. Next, fault risk features are extracted from this preliminary feature vector. Finally, a causal attention mechanism is used to analyze these risk features and output fault warning information. This solves the problems of delayed fault detection, high maintenance costs, and slow fault response speed in existing highway electromechanical equipment. It achieves real-time analysis and accurate prediction of the fault probability of key highway equipment, thereby improving the safety and efficiency of highway operation. Attached Figure Description

[0017] Figure 1 A flowchart illustrating the method provided in this application embodiment; Figure 2 This is a structural diagram of the device provided in the embodiments of this application; Figure 3 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0018] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0019] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0020] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0021] This application provides a method and apparatus for predicting the probability of equipment failure based on IoT time-series status data, so as to realize real-time monitoring and accurate fault warning of the status of key equipment on highways.

[0022] The technical solutions of this application will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0023] like Figure 1 The diagram shown is a flowchart of a device failure probability prediction method based on IoT time-series state data provided in this application. The process may include the following steps: Step S101: Obtain discrete state data and continuous physical quantity data of the target electromechanical equipment, and determine a continuous statistical time series feature vector through the discrete state data, and determine a continuous analytical time series feature vector through the continuous physical quantity data.

[0024] In this embodiment, discrete state data includes discontinuous data such as device on / off status and communication status. For example, device on / off status can be on / off, running / stopping, powered on / power off; communication status can be connection successful / connection failed, online / offline, normal / faulty, etc. These data do not have any transition or intermediate values ​​between two different values. Because there are no intermediate values, the change in discrete state data is not gradual but abrupt. Continuous physical quantity data can be continuous physical quantity data such as device temperature, voltage, and current, and these data change gradually within their effective numerical range. This application uses different methods to process these two types of data.

[0025] For discrete state data, the processing method is as follows: 1. Receive discrete status data from the real-time status monitoring interface of highway electromechanical equipment. ; 2. Initiate sliding window analysis to calculate the frequency and persistence characteristics of abnormal events for each state type in real time. The specific process is as follows: 2.1 Calculate the sliding window anomaly count for each state type over the past N time units. , used to characterize the density of short-term anomalous outbreaks; 2.2 Calculate the time interval since the most recent anomaly occurred, i.e., the time since the most recent anomaly occurred. , used to characterize the immediate impact of anomalies; 2.3 Calculate the duration of the current abnormal state, i.e., the duration of the abnormality. It is used to distinguish between transient noise and structural anomalies.

[0026] Repeat steps 2.1 through 2.3 to determine the anomaly count. Recent anomaly occurrence time Abnormal duration This is the statistical time-series feature vector determined by the aforementioned discrete-state data, denoted as... .

[0027] For continuous physical quantity data, the processing method is as follows: 1. Receive continuous physical quantity data from the acquisition interface of highway electromechanical equipment. ; 2. Initiate time-frequency domain analysis to extract trend, fluctuation, and residual characteristics from continuous physical quantity data. The specific process is as follows: 2.1 Using an exponential smoothing kernel, the long-term trend characteristics of physical quantity data are extracted. This is to adapt to the dynamic changes in equipment status over time. The calculation process is as follows: ,in, This is the raw data at the current moment. This is the smoothed trend value from the previous time step. The smoothing coefficient is determined based on the response speed requirements of trend following. The larger the value, the faster the response to new data.

[0028] 2.2 Calculate the short-term fluctuation characteristics of physical quantity data This refers to the local variance within a fixed time window W, used to characterize the local stability of the signal. The calculation process is as follows: Where W is the window size, For window A moving average within a certain range is used to calculate the fluctuation around a local mean. This method is suitable for measuring the degree of local jitter in non-steady-state signals, and its window size... The settings are based on the duration of the minimum abnormal event.

[0029] 2.3 Discrete wavelet transform (DWT) is used to extract the frequency domain energy distribution characteristics of physical quantity data. It is used to capture periodic or transient anomalous patterns. The calculation process is as follows: ,in, For the first Energy of detail coefficients in wavelet decomposition (capturing high-frequency transient anomalies). The energy of the approximate coefficients (capturing low-frequency, slowly varying trends). This method is suitable for analyzing instantaneous impacts and local characteristics in non-stationary equipment signals. The selection of its decomposition series and basis functions is determined based on the typical fault frequency characteristics of equipment vibration or electronic signals.

[0030] Repeat steps 2.1 to 2.3 above to determine the long-term trend characteristics. Short-term fluctuation characteristics and frequency domain energy distribution characteristics This is the analytical time series characteristic vector determined by the aforementioned continuous physical quantity data, denoted as... .

[0031] Step S102: Determine the preliminary feature vector of the target electromechanical equipment through the statistical time series feature vector and the analytical time series feature vector, and then extract the preliminary feature vector based on the spatiotemporal dependence to obtain the core state feature vector containing fault risk features.

[0032] In this embodiment, the statistical time series feature vector determined in step S101 is first... and analysis of time series feature vectors The features are concatenated along the feature dimension. For example, they are first aligned in terms of data structure, and then directly connected along the feature dimension using the np.concatenate(axis=0) or np.hstack() functions to form a composite feature vector that integrates statistical and deep analysis characteristics. This feature vector is the preliminary feature vector of the aforementioned electromechanical equipment. .

[0033] After determining the initial feature vector, the spatiotemporal dependencies of this feature vector are further explored, and fault risk features with high-order interpretability are extracted. The specific steps are as follows: 1. Start the one-dimensional convolutional neural network It captures local patterns and transient correlations of feature vectors within a short time window. The calculation formula is: Where ReLU() is the linear rectified function, Let be the local feature vector at the current time t. These are the convolution kernel weights. The size of the convolution kernel window (representing the time span of the local pattern). This is a one-dimensional convolution operation. It is the convolution bias vector. This step achieves compression and abstraction of the local correlations of multimodal features.

[0034] 2. Determine the local feature vector sequence using the aforementioned local feature vectors. The gated recurrent unit (GRU) is activated to model long-term temporal dependencies and state evolution. The calculation formula is as follows: Where GRU() is the gated loop unit function, The hidden state at the current moment contains the state from the beginning of the sequence to... The global temporal dependency information at each moment forms a global temporal feature sequence. .

[0035] 3. Activate the Long Short-Term Memory (LSTM) network to predict the future. Feature sequences at each time step . .in, This step explicitly reflects the potential evolutionary trends of the device in the feature space.

[0036] 4. In the current feature and future characteristics The extended sequence formed The self-attention mechanism is applied to generate a draft of fused risk features. The formula is: , Among them, attention weights Measured the target time Features For any source time in the sequence Features Attention level, query characteristics It is the current moment. Input feature vector, key features Any source time The input feature vector, It is used to input features Projected to query space The trainable weight matrix, It is used to extract input features Projected to key space Trainable weight matrix, scaling factor This is used to scale the dot product result to prevent the gradient from vanishing when the input value is too large in the Softmax function. It is this historical moment. It is the time span of future predictions. It is the current moment. A contextual feature vector that integrates information from the entire sequence. It is used to extract input features Projected onto value space The trainable weight matrix.

[0037] 5. Initial draft of the above risk characteristics Applying L1 sparsity constraints, core state feature vectors are generated. The formula is: .in, It is the total loss predicted by the model. Is it the present or a historical moment? Predicted loss It is about the future Predicted loss at any given time These are the sparsity constraint coefficients. It is an eigenvector. This constraint ensures... It exhibits high sparsity, where the activation dimension represents a high-order fault combination pattern that has been deeply fused and predicted for the future.

[0038] By following the steps above, a core state feature vector containing fault risk characteristics can be obtained. Furthermore, because this feature vector is interpretable, its prediction logic and basis can be clearly understood and explained by the user.

[0039] Step S103: The core state feature vector is parsed through a causal attention mechanism to obtain a risk parsing vector. The risk parsing vector is then input into a classifier network to obtain the predicted fault type of the target electromechanical equipment, and a warning message is generated based on the fault type.

[0040] This embodiment utilizes the interpretability of the aforementioned core state feature vectors to predict the fault types of the aforementioned electromechanical equipment. The specific process is as follows: 1. Using a causal attention mechanism to analyze the core state feature vector Perform analysis to generate risk analysis vectors. .because Since it's already a fused vector containing future prediction information, there's no need to perform a complete temporal attention aggregation again; instead, it can be directly mapped and parsed. .in, To analyze the weight matrix, It is the analytic layer bias vector, this analytic vector It encapsulates current and future predicted failure risk information.

[0041] 2. Input is fed into a classifier network to predict the fault type. The formula is: .

[0042] 3. Fault Type Determined: The fault type with the highest probability can be determined as the final prediction result. The formula is as follows: And record the confidence level of the prediction. That is, the highest probability value itself.

[0043] 4. Higher-order failure modes ( Analysis and Description: The core state feature vector generated in Step 2 of the system analysis The process involves identifying the non-zero activated dimensions and associating them with a predefined dictionary of higher-order fault combination patterns to generate interpretable textual descriptions. . , .in, To activate the set of dimension indices, This is the sparsity threshold (minimum value). It is a set of textual descriptions (e.g., {“The abnormal count trend of sensor A continues to rise”, “The voltage of key components fluctuates at high frequency”}) used to explain to maintenance personnel the underlying driving factors of the current risk prediction.

[0044] 5. Develop the above results into a structured early warning message. ,in, It is the warning timestamp, indicating the current system time when the message was generated; It predicts the fault type, representing the fault category with the highest probability determined by the classifier (e.g., communication module failure, sensor drift). It is the prediction confidence level, representing the confidence level for predicting the type of failure. The degree of probability certainty (i.e., the highest probability value output by the classifier). These are higher-order fault drivers, representing the core state characteristics of system resolving activation. The resulting set of interpretable textual descriptions is used to explain the specific combinations of abnormal patterns that lead to increased risk.

[0045] This concludes the process. Figure 1 The process is shown below.

[0046] In this embodiment, for electromechanical equipment requiring monitoring on highways, discrete state data such as equipment switching and communication status are statistically analyzed, while continuous physical quantity data such as equipment temperature, voltage, and current are time-series analyzed. Features are then extracted to generate a unified preliminary feature vector. Fault risk features are then extracted from this preliminary feature vector, and finally, a causal attention mechanism is used to analyze these risk features and output fault warning information. This solves the problems of delayed fault detection, high maintenance costs, and slow fault response speed in existing highway electromechanical equipment. It achieves real-time analysis and accurate prediction of the failure probability of key highway equipment, thereby improving the safety and efficiency of highway operation.

[0047] In another embodiment, after obtaining the risk resolution vector, the method further includes: The risk analysis vector is input into a regression network to predict the shape parameters of the Weibull distribution. and scale parameters ; Through the shape parameters and scale parameters Generating the failure probability time function Then through the fault probability time function The warning information is generated based on the fault type, where T is the current time. The remaining running time from the start until the probability of failure exceeds the set threshold.

[0048] This embodiment can also generate risk analysis vectors. Then, the failure probability time function is determined based on this vector. The function predicts the failure time MT and generates early warning information for the platform, so as to provide early warning to maintenance personnel.

[0049] Furthermore, in this embodiment, a fault probability time function can also be generated. Then, through the shape parameters and scale parameters Determine the estimated remaining lifespan of the target electromechanical equipment. The calculation formula is: Then, warning information is generated using the failure probability time function, the estimated remaining lifespan, and the failure type, wherein, This is the Gamma function.

[0050] Furthermore, this embodiment can not only predict the failure time of the equipment through risk analysis vectors, but also determine the estimated remaining lifespan of the equipment. Subsequently, the predicted failure time and estimated remaining lifespan can be added to the early warning information and sent to the platform, such as the early warning message. Modified to This information is then sent to maintenance personnel, allowing them to repair or replace the equipment in advance, thereby reducing the probability of accidents.

[0051] In another embodiment, after obtaining the risk resolution vector, the above method further includes: The risk analysis vector is input into a regression network to predict the shape parameters of the Weibull distribution. and scale parameters ; Through the shape parameters and scale parameters Determine the estimated remaining lifespan of the target electromechanical equipment. The calculation formula is: Then, warning information is generated based on the estimated remaining lifespan and the fault type, wherein, This is the Gamma function.

[0052] This embodiment can also generate risk analysis vectors. The estimated remaining lifespan of the target electromechanical equipment is then determined based on the vector. This estimated remaining lifespan can then be added to the early warning information and sent to the platform, allowing maintenance personnel to repair or replace the equipment in advance, thereby reducing the probability of accidents.

[0053] This application also provides a device for predicting the probability of device failure based on IoT time-series status data, such as... Figure 2 As shown, the device includes: The data acquisition unit 201 is used to acquire discrete state data and continuous physical quantity data of the target electromechanical equipment, and to determine a continuous statistical time series feature vector through the discrete state data and a continuous analytical time series feature vector through the continuous physical quantity data. The feature extraction unit 202 is used to determine the preliminary feature vector of the target electromechanical equipment through the statistical time series feature vector and the analytical time series feature vector, and then extract the preliminary feature vector based on the spatiotemporal dependence to obtain the core state feature vector containing fault risk features. The early warning information generation unit 203 is used to analyze the core state feature vector through a causal attention mechanism to obtain a risk analysis vector, then input the risk analysis vector into a classifier network to obtain the predicted fault type of the target electromechanical equipment, and generate early warning information based on the fault type.

[0054] In another embodiment, after the early warning information generation unit obtains the risk analysis vector, the device further includes: The second early warning information generation unit 204 is used to input the risk analysis vector into a regression network to predict the shape parameters of the Weibull distribution. and scale parameters ; Through the shape parameters and scale parameters Generating the failure probability time function Then through the fault probability time function A warning message is generated based on the fault type, where T is the current time.

[0055] In another embodiment, after the early warning information generation unit obtains the risk analysis vector, the device further includes: The third early warning information generation unit 205 is used to input the risk analysis vector into the regression network to predict the shape parameters of the Weibull distribution. and scale parameters ; Through the shape parameters and scale parameters Determine the estimated remaining lifespan of the target electromechanical equipment. The calculation formula is: Then, warning information is generated based on the estimated remaining lifespan and the fault type, wherein, This is the Gamma function.

[0056] In another embodiment, the second early warning information generation unit generates a fault probability time function. Subsequently, the device further includes: The fourth early warning information generation unit 206 is used to generate early warning information by means of the shape parameters. and scale parameters Determine the estimated remaining lifespan of the target electromechanical equipment. The calculation formula is: Then, warning information is generated using the failure probability time function, the estimated remaining lifespan, and the failure type, wherein, This is the Gamma function.

[0057] The above embodiments of the present invention provide a method for predicting the probability of equipment failure based on IoT time-series status data, and a device for predicting the probability of equipment failure based on IoT time-series status data based on the method. Through the above method and device, real-time monitoring and accurate fault warning of the status of key equipment on highways can be realized, thereby improving the safety and efficiency of highway operation.

[0058] This embodiment also discloses a computer device, such as... Figure 3 As shown, the computer device includes a processor and a memory, the memory storing at least one instruction, which is loaded and executed by the processor to implement any of the above-described alarm analysis methods based on multimodal machine learning.

[0059] Furthermore, in the above-described embodiment of the device failure probability prediction device based on IoT time-series status data, the logical division of each program module is merely illustrative. In practical applications, the above functions can be assigned to different program modules as needed, for example, for the sake of corresponding hardware configuration requirements or the convenience of software implementation. That is, the internal structure of the device failure probability prediction device based on IoT time-series status data can be divided into different program modules to complete all or part of the functions described above.

[0060] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for predicting the probability of device failure based on IoT time-series state data, applied to an intelligent transportation platform, characterized in that, The method includes: The discrete state data and continuous physical quantity data of the target electromechanical equipment are acquired, and a continuous statistical time series feature vector is determined through the discrete state data, and a continuous analytical time series feature vector is determined through the continuous physical quantity data. The preliminary feature vector of the target electromechanical equipment is determined by the statistical time series feature vector and the analytical time series feature vector. Then, the preliminary feature vector is extracted based on the spatiotemporal dependence to obtain the core state feature vector containing fault risk characteristics. The core state feature vector is analyzed by a causal attention mechanism to obtain a risk analysis vector. The risk analysis vector is then input into a classifier network to obtain the predicted fault type of the target electromechanical equipment, and a warning message is generated based on the fault type.

2. The method according to claim 1, characterized in that, After obtaining the risk resolution vector, the method further includes: The risk analysis vector is input into a regression network to predict the shape parameters of the Weibull distribution. and scale parameters ; Through the shape parameters and scale parameters Generating the failure probability time function Then through the fault probability time function The warning information is generated based on the fault type, where T is the current time. The remaining running time from the start until the probability of failure exceeds the set threshold.

3. The method according to claim 1, characterized in that, After obtaining the risk resolution vector, the method further includes: The risk analysis vector is input into a regression network to predict the shape parameters of the Weibull distribution. and scale parameters ; Through the shape parameters and scale parameters Determine the estimated remaining lifespan of the target electromechanical equipment. The calculation formula is: Then, warning information is generated based on the estimated remaining lifespan and the fault type, wherein, This is the Gamma function.

4. The method according to claim 2, characterized in that, Generating the failure probability time function Subsequently, the method further includes: Through the shape parameters and scale parameters Determine the estimated remaining lifespan of the target electromechanical equipment. The calculation formula is: Then, warning information is generated using the failure probability time function, the estimated remaining lifespan, and the failure type, wherein, This is the Gamma function.

5. The method according to claim 1, characterized in that, The process of determining continuous statistical time-series feature vectors using the discrete state data includes: The discrete state data is analyzed by a sliding window, and the frequency and persistence characteristics of abnormal events of each state type are calculated in real time. The number of abnormalities in the sliding window of each state type in the past N time units is calculated, and the immediate impact and duration of each abnormality are determined. The number of abnormalities, the immediate impact and duration of each abnormality are then determined as the statistical time series feature vector.

6. The method according to claim 4, characterized in that, The step of generating early warning information using the failure probability-time function, estimated remaining lifetime, and failure type includes: The target fault type with the highest probability and its corresponding confidence level are determined from the fault types, and the core state feature vector is associated with a predefined high-order fault combination pattern dictionary to generate a textual description representing the fault mode. The text description, the failure probability time function, the estimated remaining lifespan, the target failure type, and the corresponding confidence level are combined into a warning message and sent.

7. A device for predicting the probability of equipment failure based on Internet of Things (IoT) time-series status data, applied to an intelligent transportation platform, characterized in that, The device includes: The data acquisition unit is used to acquire discrete state data and continuous physical quantity data of the target electromechanical equipment, and to determine continuous statistical time series feature vectors through the discrete state data and continuous analytical time series feature vectors through the continuous physical quantity data. The feature extraction unit is used to determine the preliminary feature vector of the target electromechanical equipment through the statistical time series feature vector and the analytical time series feature vector, and then extract the preliminary feature vector based on the spatiotemporal dependence to obtain the core state feature vector containing fault risk features. The early warning information generation unit is used to parse the core state feature vector through a causal attention mechanism to obtain a risk analysis vector, and then input the risk analysis vector into a classifier network to obtain the predicted fault type of the target electromechanical equipment, and generate early warning information based on the fault type.

8. The apparatus according to claim 7, characterized in that, After the early warning information generation unit obtains the risk analysis vector, the device further includes: The second early warning information generation unit is used to input the risk analysis vector into the regression network to predict the shape parameters of the Weibull distribution. and scale parameters ; Through the shape parameters and scale parameters Generating the failure probability time function Then through the fault probability time function The warning information is generated based on the fault type, where T is the current time. The remaining running time from the start until the probability of failure exceeds the set threshold.

9. The apparatus according to claim 7, characterized in that, After the early warning information generation unit obtains the risk analysis vector, the device further includes: The third early warning information generation unit is used to input the risk analysis vector into the regression network to predict the shape parameters of the Weibull distribution. and scale parameters ; Through the shape parameters and scale parameters Determine the estimated remaining lifespan of the target electromechanical equipment. The calculation formula is: Then, warning information is generated based on the estimated remaining lifespan and the fault type, wherein, This is the Gamma function.

10. The apparatus according to claim 8, characterized in that, The second early warning information generation unit generates the fault probability time function. Subsequently, the device further includes: The fourth early warning information generation unit is used to generate early warning information through the shape parameters. and scale parameters Determine the estimated remaining lifespan of the target electromechanical equipment. The calculation formula is: Then, warning information is generated using the failure probability time function, the estimated remaining lifespan, and the failure type, wherein, This is the Gamma function.

Images