Fault prediction method and device based on 5G MEC and medium

Abstract

The invention provides a 5G MEC-based fault prediction method and device and a medium, and relates to the technical field of equipment monitoring and analysis, and the method comprises the steps: continuously obtaining real-time values of various parameters of equipment; acquiring the real-time change rate of each parameter according to the continuous real-time value; according to the real-time change rate of each parameter, obtaining the real-time weight of each parameter for predicting the equipment fault; according to the real-time weight of each parameter, acquiring a comprehensive state prediction value and a comprehensive anomaly detection threshold value of the equipment based on the various parameters; and obtaining equipment fault early warning information according to the comprehensive state prediction value and the comprehensive anomaly detection threshold value. According to the method and the device, the weight of each parameter for predicting the equipment fault is dynamically adjusted through the real-time change rate of each parameter, so that the comprehensive state prediction value and the comprehensive anomaly detection threshold value integrating different degrees of influences of the various parameters are obtained, the equipment state is evaluated more accurately, the equipment anomaly is captured more accurately, and the accuracy of fault prediction is improved.

Description

Technical Field

[0001] The present disclosure at least relates to the technical field of equipment monitoring and analysis, and in particular to a 5G MEC-based fault prediction method, a 5G MEC-based fault prediction device, and a computer-readable storage medium. Background Art

[0002] Condition monitoring of running equipment often involves multiple parameters, such as temperature, vibration, pressure, humidity, etc. Traditional prediction models usually set fixed weights for each parameter, which makes it difficult to flexibly respond to fluctuations in different parameters during equipment operation. When a parameter changes dramatically in a short period of time (such as a sudden rise in temperature or a sharp fluctuation in vibration frequency), traditional prediction models are difficult to accurately capture the potential impact of such drastic changes. Summary of the invention

[0003] The technical problem to be solved by the present disclosure is to provide a 5G MEC-based fault prediction method, a 5G MEC-based fault prediction device and a computer-readable storage medium in response to the above-mentioned shortcomings, so as to solve the problem of how to comprehensively analyze equipment failures by using multiple parameters.

[0004] In a first aspect, the present disclosure provides a 5G MEC-based fault prediction method, the method comprising:

[0005] Continuously obtain real-time values ​​of various parameters of the equipment;

[0006] Based on the continuous real-time value, obtain the real-time change rate of each parameter;

[0007] According to the real-time change rate of each parameter, a real-time weight of each parameter used to predict equipment failure is obtained;

[0008] According to the real-time weight of each parameter, the comprehensive state prediction value and comprehensive anomaly detection threshold of the device based on multiple parameters are obtained;

[0009] Obtain equipment failure warning information based on the comprehensive state prediction value and comprehensive anomaly detection threshold.

[0010] Further, wherein:

[0011] The real-time values ​​of various parameters of the equipment are continuously received by the MEC from sensors. The sensors are deployed to collect the real-time values ​​of various parameters of industrial equipment. The period for collecting the real-time values ​​of the parameters is set to Δt;

[0012] The real-time change rate, real-time weight, comprehensive state prediction value, comprehensive anomaly detection threshold and equipment failure warning information are all calculated and obtained by MEC. MEC is deployed at the factory where the industrial equipment is located. A single-parameter state prediction model and a single-parameter anomaly detection threshold model are also deployed in MEC to obtain the state prediction value and anomaly detection threshold of each parameter.

[0013] The single-parameter state prediction model and the single-parameter anomaly detection threshold model are sent from the cloud server to the MEC. The optimization functions of the single-parameter state prediction model and the single-parameter anomaly detection threshold model are deployed in the cloud server. The single-parameter state prediction model and the single-parameter anomaly detection threshold model in the MEC are the latest optimized by the cloud server.

[0014] Sensors and MEC, MEC and cloud servers are connected through 5G private networks. MEC is a multi-access edge computing device, and 5G is the fifth generation of mobile communication technology.

[0015] Furthermore, the real-time values ​​of various parameters of the device are continuously obtained, including:

[0016] Periodically receive the real-time value of each parameter;

[0017] The acquisition time length T = n·Δt, where n represents the number of received real-time values ​​of each parameter within T, and Δt represents the receiving period;

[0018] Obtained from the current time t 0 The n x in the previous T i (t′), i=[1,2,…,m], represents the i-th parameter, m represents a total of m parameters, t′=[t i ,t 0 -Δt,t 0 -2·Δt,…,t 0 -n·Δt], which represents n past moments.

[0019] Furthermore, based on the continuous real-time value, the real-time change rate of each parameter is obtained, including:

[0020] Get the real-time change rate of each parameter according to the following formula:

[0021]

[0022] ΔD i (t 0 ) indicates that the i-th parameter is accumulated within T to the current time t 0 The real-time rate of change.

[0023] Furthermore, according to the real-time change rate of each parameter, the real-time weight of each parameter used to predict equipment failure is obtained, specifically including:

[0024] The real-time weight of each parameter used to predict equipment failure is obtained according to the following formula:

[0025]

[0026] w i (t 0 ) indicates that the i-th parameter is at the current time t 0 The real-time weight.

[0027] Furthermore, according to the real-time weight of each parameter, the comprehensive state prediction value and comprehensive anomaly detection threshold of the device based on multiple parameters are obtained, including:

[0028] According to the single parameter state prediction model, the single parameter state prediction value P of each parameter is obtained. i (t), according to P i (t) and the real-time weight w of each parameter i (t 0 ) Obtain the comprehensive status prediction value of the device based on multiple parameters:

[0029]

[0030] According to the single parameter anomaly detection threshold model, the single parameter anomaly detection threshold θ of each parameter is obtained. i (t), according to θ i (t) and the real-time weight w of each parameter i (t 0 ) Get the device's comprehensive anomaly detection threshold based on multiple parameters:

[0031]

[0032] Among them, i represents the i-th parameter, t represents the prediction time, and t 0 represents the current moment, and m represents the total number of types of various parameters.

[0033] Further, the single parameter anomaly detection threshold θ of each parameter is obtained i (t), including:

[0034] The single parameter anomaly detection threshold for each parameter is obtained according to the following formula:

[0035] θ i (t)=(μ i +k 1 ·σ i +k 2 ΔD i (t 0 )+k 3 ·U i (t 0 ))

[0036] Among them, μ i represents the historical mean of the i-th parameter, σ i represents the historical standard deviation of the i-th parameter, U i (Δt 0 ) represents the current uncertainty of the single parameter state prediction model of the i-th parameter, k 1 , k 2 , k 3 There are three adjustment coefficients respectively.

[0037] Furthermore, the method further comprises:

[0038] In response to P final (t)≥θ(t), send equipment failure warning information to equipment maintenance personnel, and receive feedback information from equipment maintenance personnel after maintaining the equipment;

[0039] In response to the feedback information showing that the equipment fault warning information is inaccurate, the single parameter state prediction model and the single parameter anomaly detection threshold model are optimized, and the uncertainty U of the optimized single parameter state prediction model of each parameter is obtained. i (t 0 ), uncertainty U i (t 0 ) is expressed using prediction error or confidence.

[0040] In a second aspect, the present disclosure provides a 5G MEC-based fault prediction device, the device comprising:

[0041] Numerical module, used to continuously obtain the real-time values ​​of various parameters of the equipment;

[0042] A change rate module, connected to the value module, is used to obtain the real-time change rate of each parameter based on the continuous real-time value;

[0043] A weight module, connected to the change rate module, for obtaining a real-time weight of each parameter for predicting equipment failure according to the real-time change rate of each parameter;

[0044] The comprehensive module is connected with the weight module and is used to obtain the comprehensive state prediction value and comprehensive abnormality detection threshold of the device based on multiple parameters according to the real-time weight of each parameter;

[0045] The early warning module is connected to the comprehensive module and is used to obtain equipment failure early warning information based on the comprehensive state prediction value and the comprehensive anomaly detection threshold.

[0046] In a third aspect, the present disclosure provides a computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, the 5G MEC-based fault prediction method as described above is implemented.

[0047] The present disclosure provides a fault prediction method based on 5G MEC, a fault prediction device based on 5G MEC, and a computer-readable storage medium. The method obtains multiple parameter values ​​of a device in real time, calculates the real-time change rate of each parameter, and dynamically adjusts the weight of each parameter for predicting device failures, thereby obtaining a comprehensive state prediction value and a comprehensive anomaly detection threshold that integrate the different degrees of influence of multiple parameters, so as to more accurately evaluate the device state and capture device anomalies, thereby improving the accuracy of fault prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is a flow chart of a fault prediction method based on 5G MEC according to an embodiment of the present disclosure;

[0049] Figure 2 It is a structural diagram of a fault prediction device based on 5G MEC according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0050] In order to enable those skilled in the art to better understand the technical solution of the present disclosure, the embodiments of the present disclosure will be further described in detail below with reference to the accompanying drawings.

[0051] It should be understood that the specific embodiments and drawings described herein are only used to explain the present disclosure rather than to limit the present disclosure.

[0052] It can be understood that, in the absence of conflict, the various embodiments of the present disclosure and the various features in the embodiments can be combined with each other.

[0053] It can be understood that, for the convenience of description, the drawings of the present disclosure only show parts related to the present disclosure, while parts irrelevant to the present disclosure are not shown in the drawings.

[0054] It can be understood that each module or unit involved in the embodiments of the present disclosure may correspond to only one entity structure, or may be composed of multiple entity structures, or multiple modules or units may be integrated into one entity structure.

[0055] It will be understood that, in the absence of conflict, the functions and steps marked in the flowcharts and block diagrams of the present disclosure may occur in an order different from that marked in the drawings.

[0056] It is understood that the flowcharts and block diagrams of the present disclosure illustrate the possible architectures, functions, and operations of the systems, devices, equipment, and methods according to the various embodiments of the present disclosure. Each box in the flowchart or block diagram may represent a module, unit, program segment, or code, which contains executable instructions for implementing the specified functions. Moreover, each box or combination of boxes in the block diagram and flowchart may be implemented by a hardware-based device that implements the specified functions, or may be implemented by a combination of hardware and computer instructions.

[0057] It can be understood that the modules and units involved in the embodiments of the present disclosure can be implemented by software or hardware, for example, the modules and units can be located in a processor.

[0058] In order to facilitate the understanding of the present disclosure, the main inventive concept of the present disclosure is first introduced in combination with specific application methods.

[0059] The present disclosure can specifically provide a fault prediction method based on 5G (fifth generation mobile communication technology) MEC (multi-access edge computing), which is applied to the industrial Internet.

[0060] The Industrial Internet forms an intelligent industrial system by interconnecting sensors, control systems and equipment. With the maturity of technologies such as 5G, cloud computing, big data and artificial intelligence, the Industrial Internet has become an important means to improve the automation and intelligence of production processes.

[0061] Equipment condition monitoring is an important part of the Industrial Internet. It aims to monitor the operating status of equipment in real time, predict potential failures, and optimize maintenance strategies through sensors and data analysis technology. Traditional maintenance models often rely on regular manual inspections or post-failure repairs, which are inefficient and costly. Condition monitoring technology enables companies to achieve "predictive maintenance", that is, identifying potential problems before failures occur, thereby reducing equipment downtime and maintenance costs.

[0062] The main application scenarios at present include: Manufacturing: In the production line, real-time monitoring of the status of machinery and equipment can be used to prevent unexpected downtime, and predictive maintenance can be used to reduce losses caused by equipment failure; Energy industry: Monitoring the operating status of large-scale mechanical equipment such as generators and wind power generation equipment to ensure the continuity and stability of energy production; Transportation and logistics: Status monitoring of automated warehousing equipment, transport vehicles, etc. to ensure the efficient operation of logistics and transportation.

[0063] The core technologies of the condition monitoring method include: sensor technology: real-time collection of equipment operation data through sensors such as temperature, vibration, pressure, and humidity; digital twin technology: by building a virtual model of the equipment (digital twin), real-time monitoring and simulation of the equipment status throughout its life cycle are carried out to provide data support for equipment maintenance and optimization.

[0064] Although the existing industrial equipment condition monitoring technology has made significant progress, it still has some shortcomings and challenges. Although the condition monitoring system can automatically monitor the operation of equipment, it still requires human intervention for analysis, judgment and maintenance. In particular, the cause analysis and repair of complex faults rely on expert experience. The impact of these shortcomings is that human intervention may reduce the degree of automation of the system, affect maintenance efficiency in complex industrial environments, and prolong downtime.

[0065] In view of the above defects, the technical concept of the technical method disclosed in the present invention is: deploy sensors on industrial equipment of the industrial Internet to realize the collection, transmission and processing of equipment data, etc., arrange MEC based on 5G private network in the entire industrial area, perform data transmission and calculation, build a multi-parameter dynamic adjustment model based on parameter change rate, and deploy it on MEC. MEC calculates the dynamic weight of each parameter, calculates the comprehensive prediction value of the equipment status through the data transmitted by 5G and the prediction model of MEC, and then calculates based on the comprehensive abnormal detection threshold of 5G and MEC. Once it is predicted that the threshold is exceeded, 5G MEC will issue an early warning in time and notify human intervention. Finally, the collaborative work of edge computing and cloud computing is realized through the 5G network, and global data analysis and model optimization are performed on the cloud. Through this method, the efficiency of fault detection and processing can be effectively improved, and the dependence on manual work can be reduced.

[0066] It can be understood that the prediction model, cloud-edge collaboration and other technologies of the technical solution of the present invention can be implemented with reference to the existing technology, and the method of the present invention can be used in combination with the digital twin technology.

[0067] Embodiment 1:

[0068] like Figure 1 As shown, the present disclosure provides a fault prediction method based on 5G MEC, the method comprising:

[0069] S1. Continuously obtain the real-time values ​​of various parameters of the equipment;

[0070] S2. Obtain the real-time change rate of each parameter based on the continuous real-time value;

[0071] S3. Obtaining a real-time weight of each parameter for predicting equipment failure according to the real-time change rate of each parameter;

[0072] S4. According to the real-time weight of each parameter, obtain the comprehensive state prediction value and comprehensive anomaly detection threshold of the device based on multiple parameters;

[0073] S5. Obtain equipment failure warning information based on the comprehensive state prediction value and the comprehensive abnormality detection threshold.

[0074] In this embodiment, by acquiring multiple parameter values ​​of the equipment in real time, calculating the real-time change rate of each parameter, and dynamically adjusting the weight of each parameter for predicting equipment failure, a comprehensive state prediction value and a comprehensive anomaly detection threshold that integrate the different degrees of influence of multiple parameters are obtained, thereby achieving more accurate assessment of equipment status and capture of equipment anomalies, thereby improving the accuracy of fault prediction.

[0075] Specifically, this embodiment can provide an industrial equipment status monitoring and prediction method, which can be applied in the field of industrial Internet to predict equipment failures and issue failure warning information in a timely manner. The main innovations are:

[0076] 1) Dynamically adjust parameter weights: Traditional prediction models usually set fixed weights for each parameter, which makes it difficult to flexibly respond to fluctuations in different parameters during equipment operation. The model of this embodiment introduces a dynamic weight adjustment mechanism, which adjusts the weight according to the change rate of each parameter, so that the parameters with more significant changes have a greater impact on the prediction results.

[0077] Benefits: When a parameter changes dramatically in a short period of time (such as a sudden rise in temperature or a sharp fluctuation in vibration frequency), the model will automatically increase the weight of the parameter and improve the sensitivity to the parameter, thereby more accurately capturing potential equipment anomalies.

[0078] 2) Multi-parameter fusion prediction: The model comprehensively considers multiple parameters (such as temperature, vibration, speed, etc.), not only analyzing the status of each parameter one by one, but also weighting and synthesizing the prediction results of each parameter. The impact of each parameter on the overall prediction is dynamically adjusted according to the rate of change, so it can reflect the health status of the equipment more comprehensively and three-dimensionally.

[0079] Benefits: Different parameters often influence each other. Through multi-parameter fusion prediction, the system can more accurately evaluate the overall status of the equipment instead of relying on the change of a single parameter, thus avoiding misjudgment due to the abnormality of a single parameter.

[0080] 3) Introducing the rate of change as the core driver: Traditional prediction models rely more on historical averages or fixed thresholds, while this model uses the rate of change as the core driver. The weight and prediction result of each parameter are directly determined by the current rate of change, reflecting the dynamic changes in the device status. This rate-based mechanism makes the model more sensitive to abnormal changes in device operation.

[0081] Benefits: The rate of change captures dramatic fluctuations in parameters in the short term, so even if the equipment parameters are historically within the normal range, the model can still issue early warnings based on rapidly changing trends, avoiding delayed responses caused by relying on fixed thresholds.

[0082] 4) Dynamic calculation of comprehensive anomaly detection threshold: A comprehensive anomaly detection threshold is introduced into the model, which is obtained by dynamically calculating the weighted sum of multiple parameters. Different from the traditional single parameter threshold, the comprehensive threshold is calculated based on the historical mean, change rate, prediction uncertainty and other factors of each parameter, and is updated at any time, making the detection process more flexible and accurate.

[0083] Benefits: The comprehensive threshold takes into account the interactive effects of multi-parameter data and can more accurately detect anomalies when multiple parameters change at the same time. In addition, the threshold can be dynamically adjusted as the device status changes, reducing the risk of false positives and false negatives.

[0084] In one embodiment, wherein:

[0085] The real-time values ​​of various parameters of the equipment are continuously received by the MEC from sensors. The sensors are deployed to collect the real-time values ​​of various parameters of industrial equipment. The period for collecting the real-time values ​​of the parameters is set to Δt;

[0086] The real-time change rate, real-time weight, comprehensive state prediction value, comprehensive anomaly detection threshold and equipment failure warning information are all calculated and obtained by MEC. MEC is deployed at the factory where the industrial equipment is located. A single-parameter state prediction model and a single-parameter anomaly detection threshold model are also deployed in MEC to obtain the state prediction value and anomaly detection threshold of each parameter.

[0087] The single-parameter state prediction model and the single-parameter anomaly detection threshold model are sent from the cloud server to the MEC. The optimization functions of the single-parameter state prediction model and the single-parameter anomaly detection threshold model are deployed in the cloud server. The single-parameter state prediction model and the single-parameter anomaly detection threshold model in the MEC are the latest optimized by the cloud server.

[0088] Sensors and MEC, MEC and cloud servers are connected through 5G private networks. MEC is a multi-access edge computing device, and 5G is the fifth generation of mobile communication technology.

[0089] In this embodiment, an industrial equipment status monitoring and prediction method is implemented based on 5G MEC technology. First, it is necessary to deploy the corresponding 5G private network and MEC architecture, which specifically includes the following deployment contents:

[0090] Goal: Combine 5G networks with MEC technology to provide high-speed, low-latency data transmission and real-time computing capabilities for sensor data collection and prediction.

[0091] 5G private network deployment: Deploy 5G base stations in industrial parks to ensure that devices and edge computing nodes are connected through 5G private networks, achieving high bandwidth, low latency and high reliability of data transmission. In addition, 5G private networks provide dedicated network resources to ensure stable and fast data transmission in industrial environments without interference from the public network.

[0092] Sensor deployment: Deploy sensors on each industrial device to collect multi-modal data status data of certain key components of the device every second or every minute, such as temperature, pressure, vibration, etc.

[0093] MEC deployment: Edge computing nodes are closely integrated with 5G base stations and deployed in workshops or near equipment. MEC is responsible for local processing and real-time analysis of multi-parameter data of equipment, reducing the time for data transmission to the cloud and improving response speed.

[0094] MEC introduction: MEC can be simply described by the following equation: MEC = (connection + computing) × capacity.

[0095] Connection: Solve the problem of multiple access. Regardless of whether users access through 4G network, 5G network or fixed network, they can access the same service deployed at the edge and obtain the same user perception.

[0096] Computing: Provides computing resources required for edge deployment services, including general computing capabilities of CPU (Central Processing Unit), as well as heterogeneous accelerated computing capabilities provided by GPU (Graphic Processing Unit), FPGA (Field Programmable Gate Array), etc. At the same time, more importantly, computing power can be flexibly scheduled according to business needs.

[0097] Capabilities: Provide edge cloud resource capabilities, wireless network capabilities, business enabling capabilities, etc., to provide various conveniences for local computing and processing of edge applications.

[0098] Usually, the UPF (User Plane Function) network element in 5G is sunk closer to the user, and then the MEC function is added on the basis of UPF to provide the IT (Information Technology) services and cloud computing functions required by telecom users nearby, thereby creating a telecom-grade service environment with high performance, low latency and high bandwidth, allowing users to enjoy high-quality business experience.

[0099] Equipment data collection, real-time transmission and processing: The real-time transmission of equipment status data is achieved through the 5G network, and it is processed locally in the MEC. That is, the sensors on the equipment collect data such as temperature, vibration, and speed, and these data are transmitted to the MEC for pre-processing through the 5G network with low latency.

[0100] Cloud collaboration and global optimization: The 5G network enables the collaboration between edge computing and the cloud, and the cloud performs data analysis and global optimization. Specifically, MEC performs data processing and anomaly detection, and the cloud is responsible for long-term analysis of historical data and optimization of model parameters, and / or the cloud optimizes the prediction model by analyzing data from multiple devices, and sends the optimization results to MEC through the 5G network. Through model optimization and feedback, the prediction accuracy of the system is further improved, and edge and cloud collaboration is achieved.

[0101] Implementation case: 5G private network sensor deployment in smart manufacturing

[0102] In an intelligent manufacturing workshop, multiple industrial equipment are used, including CNC machine tools, automated assembly lines, mobile robots, etc. 5G base stations are deployed in the workshop to ensure that each sensor can be connected to the network through the 5G private network. Sensors are deployed on CNC machine tools and automation equipment to collect data such as temperature, vibration, and speed of certain key components. The data is transmitted to the MEC server through the 5G private network. MEC uses a multi-modal collaborative prediction model (multi-parameter dynamic adjustment prediction model) to predict the future state of key components of the equipment and uses a warning state propagation model to predict the impact between components. For example, using a multi-parameter dynamic adjustment prediction model based on parameter change rate, the CNC machine tools in the intelligent manufacturing workshop are monitored and fault predicted through 5GMEC. By monitoring multiple parameters of the machine tools, such as temperature, vibration, and speed, the system can dynamically adjust the weight of the model, identify key factors affecting the health of the equipment, and issue warnings and take maintenance measures in a timely manner.

[0103] It can be understood that the multimodal collaborative prediction model (multi-parameter dynamic adjustment prediction model) refers to a model that adds real-time weights obtained according to the real-time change rate of each parameter. The model can first establish a single parameter model for each parameter and then integrate the influence of each parameter through weights. It can also add the influence of the weight of each parameter to the model that integrates each parameter. The specific model design can be designed according to the characteristics of the prediction of each device.

[0104] In one embodiment, continuously obtaining real-time values ​​of various parameters of the device specifically includes:

[0105] Periodically receive the real-time value of each parameter;

[0106] The acquisition time length T = n·Δt, where n represents the number of received real-time values ​​of each parameter within T, and Δt represents the receiving period;

[0107] Obtained from the current time t 0 The n x in the previous T i (t′), i=[1,2,…,m], represents the i-th parameter, m represents a total of m parameters, t′=[t i ,t 0 -Δt,t 0 -2·Δt,…,t 0 -n·Δt], which represents n past moments.

[0108] In this embodiment, the following are the specific steps and mechanisms for deploying sensor data collection and combining it with 5GMEC:

[0109] Deploy a 5G private network in the factory. At the same time, deploy an MEC server based on the 5G private network. The MEC server is responsible for processing data from industrial equipment and supports real-time processing and analysis.

[0110] The sensors of industrial equipment are connected to the MEC server through the 5G private network. After the sensors collect equipment data, they can transmit the data to the MEC server for processing. For scenarios that require high-frequency collection (such as vibration monitoring, equipment noise detection, etc.), the high bandwidth of the 5G private network supports the transmission of multiple data points per second, ensuring high-precision monitoring.

[0111] Through the 5G private network, multimodal data (vibration, temperature, pressure, speed) of certain key components of industrial equipment collected by sensors can be quickly and stably transmitted to the MEC server for processing. MEC uses a multimodal collaborative prediction model to predict the future state of key components of the equipment and a warning state propagation model to predict the impact between components.

[0112] The multimodal collaborative prediction model is a multi-parameter dynamic adjustment model built based on the parameter change rate and deployed on MEC. Its function is to calculate the change rate of each parameter based on the data transmitted from the 5G network, and dynamically adjust the weights of each parameter in the monitoring and prediction model through MEC to achieve flexible prediction. In order to calculate the real-time change rate of each parameter, it is necessary to obtain multiple data collected continuously for a period of time before the current moment, and obtain the fluctuation of this parameter based on the data changes during this period. The specific number of data required can be set based on experience.

[0113] In one embodiment, obtaining the real-time change rate of each parameter according to the continuous real-time value specifically includes:

[0114] Get the real-time change rate of each parameter according to the following formula:

[0115]

[0116] ΔD i (t 0 ) indicates that the i-th parameter is accumulated within T to the current time t 0 The real-time rate of change.

[0117] In this embodiment, MEC calculates the change rate of each device parameter (such as temperature, vibration frequency, etc.) to determine its fluctuation in the past T time period, where Δt can be a unit time, such as 1s. Parameters with a larger change rate are often more sensitive to changes in device status, and their weights should be increased.

[0118] In one embodiment, according to the real-time change rate of each parameter, the real-time weight of each parameter for predicting equipment failure is obtained, specifically including:

[0119] The real-time weight of each parameter used to predict equipment failure is obtained according to the following formula:

[0120]

[0121] w i (t 0 ) represents the i-th parameter at the current time t 0 The real-time weight.

[0122] In this embodiment, MEC calculates the change rate of each parameter of the data transmitted through the 5G private network, and dynamically adjusts its weight in the comprehensive model to ensure the leading role of key parameters in device status prediction. Parameters with high change rates indicate that they have a greater impact on the device status, and their weights should be increased accordingly.

[0123] It is understandable that the real-time weight mainly considers the impact of the change rate on the prediction results. If the sensitivity of specific equipment to different parameters is different, the original fixed weight can still be retained, such as setting the corresponding fixed weight in the state prediction model of each parameter, or obtaining the fixed weight of each parameter. i Add the above real-time weight calculation formula, such as,

[0124] In one embodiment, according to the real-time weight of each parameter, the comprehensive state prediction value and the comprehensive abnormality detection threshold of the device based on multiple parameters are obtained, specifically including:

[0125] According to the single parameter state prediction model, the single parameter state prediction value P of each parameter is obtained. i (t), according to P i (t) and the real-time weight w of each parameter i (t 0) Obtain the comprehensive status prediction value of the device based on multiple parameters:

[0126]

[0127] According to the single parameter anomaly detection threshold model, the single parameter anomaly detection threshold θ of each parameter is obtained. i (t), according to θ i (t) and the real-time weight w of each parameter i (t 0 ) Get the device's comprehensive anomaly detection threshold based on multiple parameters:

[0128]

[0129] Among them, i represents the i-th parameter, t represents the prediction time, and t 0 represents the current moment, and m represents the total number of types of various parameters.

[0130] In this embodiment, the prediction includes two parts. The first is to calculate the comprehensive prediction value of the device status through the data transmitted by 5G and the prediction model of MEC. The formula is shown above, where P final (t) is the final predicted value, P i (t) is the result predicted for each parameter (such as temperature, vibration, etc.), w i (t) is the dynamic weight of each parameter. Considering that the prediction time can be the current time or the future time, t 0 Indicates the current moment, and t indicates the predicted moment, to distinguish them. The second is the calculation of the comprehensive anomaly detection threshold based on 5G and MEC. In the status monitoring and prediction of industrial Internet equipment, the anomaly detection threshold is a key parameter used to determine whether the equipment is in an abnormal state. To ensure the accuracy of anomaly detection, the threshold setting needs to flexibly respond to the dynamic changes in the equipment status. In the scenario of multi-parameter prediction, the calculation of the anomaly detection threshold needs to consider the comprehensive impact of multiple parameters (such as temperature, vibration, speed, etc.). The threshold calculation of a single parameter is not sufficient to fully evaluate the status of the equipment, especially when multiple parameters are abnormal at the same time, their interactions should be considered.

[0131] In one embodiment, the single parameter anomaly detection threshold θ of each parameter is obtained. i (t), including:

[0132] The single parameter anomaly detection threshold for each parameter is obtained according to the following formula:

[0133] θ i (t)=(μ i +k 1 ·σ i +k 2 ΔD i(t 0 )+k 3 ·U i (t 0 ))

[0134] Among them, μ i represents the historical mean of the i-th parameter, σ i represents the historical standard deviation of the i-th parameter, U i (t 0 ) represents the current uncertainty of the single parameter state prediction model of the i-th parameter, k 1 , k 2 , k 3 There are three adjustment coefficients respectively.

[0135] In this embodiment, a comprehensive anomaly detection threshold calculation method is to dynamically adjust based on historical data of device status, parameter change rate, prediction uncertainty and other factors to improve the sensitivity and reliability of detection. MEC calculates the comprehensive anomaly detection threshold by integrating the status of multiple parameters and transmits the result to the cloud and control system through the 5G private network. The calculation formula of the comprehensive anomaly detection threshold is as follows:

[0136]

[0137] Formula explanation:

[0138] w i (t 0 ): The weight of the i-th parameter.

[0139] μ i : The historical mean of the i-th parameter, reflecting the average operating level of the equipment under normal working conditions.

[0140] σ i : The historical standard deviation of the ith parameter reflects the fluctuation range of the equipment. The normal operation mode of the equipment can be captured by combining the mean and standard deviation.

[0141] ΔD i (t 0 ): The i-th parameter at time t 0 The rate of change at that time indicates the degree of change of the current parameter relative to the previous one. A high rate of change may indicate potential anomalies, so the weight should be increased.

[0142] U i (t 0 ): The prediction uncertainty of the oth parameter (e.g., the prediction error or confidence interval of the model), which reflects the uncertainty of the prediction model at the current moment. High uncertainty means that the model may have biased predictions for the current data, so the threshold should be appropriately increased to prevent too many false positives.

[0143] k 1 , k 2 , k 3 : They are the adjustment coefficients of mean-standard deviation term, rate of change term, and uncertainty term. These three coefficients can be used to flexibly adjust the impact of various factors on the threshold according to the actual application scenario.

[0144] Usually, k 1 Used to control the response to historical data fluctuations by adjusting k 1 , can determine the system's tolerance to historical fluctuations. If the system's historical fluctuations are large, appropriately increasing the standard deviation term can reduce false positives.

[0145] k 2 Used to adjust the sensitivity to the rate of change. The parameter change rate reflects the dynamic changes of the current device state. The higher the rate of change, the more likely the current state is unstable and the greater the possibility of failure. 2 Controlling the impact of the rate of change on the threshold can increase the sensitivity of the system when there are large fluctuations. If the rate of change increases dramatically, the system will actively relax the threshold to capture potential abnormal signals in a timely manner.

[0146] k 3 Used to deal with the uncertainty of the prediction model. By adjusting the coefficient, the sensitivity of the system can be flexibly adjusted to ensure sufficient detection flexibility when the device state changes, while avoiding frequent false alarms caused by excessive sensitivity. When the confidence of the model is low (that is, the uncertainty is high), the anomaly detection threshold should be appropriately relaxed to reduce false alarms caused by model errors. By adjusting k 3 , the weight of this item can be flexibly adjusted according to the actual model performance. This is because the prediction model is usually accompanied by a certain degree of uncertainty or error.

[0147] According to the above description, the present disclosure also includes the fifth innovative point:

[0148] 5) Adjustment based on prediction uncertainty: The model is not only adjusted based on the rate of change, but also incorporates prediction uncertainty (such as the model's confidence interval or prediction error) to adjust the weights and thresholds of parameters. By considering the uncertainty of the prediction model, the system can relax the criteria for anomaly detection when the confidence level is low, reducing false positives caused by prediction errors.

[0149] Benefits: When the uncertainty of equipment status is high, the model can flexibly deal with data noise and prediction errors, avoid frequent false alarms, and improve the stability of predictions.

[0150] In one embodiment, the method further comprises:

[0151] In response to P final(t)≥θ(t), send equipment failure warning information to equipment maintenance personnel, and receive feedback information from equipment maintenance personnel after maintaining the equipment;

[0152] In response to the feedback information showing that the equipment fault warning information is inaccurate, the single parameter state prediction model and the single parameter anomaly detection threshold model are optimized, and the uncertainty U of the optimized single parameter state prediction model of each parameter is obtained. i (t 0 ), uncertainty U i (t 0 ) is expressed using prediction error or confidence.

[0153] In this embodiment, when the predicted value P final (t) exceeds the calculated comprehensive anomaly detection threshold θ(t), the system issues a fault warning through the 5G private network, indicating that the device may fail, that is:

[0154]

[0155] After the warning is issued, the maintenance personnel will take maintenance measures in a timely manner and feedback the maintenance results. Based on the feedback information, it can be determined whether the global model needs to be optimized in the cloud.

[0156] A more specific complete calculation example is shown below:

[0157] Scenario description: In a smart manufacturing workshop, multiple state parameters (temperature, vibration, speed, etc.) of CNC machine tools collected by sensors are transmitted to MEC through the 5G private network for processing and prediction. The system dynamically adjusts the model based on the parameter change rate, calculates the comprehensive anomaly detection threshold of the equipment, and issues an early warning through the 5G network when an anomaly occurs. The MEC architecture diagram can adopt the usual architecture.

[0158] 5G private network and MEC deployment: 5G base stations are deployed in the workshop, and data on multiple parameters such as temperature, vibration, and speed of CNC machine tools are transmitted to MEC through the 5G network. MEC is responsible for processing data and calculating the rate of change, weight, and anomaly detection threshold.

[0159] Data collection: The sensors on the machine tool collect temperature, vibration frequency and speed once a second. The temperature, vibration and speed data of the machine tool are collected every second and transmitted to MEC via the 5G network.

[0160] Temperature data: Current temperature: T(t) = 70°C; Historical average: μ T =60℃; Historical standard deviation: σ T =5℃.

[0161] Vibration frequency data: Current vibration frequency: V(t) = 500Hz; Historical average: μ V=450Hz; historical standard deviation: σ V =30Hz.

[0162] Speed ​​data: Current speed: R(t) = 3200RPM; Historical average: μ R =3000RPM; Historical standard deviation: σ R =100RPM.

[0163] Model prediction: To identify changes in the equipment operating status, MEC calculates the rate of change ΔD of each parameter i (t), thereby determining which parameter is most sensitive to the current state change and dynamically adjusting the weights in the model according to the rate of change.

[0164] Specifically, the vibration change rate is calculated as follows: Assuming that the vibration frequency V(t) collected by the system at time t is 500 Hz, and the vibration frequency at t-1 is 480 Hz, the following calculation is performed:

[0165]

[0166] The rate of change of the vibration frequency is 4.17%.

[0167] Temperature change rate calculation: Assuming that the temperature rises from 65°C at t-1 to 70°C, the temperature change rate is:

[0168]

[0169] The temperature change rate is 7.69%.

[0170] Speed ​​change rate calculation: Current speed R(t) = 3200RPM, previous speed R(t-1) = 3150RPM, speed change rate formula:

[0171]

[0172] The rate of change of the rotation speed is 1.59%.

[0173] According to the rate of change of the parameters, MEC dynamically adjusts the dynamic weight of each parameter to ensure that the most sensitive parameter has a greater impact on the prediction results. The total rate of change is:

[0174]

[0175] Temperature weight: w T (t)=0.0769 / 0.1345=0.572;

[0176] The weight of vibration: w V (t)=0.0417 / 0.1345=0.310;

[0177] The weight of the speed: w R (t)=0.0159 / 0.1345=0.118.

[0178] Fault prediction: MEC dynamically adjusts the prediction model through multiple parameters to calculate the final prediction value P final (t):

[0179]

[0180] Assume that the prediction model obtains: temperature prediction value: PT(t) = 75°C; vibration prediction value: PV(t) = 510Hz; speed prediction value: PR(t) = 3200RPM.

[0181] Final prediction value calculation:

[0182] P final (t)=0.572*75+0.310*510+0.118*3200=578.6.

[0183] Finally, based on the changing multi-parameter data, MEC calculates the comprehensive anomaly detection threshold as follows:

[0184]

[0185] Assume that the adjustment coefficient k 1 =1.5, k 2 =0.5, k 3 =1, assuming that the temperature prediction uncertainty UT(t) = 2°C, the vibration frequency prediction uncertainty UV(t) = 15Hz, and the speed prediction uncertainty UR(t) = 50RPM.

[0186] Temperature anomaly detection threshold:

[0187] θ T (t)=60+1.5*5+0.5*0.0769*100+1*2=73.345℃

[0188] Abnormal vibration frequency detection threshold:

[0189] θ V (t)=450+1*30+0.5*0.0417*450+1*15=504.3825Hz

[0190] Abnormal speed detection threshold:

[0191] θ R (t)=3000+1*100+0.5*0.0159*3000+1*50=3173.85RPM

[0192] MEC comprehensively calculates the threshold-weighted sum of all parameters:

[0193] θ(t)=0.572*73.345+0.310*504.3825+0.118*3173.85=572.43

[0194] Therefore, the comprehensive anomaly detection threshold of the device is 572.43, which is calculated by combining the dynamic weights of temperature, vibration, and rotation speed with their respective anomaly detection thresholds.

[0195] Because P final (t)=578.6>θ(t)=572.43. The system issues a fault warning through the 5G private network, reminding equipment maintenance personnel to take timely measures.

[0196] Cloud optimization and feedback: The cloud system analyzes the historical data of multiple devices, globally optimizes the prediction model, and sends the new model parameters to MEC through the 5G network to further improve the prediction accuracy.

[0197] In the above example, by combining 5G and MEC, the system realizes the monitoring and efficient prediction of equipment status, shortens the equipment anomaly detection time by about 20%, and advances the fault warning by 30 minutes, thus improving the equipment maintenance efficiency. Applying this method to the information system security active defense platform and combining it with the industrial equipment status monitoring and security warning method can effectively improve the efficiency of fault detection and processing, reduce the dependence on manual labor, and improve the overall operational efficiency and quality.

[0198] Embodiment 2:

[0199] like Figure 2 As shown, the present disclosure provides a 5G MEC-based fault prediction device, the device comprising:

[0200] Numerical module 1, used to continuously obtain real-time values ​​of various parameters of the equipment;

[0201] The change rate module 2 is connected to the value module 1 and is used to obtain the real-time change rate of each parameter according to the continuous real-time value;

[0202] The weight module 3 is connected to the change rate module 2 and is used to obtain the real-time weight of each parameter used to predict equipment failure according to the real-time change rate of each parameter;

[0203] The comprehensive module 4 is connected to the weight module 3 and is used to obtain the comprehensive state prediction value and the comprehensive abnormality detection threshold of the device based on multiple parameters according to the real-time weight of each parameter;

[0204] The early warning module 5 is connected to the comprehensive module 4 and is used to obtain equipment failure early warning information based on the comprehensive state prediction value and the comprehensive abnormality detection threshold.

[0205] It should be noted that, with reference to the description of Example 1, the device may specifically be an MEC, a system including an MEC, or other similar devices.

[0206] In one embodiment, wherein:

[0207] The real-time values ​​of various parameters of the equipment are continuously received by the MEC from sensors. The sensors are deployed to collect the real-time values ​​of various parameters of industrial equipment. The period for collecting the real-time values ​​of the parameters is set to Δt;

[0208] The real-time change rate, real-time weight, comprehensive state prediction value, comprehensive anomaly detection threshold and equipment failure warning information are all calculated and obtained by MEC. MEC is deployed at the factory where the industrial equipment is located. A single-parameter state prediction model and a single-parameter anomaly detection threshold model are also deployed in MEC to obtain the state prediction value and anomaly detection threshold of each parameter.

[0209] The single-parameter state prediction model and the single-parameter anomaly detection threshold model are sent from the cloud server to the MEC. The optimization functions of the single-parameter state prediction model and the single-parameter anomaly detection threshold model are deployed in the cloud server. The single-parameter state prediction model and the single-parameter anomaly detection threshold model in the MEC are the latest optimized by the cloud server.

[0210] Sensors and MEC, MEC and cloud servers are connected through 5G private networks. MEC is a multi-access edge computing device, and 5G is the fifth generation of mobile communication technology.

[0211] In one embodiment, the numerical module 1 specifically includes:

[0212] A receiving unit, used for periodically receiving the real-time value of each parameter;

[0213] The duration unit is connected to the receiving unit and is used to obtain a time length T=n·Δt, where n represents the number of received real-time values ​​of each parameter within T, and Δt represents a receiving period;

[0214] The data unit is connected to the duration unit to obtain the time from the current time t 0 The n x in the previous T i (t ′ ), i = [1, 2, ..., m], indicating the i-th parameter, m indicating that there are m parameters in total, t′ = [t 0 ,t 0 -Δt,t 0 -2·Δt,…,t 0 -n·Δt], which represents n past moments.

[0215] In one embodiment, the rate of change module 2 is specifically used for:

[0216] Get the real-time change rate of each parameter according to the following formula:

[0217]

[0218] ΔD i (t 0 ) indicates that the i-th parameter is accumulated within T to the current time t 0 The real-time rate of change.

[0219] In one embodiment, the weight module 3 is specifically used for:

[0220] The real-time weight of each parameter used to predict equipment failure is obtained according to the following formula:

[0221]

[0222] w i (r 0 ) represents the i-th parameter at the current time t 0 The real-time weight.

[0223] In one embodiment, the comprehensive module 4 specifically includes:

[0224] The prediction unit is used to obtain the single parameter state prediction value P of each parameter according to the single parameter state prediction model. i (t), according to P i (t) and the real-time weight w of each parameter i (t 0 ) Obtain the comprehensive status prediction value of the device based on multiple parameters:

[0225]

[0226] The threshold unit is used to obtain the single parameter anomaly detection threshold θ for each parameter according to the single parameter anomaly detection threshold model. i (t), according to θ i (t) and the real-time weight w of each parameter i (t 0 ) Get the device's comprehensive anomaly detection threshold based on multiple parameters:

[0227]

[0228] Among them, i represents the i-th parameter, t represents the prediction time, and t 0 represents the current moment, and m represents the total number of types of various parameters.

[0229] In one embodiment, the threshold unit specifically includes:

[0230] The single parameter threshold unit is used to obtain the single parameter anomaly detection threshold for each parameter according to the following formula:

[0231] θ i (t)=(μ i +k 1 ·σ i +k 2 ΔD i (t 0 )+k 3 ·U i (t 0 ))

[0232] Among them, μ i represents the historical mean of the i-th parameter, σ i represents the historical standard deviation of the i-th parameter, U i (t 0 ) represents the current uncertainty of the single parameter state prediction model of the i-th parameter, k 1 , k 2 , k 3 There are three adjustment coefficients respectively.

[0233] In one embodiment, the device further comprises:

[0234] The maintenance feedback unit is connected to the early warning module 5 and is used to respond to P final (t)≥θ(t), send equipment failure warning information to equipment maintenance personnel, and receive feedback information from equipment maintenance personnel after maintaining the equipment;

[0235] The model optimization unit is connected to the maintenance feedback unit, and is used to optimize the single parameter state prediction model and the single parameter anomaly detection threshold model in response to the feedback information showing that the equipment fault warning information is inaccurate, and obtain the uncertainty U of the single parameter state prediction model of each parameter after optimization. i (t 0 ), uncertainty U i (t 0 ) is expressed using prediction error or confidence.

[0236] Embodiment 3:

[0237] Embodiment 3 of the present disclosure provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the fault prediction method based on 5G MEC as described in Embodiment 1 is implemented, or the fault prediction device based on 5G MEC as described in Embodiment 2 is implemented.

[0238] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program elements or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable read only memory), flash memory or other memory technology, CD-ROM (Compact Disc Read-Only Memory), digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer.

[0239] In addition, the present disclosure may also provide a computer device, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor runs the computer program stored in the memory, the processor executes the 5G MEC-based fault prediction method as described in Example 1. The computer device may be a 5G MEC-based fault prediction device as described in Example 2.

[0240] The memory is connected to the processor, the memory may be a flash memory or a read-only memory or other memory, and the processor may be a central processing unit or a single-chip microcomputer.

[0241] Embodiments 1-3 of the present disclosure provide a fault prediction method based on 5G MEC, a fault prediction device based on 5G MEC, and a computer-readable storage medium, which obtain multiple parameter values ​​of the device in real time, calculate the real-time change rate of each parameter, and dynamically adjust the weight of each parameter for predicting device failures, thereby obtaining a comprehensive state prediction value and a comprehensive anomaly detection threshold that integrate the different degrees of influence of multiple parameters, so as to more accurately evaluate the device state and capture device anomalies, thereby improving the accuracy of fault prediction.

[0242] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present disclosure, but the present disclosure is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and substance of the present disclosure, and these modifications and improvements are also considered to be within the scope of protection of the present disclosure.

Claims

1. A fault prediction method based on 5G MEC, characterized in that: The method comprises: Continuously obtain real-time values ​​of various parameters of the equipment; Based on the continuous real-time value, obtain the real-time change rate of each parameter; According to the real-time change rate of each parameter, a real-time weight of each parameter used to predict equipment failure is obtained; According to the real-time weight of each parameter, the comprehensive state prediction value and comprehensive anomaly detection threshold of the device based on multiple parameters are obtained; Obtain equipment failure warning information based on the comprehensive state prediction value and comprehensive anomaly detection threshold.

2. The method according to claim 1, characterized in that in: The real-time values ​​of various parameters of the equipment are continuously received by the MEC from sensors. The sensors are deployed to collect the real-time values ​​of various parameters of industrial equipment. The period for collecting the real-time values ​​of the parameters is set to Δt; The real-time change rate, real-time weight, comprehensive state prediction value, comprehensive anomaly detection threshold and equipment failure warning information are all calculated and obtained by MEC. MEC is deployed at the factory where the industrial equipment is located. A single-parameter state prediction model and a single-parameter anomaly detection threshold model are also deployed in MEC to obtain the state prediction value and anomaly detection threshold of each parameter. The single-parameter state prediction model and the single-parameter anomaly detection threshold model are sent from the cloud server to the MEC. The optimization functions of the single-parameter state prediction model and the single-parameter anomaly detection threshold model are deployed in the cloud server. The single-parameter state prediction model and the single-parameter anomaly detection threshold model in the MEC are the latest optimized by the cloud server. Sensors and MEC, MEC and cloud servers are connected through 5G private networks. MEC is a multi-access edge computing device, and 5G is the fifth generation of mobile communication technology.

3. The method according to claim 1, characterized in that Continuously obtain the real-time values ​​of various parameters of the device, including: Periodically receive the real-time value of each parameter; The acquisition time length T = n·Δt, where n represents the number of received real-time values ​​of each parameter within T, and Δt represents the receiving period; Get n x within T from the current time t0 i (t′), i=[1,2,…,m], represents the i-th parameter, m represents a total of m parameters, t′=[t0,t0-Δt,t0-2·Δt,…,t0-n·Δt], represents n past moments.

4. The method according to claim 3, characterized in that Based on the continuous real-time value, obtain the real-time change rate of each parameter, including: Get the real-time change rate of each parameter according to the following formula: ΔD i (t0) represents the real-time change rate of the i-th parameter accumulated within T to the current time t0.

5. The method according to claim 4, characterized in that According to the real-time change rate of each parameter, the real-time weight of each parameter used to predict equipment failure is obtained, including: The real-time weight of each parameter used to predict equipment failure is obtained according to the following formula: w i (t0) represents the real-time weight of the i-th parameter at the current time t0.

6. The method according to any one of claims 1 to 5, characterized in that: According to the real-time weight of each parameter, the comprehensive state prediction value and comprehensive anomaly detection threshold of the device based on multiple parameters are obtained, including: According to the single parameter state prediction model, the single parameter state prediction value P of each parameter is obtained. i (t), according to P i (t) and the real-time weight w of each parameter i (t0) Obtain the comprehensive status prediction value of the equipment based on multiple parameters: According to the single parameter anomaly detection threshold model, the single parameter anomaly detection threshold θ of each parameter is obtained. i (t), according to θ i (t) and the real-time weight w of each parameter i (t0) Obtain the comprehensive anomaly detection threshold of the device based on multiple parameters: Among them, i represents the i-th parameter, t represents the prediction time, t0 represents the current time, and m represents the total number of types of multiple parameters.

7. The method according to claim 6, characterized in that Get the single parameter anomaly detection threshold θ for each parameter i (t), including: The single parameter anomaly detection threshold for each parameter is obtained according to the following formula: i i (t)=(μ i +k1·s i +k2·ΔD i (t0)+k3·U i (t0)) Among them, μ i represents the historical mean of the i-th parameter, σ i represents the historical standard deviation of the i-th parameter, U i (t0) represents the current uncertainty of the single parameter state prediction model of the i-th parameter, and k1, k2, and k3 are three adjustment coefficients respectively.

8. The method according to claim 7, characterized in that The method further comprises: In response to P final (t)≥θ(t), send equipment failure warning information to equipment maintenance personnel, and receive feedback information from equipment maintenance personnel after maintaining the equipment; In response to the feedback information showing that the equipment fault warning information is inaccurate, the single parameter state prediction model and the single parameter anomaly detection threshold model are optimized, and the uncertainty U of the optimized single parameter state prediction model of each parameter is obtained. i (t0), uncertainty U i (t0) is expressed using prediction error or confidence.

9. A fault prediction device based on 5G MEC, characterized in that: The device comprises: Numerical module, used to continuously obtain the real-time values ​​of various parameters of the equipment; A change rate module, connected to the value module, is used to obtain the real-time change rate of each parameter based on the continuous real-time value; A weight module, connected to the change rate module, for obtaining a real-time weight of each parameter for predicting equipment failure according to the real-time change rate of each parameter; The comprehensive module is connected with the weight module and is used to obtain the comprehensive state prediction value and comprehensive abnormality detection threshold of the device based on multiple parameters according to the real-time weight of each parameter; The early warning module is connected to the comprehensive module and is used to obtain equipment failure early warning information based on the comprehensive state prediction value and the comprehensive anomaly detection threshold.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the 5G MEC-based fault prediction method as described in any one of claims 1 to 8 is implemented.

Images