Equipment failure prediction method and related equipment based on industrial Internet of Things

Abstract

The present invention relates to an equipment failure prediction method based on industrial Internet of Things and related equipment, comprising the following steps: collecting physical signals of a target device through a preset industrial Internet of Things to obtain an original sequence signal; identifying whether there is an abnormal signal in the original sequence signal to obtain an identification result; if the identification result is that there is an abnormal signal, performing a fault type analysis on the target device based on the abnormal signal to obtain a fault location and a fault type corresponding to the fault location; performing a fault transfer prediction on the target device based on the abnormal signal, the fault location and the fault type corresponding to the fault location to obtain a predicted fault transfer path; generating an early warning instruction based on the predicted fault transfer path, and performing a fault alarm based on the early warning instruction, thereby solving the technical problem that traditional equipment maintenance methods mostly adopt a method of regular inspection or post-fault repair, which is not only costly but also cannot effectively prevent the occurrence of sudden failures.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial Internet of Things, and in particular to an equipment failure prediction method based on industrial Internet of Things and related equipment. Background Art

[0002] With the continuous advancement of industrialization, the manufacturing industry is increasingly dependent on production equipment. However, various failures are inevitable in the long-term operation of equipment, which not only affect production efficiency, but also may lead to a decline in product quality and even production safety accidents. Traditional equipment maintenance methods mostly use regular inspections or post-failure repairs, which is not only costly, but also unable to effectively prevent sudden failures, causing huge economic losses to enterprises.

[0003] In recent years, with the development of IoT technology, the Industrial Internet of Things (IIoT) has gradually become an important means to improve equipment management efficiency and reduce maintenance costs. By deploying a large number of sensors and data acquisition devices, the Industrial Internet of Things can monitor the operating status of equipment in real time and collect a large amount of equipment operation data. However, faced with such a huge amount of data, how to efficiently and accurately extract valuable information from it, especially to detect and predict potential equipment failures at an early stage, has become a difficulty in current research. Although some existing fault prediction models can achieve early warning of faults to a certain extent, they often require a large amount of historical data as a training basis. For new equipment or when there is a lack of sufficient historical data, the performance of these models is often unsatisfactory.

[0004] In addition, even when sufficient data is available, existing fault prediction methods face problems such as low recognition accuracy and high false alarm rate. This is mainly because the manifestations of equipment failures are complex and diverse. Different types of failures may have similar characteristic manifestations, and the same type of failure may also exhibit different characteristics under different working conditions. Therefore, developing a method that can adapt to a variety of working conditions, accurately identify equipment failure types and predict failure development trends is of great significance to improving the reliability and safety of equipment. The above-mentioned equipment failure prediction method based on industrial Internet of Things is designed to solve these problems. It combines anomaly detection algorithms with fault type analysis to realize an integrated process from data collection to fault warning, providing a new solution for the intelligent management of industrial equipment. Summary of the invention

[0005] The main purpose of the present invention is to provide an equipment failure prediction method and related equipment based on the industrial Internet of Things, which solves the technical problem that traditional equipment maintenance methods mostly use regular inspections or post-failure repairs, which is not only costly but also unable to effectively prevent sudden failures.

[0006] To achieve the above object, the present invention provides an equipment failure prediction method based on industrial Internet of Things, comprising the following steps:

[0007] The physical signal of the target device is collected through the preset industrial Internet of Things to obtain the original sequence signal;

[0008] Identify whether there is an abnormal signal in the original sequence signal by using a preset abnormality detection algorithm to obtain an identification result;

[0009] If the identification result is that an abnormal signal exists, a fault type analysis is performed on the target device based on the abnormal signal to obtain a fault location and a fault type corresponding to the fault location;

[0010] Performing a fault transfer prediction on the target device based on the abnormal signal, the fault location and the fault type corresponding to the fault location to obtain a predicted fault transfer path;

[0011] A warning instruction is generated based on the predicted fault transfer path, and a fault alarm is issued based on the warning instruction.

[0012] Furthermore, the physical signal collection of the target device through the preset industrial Internet of Things to obtain the original sequence signal includes:

[0013] Physical signals of the device are collected through multiple types of sensors deployed on the target device to obtain preliminary collected data;

[0014] Performing denoising processing on the preliminary collected data to obtain denoised preliminary collected data;

[0015] The denoised preliminary collected data is segmented by a preset sliding window mechanism to generate data segments within a plurality of time windows, wherein each of the data segments contains a continuous signal within a certain time period;

[0016] Using wavelet transform technology, each of the data segments is decomposed into multiple scales to extract signal features in different frequency ranges to obtain multi-scale feature signals;

[0017] Performing dimensionality reduction processing on the multi-scale feature signal to extract the most representative feature signal to obtain a feature signal after dimensionality reduction;

[0018] The characteristic signals after dimension reduction are combined in time sequence to obtain original sequence signals; wherein the original sequence signals include vibration signals, temperature signals and electrical signals.

[0019] Furthermore, the identifying whether there is an abnormal signal in the original sequence signal by using a preset abnormality detection algorithm to obtain an identification result includes:

[0020] The original sequence signal is subjected to feature compression and reconstruction by a preset deep learning autoencoder to obtain a compressed feature vector;

[0021] Performing radius format conversion on the compressed feature vector using a preset hypersphere description model to obtain a conversion format radius;

[0022] Coinciding one end of the conversion format radius with the center of the ball in the hypersphere description model to obtain a center point coincidence radius; wherein the center point coincidence radius is a conversion format radius that coincides with the center of the ball;

[0023] Using a preset anomaly detection algorithm, identifying whether the other end of the center point coincidence radius reaches the surface boundary of the hypersphere description model;

[0024] If so, the original sequence signal corresponding to the center point coincidence radius that does not reach the surface boundary of the hypersphere description model is marked as abnormal to obtain a recognition result.

[0025] Further, the performing of a fault type analysis on the target device based on the abnormal signal to obtain a fault location and a fault type corresponding to the fault location includes:

[0026] Performing signal layering processing on the abnormal signal through a preset hidden Markov model to obtain a layered abnormal signal;

[0027] Performing physical structure mapping on the target device based on the layered abnormal signal to obtain a signal-structure mapping relationship;

[0028] By using a preset Bayesian network, based on the signal-structure mapping relationship, the probability of the fault propagation path of the target device is analyzed to obtain a fault propagation probability distribution;

[0029] Using a preset density space clustering algorithm, hot spot areas of the target device are identified based on the fault propagation probability distribution to obtain fault hot spot areas;

[0030] Determine the fault location of the target device based on the fault hotspot area; wherein the fault location is the location with the highest probability of fault propagation in the fault hotspot area;

[0031] In the fault hot spot area corresponding to the fault location, the fault type of the target device is judged based on the abnormal signal to obtain the fault type corresponding to the fault location.

[0032] Further, the performing a failover prediction on the target device based on the abnormal signal, the fault location and the fault type corresponding to the fault location to obtain a predicted failover path includes:

[0033] By using a preset fault tree analysis method, a fault logic structure of the target device is constructed based on the fault location and fault type;

[0034] Calling historical fault data in the database, and performing fault probability condition analysis on the fault logic structure based on the historical fault data to obtain a fault condition probability table; wherein the fault condition probability table includes a fault type identifier, a historical fault count, and a fault impact factor, and the fault type identifier includes a node identifier and a parent node identifier;

[0035] By using a preset Markov chain technology, a fault state transition analysis is performed on the fault type based on the fault condition probability table to obtain a state combination space and a fault transition probability;

[0036] A dynamic programming algorithm is used to predict the failure transfer of the target device based on the abnormal signal, the state combination space and the failure transfer probability to obtain a predicted failure transfer path.

[0037] Furthermore, the fault tree analysis method is used to construct a fault logic structure of the target device based on the fault location and the fault type, including:

[0038] By using the hierarchical analysis method, the fault location is divided into regions to obtain a fault region hierarchy structure; wherein the fault region hierarchy structure includes fault region boundaries, connection relationships between fault region boundaries, and weights of fault region boundaries;

[0039] By using a preset fault tree analysis method, the fault factor is determined based on the fault area hierarchy and the fault type, and the minimum cut set analysis is performed on the fault factor to obtain a set of fault basic events; wherein the set of fault basic events includes fault basic events and fault basic event combinations;

[0040] Performing causal correlation analysis on the set of basic fault events to obtain a causal chain of fault events;

[0041] Through Boolean algebra operations, based on the causal chain of the fault events, logic gate analysis is performed on the combination of the fault basic events to obtain a fault event logic relationship table;

[0042] The fault event logic relationship table is dynamically modeled by a preset petri net modeling method to obtain a fault propagation network structure; wherein the fault propagation network structure includes a place node and a transition node;

[0043] Based on the fault propagation network structure, a fault logic structure is obtained.

[0044] Further, generating a warning instruction based on the predicted fault transfer path, and performing a fault alarm based on the warning instruction, includes:

[0045] Based on the predicted fault transfer path, a quantitative assessment is performed on the risk level in the fault propagation process to obtain a risk assessment value sequence;

[0046] Determining a warning level according to the risk assessment value sequence, and constructing warning information content according to the warning level and the predicted fault transfer path;

[0047] Based on the message queue technology, a transmission queue is constructed for the warning information content to obtain a warning information queue to be sent;

[0048] Based on a multi-channel communication protocol, the early warning information queue is sent to a receiving end of the target device to obtain an early warning instruction of sending completion;

[0049] By using visualization technology, a fault alarm is issued at the receiving end of the target device according to the early warning instruction.

[0050] The present invention also provides an equipment failure prediction device based on industrial Internet of Things, comprising:

[0051] The acquisition module is used to collect physical signals from the target device through the preset industrial Internet of Things to obtain the original sequence signal;

[0052] An identification module is used to identify whether there is an abnormal signal in the original sequence signal through a preset abnormality detection algorithm to obtain an identification result;

[0053] An analysis module, configured to, if the identification result indicates that an abnormal signal exists, analyze the fault type of the target device based on the abnormal signal to obtain a fault location and a fault type corresponding to the fault location;

[0054] A prediction module, configured to predict a fault transfer path for the target device based on the abnormal signal, the fault location and the fault type corresponding to the fault location, and obtain a predicted fault transfer path;

[0055] A generation module is used to generate an early warning instruction based on the predicted fault transfer path, and to issue a fault alarm based on the early warning instruction.

[0056] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0057] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0058] The device fault prediction method based on industrial Internet of Things provided by the present invention comprises the following steps: physical signal collection of target device through preset industrial Internet of Things to obtain original sequence signal; identifying whether there is abnormal signal in the original sequence signal through preset abnormal detection algorithm to obtain identification result; if the identification result is that there is abnormal signal, then fault type analysis is performed on the target device based on the abnormal signal to obtain fault location and fault type corresponding to the fault location; fault transfer prediction is performed on the target device based on the abnormal signal, fault location and fault type corresponding to the fault location to obtain predicted fault transfer path; early warning instruction is generated based on the predicted fault transfer path, and fault alarm is performed based on the early warning instruction. Through the above technical means, the traditional equipment maintenance method mostly adopts the method of regular inspection or post-fault maintenance, which is not only costly but also cannot effectively prevent the occurrence of sudden faults. The technical problem is solved, and the predicted fault transfer path is realized. The method can generate early warning instruction and issue fault alarm accordingly. This real-time early warning mechanism enables operators to take necessary preventive measures before the fault occurs, effectively prevents the occurrence of accidents, and ensures the beneficial effect of production safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 This is a schematic diagram of the steps of a method for predicting equipment failure based on the industrial Internet of Things in one embodiment of the present invention;

[0060] Figure 2 It is a structural block diagram of an equipment failure prediction device based on industrial Internet of Things in one embodiment of the present invention;

[0061] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0062] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0064] like Figure 1 As shown, Figure 1 This is a schematic diagram of the steps of a method for predicting equipment failure based on industrial Internet of Things in one embodiment of the present invention;

[0065] In one embodiment of the present invention, a method for predicting equipment failure based on industrial Internet of Things is provided, comprising the following steps:

[0066] Step S1, collect physical signals of the target device through the preset industrial Internet of Things to obtain the original sequence signal.

[0067] Specifically, when implementing the above-mentioned step of "collecting physical signals from the target device through the preset industrial Internet of Things to obtain the original sequence signal", it is first necessary to install a series of sensors at the key parts of the target device. These sensors can monitor various physical parameters during the operation of the device, such as temperature, vibration, pressure, etc. The selection and arrangement of sensors should be determined according to the specific type of equipment and the working environment to ensure that the main operating status of the equipment can be fully covered. For example, in a large manufacturing machine, temperature sensors and vibration sensors may be installed on key components such as motors, bearings, and gearboxes to facilitate real-time monitoring of the working conditions of these components. When these sensors are activated, they will continuously send the collected data to the industrial Internet of Things platform. As the hub of the entire system, the industrial Internet of Things platform is responsible for receiving the data streams from each sensor and performing preliminary processing on them, such as filtering and denoising, format conversion, etc., to ensure the accuracy of subsequent processing. These processed data constitute the so-called "original sequence signal". In practical applications, such as in the production line of an automobile manufacturer, each assembly robot is an important target device. By installing vibration sensors at its joints, the robot's motion status can be monitored in real time. Once abnormal vibration is detected, the system can respond quickly and take timely measures to avoid the occurrence and development of potential faults, thereby ensuring the stable operation of the production line. In this way, the Industrial Internet of Things can not only help factory managers understand the health status of equipment in real time, but also provide reliable data support for fault prediction.

[0068] Step S2, identifying whether there is an abnormal signal in the original sequence signal through a preset abnormality detection algorithm to obtain an identification result.

[0069] Specifically, we first need to select a suitable anomaly detection algorithm. These algorithms are usually built based on methods such as statistics, machine learning, or deep learning to identify abnormal patterns in data. In the industrial Internet of Things environment, the original sequence signal contains various physical parameters of the equipment operation, such as temperature, vibration, pressure, etc., which follow certain distribution rules under normal circumstances. When the equipment fails, these parameters often deviate from the normal range and generate abnormal signals. In actual operation, the preset anomaly detection algorithm will first pre-process the collected original sequence signals, including but not limited to data cleaning, normalization, and dimensionality reduction, to remove noise and irrelevant information, so that the data is more suitable for algorithm processing. Then, the algorithm will analyze the processed data to find data points or data sequences that do not conform to the normal pattern. For example, on the production line of an automobile manufacturer, if the data sent back by the vibration sensor at a joint of the assembly robot suddenly shows an abnormally high vibration level, the preset anomaly detection algorithm can identify this change and mark it as an abnormal signal. At this time, the system will generate a recognition result indicating that there is an abnormal signal, which will trigger the subsequent fault type analysis and fault transfer prediction process. This method of automatically identifying abnormal signals through a preset abnormal detection algorithm can not only quickly and accurately discover potential equipment problems, but also reduce the workload of manual inspections, improve fault response speed, and ensure the smooth operation of the production process. At the same time, with the continuous optimization and training of the algorithm, its accuracy and sensitivity in identifying abnormal signals will continue to improve, providing strong technical support for the intelligent maintenance and fault prevention of industrial equipment.

[0070] Step S3: If the identification result shows that an abnormal signal exists, a fault type analysis is performed on the target device based on the abnormal signal to obtain a fault location and a fault type corresponding to the fault location.

[0071] Specifically, if the identification result is that there is an abnormal signal, the target device is subjected to a fault type analysis based on the abnormal signal to obtain the fault location and the fault type corresponding to the fault location. This process involves an in-depth analysis of the abnormal signal, the purpose of which is to determine the specific fault cause behind the abnormal signal and the precise location of the fault. First, the system uses a preset fault diagnosis model to further process the identified abnormal signal. These models may be rule bases built based on expert knowledge, or machine learning models trained with a large amount of historical data, or even deep learning networks. Regardless of which model is used, its core task is to extract features from the abnormal signal and match these features with known fault modes to determine the fault type. For example, on the production line of an automobile manufacturing plant, if a vibration sensor at a joint of an assembly robot detects an abnormally high vibration level, and the anomaly detection algorithm has confirmed the existence of this signal, then the fault type analysis module will intervene next. This module analyzes the spectral characteristics, amplitude changes and other characteristics of the vibration signal, combines the historical operation data and maintenance records of the robot, and compares the various fault modes stored in the database to finally determine which specific component caused the abnormal vibration, such as bearing wear or loose gears. At the same time, the system will also give the specific location of the fault, such as "the second joint of the left arm of the assembly robot." This method can not only quickly locate the source of the fault, but also accurately determine the type of fault, providing clear guidance for subsequent repairs and maintenance. For example, maintenance personnel can go directly to the designated robot joint for inspection and repair based on the fault location and fault type information provided by the system, without having to conduct a comprehensive inspection of the entire equipment, which greatly improves maintenance efficiency and shortens fault handling time. In addition, this precise fault type analysis can also help to formulate more targeted preventive maintenance plans, further improving equipment reliability and production efficiency.

[0072] Step S4: performing a fault transfer prediction on the target device based on the abnormal signal, the fault location and the fault type corresponding to the fault location to obtain a predicted fault transfer path.

[0073] Specifically, after the system identifies an abnormal signal through an anomaly detection algorithm and determines the specific location and type of the fault through fault type analysis, it is necessary to use this information to predict fault transfer. Fault transfer prediction mainly relies on preset fault propagation models. These models can be built through historical data and expert experience, or they can be learned and optimized from a large number of fault cases through machine learning and deep learning technologies. During the prediction process, the system will consider multiple factors, such as the working conditions of the equipment, environmental factors, historical operation data of the equipment, etc., and input these data into the fault propagation model, and the model will output the possible development path of the fault. For example, on the production line of an automobile manufacturing plant, if the vibration sensor at a joint of the assembly robot detects an abnormal signal, and the fault type analysis determines that the fault is caused by bearing wear, then the fault transfer prediction module will further analyze whether this wear will trigger a chain reaction of other components, such as whether it will cause gear damage or motor overheating. To achieve this, the fault transfer prediction model will combine the structural characteristics and operating parameters of the equipment to simulate the propagation path of the fault under different conditions. For example, the model may predict that bearing wear will cause poor gear meshing after continuing to run for a period of time, which will cause greater vibration and noise, and may eventually lead to the failure of the entire transmission system. Through this prediction, the system can generate early warning information in advance to inform operators of possible fault development trends and remind them to take corresponding preventive measures, such as adjusting the operating parameters of the equipment, increasing the lubrication frequency, or arranging early maintenance. Fault transfer prediction can not only help factory managers better understand the health of the equipment, but also provide an important reference for formulating scientific and reasonable maintenance plans. For example, the maintenance team can prioritize key components that may cause larger failures based on the predicted fault transfer path, thereby avoiding unplanned downtime of the production line and improving production efficiency and equipment reliability. This forward-looking fault management method can significantly reduce the maintenance cost of equipment, extend the service life of equipment, and bring more economic benefits to the enterprise.

[0074] Step S5: generating a warning instruction based on the predicted fault transfer path, and performing a fault alarm based on the warning instruction.

[0075] Specifically, an early warning instruction is generated based on the predicted fault transfer path, and a fault alarm is performed based on the early warning instruction. This process is to convert the results of fault prediction into a specific action guide to ensure that timely measures are taken to prevent the fault from further spreading and worsening. Specifically, when the system obtains the possible development path of the fault through the fault transfer prediction model, the next step is to generate an early warning instruction based on these prediction results. The generation of the early warning instruction is based on preset rules and thresholds. The system will evaluate the severity and urgency of the fault development based on the predicted fault transfer path. For example, if the prediction results show that a certain fault may cause the entire system to shut down in a short period of time, the system will generate a high-priority early warning instruction; if the fault develops slowly, the system will generate a lower-priority early warning instruction. These early warning instructions usually include the specific location of the fault, the type of fault, the expected development trend, and the recommended countermeasures. After the early warning instructions are generated, the system will immediately send these instructions to the relevant operators or maintenance teams. These instructions can be conveyed in a variety of ways, such as being sent to the operator's mobile device through the industrial Internet of Things platform, displayed on the monitoring screen in the control room, or sounding an alarm. For example, on the production line of an automobile manufacturing plant, if the system predicts that bearing wear at a joint of an assembly robot may cause gear damage, the system will generate an early warning instruction, which may be as follows: "Warning: The bearing of the second joint of the left arm of the assembly robot is worn, which is expected to cause gear damage in the next 24 hours. Please check and replace it immediately." The operator or maintenance team who receives the early warning instruction can take corresponding measures according to the suggestions in the instruction. For example, they can suspend the operation of related equipment, conduct detailed inspection and repair, or adjust the operating parameters of the equipment to slow down the development of the fault. This timely fault alarm mechanism can not only help factory managers respond quickly to potential fault risks, but also effectively avoid further spread of faults, reduce production interruption time and maintenance costs. Through this method of generating early warning instructions based on the predicted fault transfer path and performing fault alarm, the factory can achieve active management and prevention of equipment failures, improve equipment reliability and production efficiency, and ensure production continuity and safety. This intelligent fault management method not only improves the management level of equipment, but also brings significant economic and social benefits to enterprises.

[0076] In a specific embodiment, the physical signal acquisition of the target device through the preset industrial Internet of Things to obtain the original sequence signal includes:

[0077] Physical signals of the device are collected through multiple types of sensors deployed on the target device to obtain preliminary collected data;

[0078] Performing denoising processing on the preliminary collected data to obtain denoised preliminary collected data;

[0079] The denoised preliminary collected data is segmented by a preset sliding window mechanism to generate data segments within a plurality of time windows, wherein each of the data segments contains a continuous signal within a certain time period;

[0080] Using wavelet transform technology, each of the data segments is decomposed into multiple scales to extract signal features in different frequency ranges to obtain multi-scale feature signals;

[0081] Performing dimensionality reduction processing on the multi-scale feature signal to extract the most representative feature signal to obtain a feature signal after dimensionality reduction;

[0082] The characteristic signals after dimension reduction are combined in time sequence to obtain original sequence signals; wherein the original sequence signals include vibration signals, temperature signals and electrical signals.

[0083] Specifically, it is necessary to first install various types of sensors on the key parts of the target equipment. These sensors can monitor various physical parameters during the operation of the equipment, such as vibration, temperature, and electrical signals. For example, on the production line of an automobile manufacturing plant, the assembly robot is an important target device. Vibration sensors, temperature sensors, and current sensors can be installed on its key components such as motors, bearings, and gearboxes to monitor the working conditions of these components in real time. These sensors will continuously send the collected data to the industrial Internet of Things platform to form preliminary collected data. The preliminary collected data usually contains a lot of noise and interference information, which will seriously affect the accuracy of subsequent analysis. Therefore, it is necessary to denoise the preliminary collected data. Denoising can be achieved through a variety of methods, such as Fourier transform, wavelet transform, Kalman filtering, etc. In this method, a suitable wavelet transform technology can be selected to denoise the preliminary collected data, remove high-frequency noise and random interference, and retain useful signal components, thereby obtaining denoised preliminary collected data. Next, in order to better analyze and process the denoised preliminary collected data, it is necessary to segment the data through a preset sliding window mechanism. Sliding window is a commonly used data processing technology that divides the continuous data stream into multiple data segments in a fixed-length time window. Each data segment contains a continuous signal within a certain period of time, which can better capture the dynamic changes of the signal. For example, a 10-second time window can be set and slide every 5 seconds to generate multiple 10-second data segments. This method can ensure the continuity and integrity of the data, while facilitating subsequent feature extraction and analysis. After generating data segments in multiple time windows, it is necessary to use wavelet transform technology to perform multi-scale decomposition on each data segment. Wavelet transform is a powerful signal analysis tool that can decompose signals into sub-signals in different frequency ranges to extract multi-scale features of the signal. Through multi-scale decomposition, the changes of signals at different time scales can be observed more clearly, which is very useful for fault detection and diagnosis. For example, when analyzing the vibration signal of an assembly robot, the slow-changing signal in the low-frequency part and the fast-changing signal in the high-frequency part can be extracted through wavelet transform, so as to better understand the characteristics of the vibration and potential fault modes. After extracting the multi-scale feature signal, in order to reduce the data dimension and improve the computational efficiency, the multi-scale feature signal needs to be reduced in dimension. Dimensionality reduction can be achieved through methods such as principal component analysis (PCA) and linear discriminant analysis (LDA). These methods can extract the most representative feature signals, remove redundant information, and retain key features. For example, the PCA method can be used to reduce the dimension of multi-scale feature signals and extract the first few principal components, which contain the main features of the signal and can be used for subsequent fault detection and analysis. Finally, the feature signals after dimensionality reduction are combined in chronological order to form the original sequence signal.The original sequence signal is a time series containing multiple characteristic signals, including vibration signals, temperature signals, and electrical signals. In this way, different types of physical signals can be integrated together to form a complete time series data, providing reliable data support for subsequent anomaly detection, fault type analysis, and fault transfer prediction. For example, on the production line of an automobile manufacturer, the original sequence signal processed by the above steps can be transmitted to the central control system in real time, and the system can monitor and analyze these signals in real time. Once an abnormal signal is detected, the system will immediately generate an early warning instruction and notify the operator in a variety of ways, such as sending a text message, displaying on the monitoring screen in the control room, or sounding an alarm. The operator can take corresponding measures according to the suggestions in the early warning instruction, such as suspending the operation of the equipment, conducting inspections and repairs, or adjusting the operating parameters of the equipment, thereby effectively avoiding the occurrence and development of faults and ensuring the continuity and safety of production. In this way, the industrial Internet of Things can not only help factory managers to grasp the health status of equipment in real time, but also provide strong technical support for fault prediction and prevention, improve the reliability and production efficiency of equipment, reduce maintenance costs, and bring significant economic and social benefits to enterprises.

[0084] In a specific embodiment, the step of identifying whether an abnormal signal exists in the original sequence signal by using a preset abnormality detection algorithm to obtain an identification result includes:

[0085] The original sequence signal is subjected to feature compression and reconstruction by a preset deep learning autoencoder to obtain a compressed feature vector;

[0086] Performing radius format conversion on the compressed feature vector using a preset hypersphere description model to obtain a conversion format radius;

[0087] Coinciding one end of the conversion format radius with the center of the ball in the hypersphere description model to obtain a center point coincidence radius; wherein the center point coincidence radius is a conversion format radius that coincides with the center of the ball;

[0088] Using a preset anomaly detection algorithm, identifying whether the other end of the center point coincidence radius reaches the surface boundary of the hypersphere description model;

[0089] If so, the original sequence signal corresponding to the center point coincidence radius that does not reach the surface boundary of the hypersphere description model is marked as abnormal to obtain a recognition result.

[0090] Specifically, first, the original sequence signal needs to be feature compressed and reconstructed by a preset deep learning autoencoder to obtain a compressed feature vector. The deep learning autoencoder is an unsupervised learning model that can compress a high-dimensional original sequence signal into a low-dimensional feature vector while retaining the main features of the signal. For example, on the production line of an automobile manufacturing plant, the original sequence signals such as vibration signals, temperature signals and electrical signals of the assembly robot will be input into the autoencoder. The autoencoder compresses the high-dimensional original signal into a low-dimensional feature vector by learning the internal features of these signals. This process not only reduces the dimension of the data, but also extracts the key features of the signal, providing a basis for subsequent anomaly detection. Next, the compressed feature vector is converted into a radius format by a preset hypersphere description model to obtain a conversion format radius. The hypersphere description model is a method for anomaly detection, assuming that normal data points are distributed inside the hypersphere, while abnormal data points are located outside the hypersphere. In this process, the compressed feature vector will be converted into a radius value representing the distance, that is, the conversion format radius. This radius value reflects the distance from the data point to the center of the hypersphere. For example, a radius value can be obtained by calculating the Euclidean norm of the compressed feature vector, and this radius value represents the distance from the feature vector to the center of the hypersphere. Then, one end of the conversion format radius is overlapped with the center of the ball in the hypersphere description model to obtain the center point overlap radius. This means that the radius value of each data point is aligned with the center point of the hypersphere to form a vector starting from the center of the ball. In this way, the position of each data point relative to the center of the hypersphere can be intuitively seen. For example, assuming that the center point of the hypersphere is the origin (0,0,0), then the radius value of each data point can be regarded as a vector starting from the origin, and the length of this vector is the distance from the data point to the origin. Through the preset anomaly detection algorithm, it is identified whether the other end of the center point overlap radius reaches the surface boundary of the hypersphere description model. Specifically, the anomaly detection algorithm checks whether the radius value of each data point exceeds the preset hypersphere radius. If the radius value of a data point is less than or equal to the radius of the hypersphere, it means that the data point is located inside the hypersphere and is a normal data point; conversely, if the radius value of a data point is greater than the radius of the hypersphere, it means that the data point is located outside the hypersphere and is an abnormal data point. For example, on the production line of an automobile manufacturer, if the vibration signal of an assembly robot exceeds the preset hypersphere radius after the above processing, the system will mark the signal as an abnormal signal. If so, the original sequence signal corresponding to the center point coincidence radius that does not reach the surface boundary of the hypersphere description model is marked as abnormal to obtain the recognition result. This means that the system will record all data points whose radius values ​​are greater than the hypersphere radius and mark these data points as abnormal signals.For example, suppose that within a certain period of time, the radius values ​​of multiple data points in the vibration signal of the assembly robot exceed the preset hypersphere radius, then these data points will be marked as abnormal signals, and the system will generate corresponding warning information to notify the operator to conduct further inspection and processing.

[0091] Through this method, the industrial Internet of Things system can monitor the operating status of the equipment in real time, detect potential abnormal signals in time, and generate early warning information. For example, on the production line of an automobile manufacturer, once the system detects an abnormal vibration signal at a joint of an assembly robot, it will immediately generate an early warning instruction to notify the operator to check and repair, thereby avoiding further development of the fault and ensuring the continuity and safety of production. This anomaly detection method based on deep learning autoencoders and hypersphere description models can not only improve the accuracy of fault detection, but also reduce the false alarm rate, providing strong technical support for the intelligent management of industrial equipment.

[0092] In a specific embodiment, the performing fault type analysis on the target device based on the abnormal signal to obtain the fault location and the fault type corresponding to the fault location includes:

[0093] Performing signal layering processing on the abnormal signal through a preset hidden Markov model to obtain a layered abnormal signal;

[0094] Performing physical structure mapping on the target device based on the layered abnormal signal to obtain a signal-structure mapping relationship;

[0095] By using a preset Bayesian network, based on the signal-structure mapping relationship, the probability of the fault propagation path of the target device is analyzed to obtain a fault propagation probability distribution;

[0096] Using a preset density space clustering algorithm, hot spot areas of the target device are identified based on the fault propagation probability distribution to obtain fault hot spot areas;

[0097] Determine the fault location of the target device based on the fault hotspot area; wherein the fault location is the location with the highest probability of fault propagation in the fault hotspot area;

[0098] In the fault hot spot area corresponding to the fault location, the fault type of the target device is judged based on the abnormal signal to obtain the fault type corresponding to the fault location.

[0099] Specifically, firstly, the abnormal signal needs to be processed by signal stratification through the preset hidden Markov model to obtain the stratified abnormal signal. Hidden Markov model (HMM) is a statistical model that is often used to process time series data. In this method, HMM can decompose complex abnormal signals into multiple levels, each of which corresponds to different fault characteristics. For example, on the production line of an automobile manufacturing plant, if the vibration sensor at a joint of an assembly robot detects an abnormal signal, HMM can divide these abnormal signals into multiple levels, each of which may represent different fault characteristics, such as low-frequency vibration, high-frequency vibration, periodic fluctuation, etc. Next, the target device is mapped to the physical structure based on the stratified abnormal signal to obtain the signal-structure mapping relationship. Physical structure mapping refers to corresponding the stratified abnormal signal to the physical structure of the device and establishing the association relationship between the signal and the structure. For example, low-frequency vibration signals can be mapped to bearings, high-frequency vibration signals can be mapped to gears, and periodic fluctuation signals can be mapped to motors. In this way, abstract signal features can be linked to specific physical components, providing a basis for subsequent fault analysis. Then, the fault propagation path probability of the target device is analyzed based on the signal-structure mapping relationship through the preset Bayesian network to obtain the fault propagation probability distribution. The Bayesian network is a probabilistic graph model that can be used to represent the dependency relationship between variables. In this method, the Bayesian network can calculate the probability of fault propagation between different components based on the signal-structure mapping relationship. For example, if the vibration signal is mapped to the bearing, and the failure of the bearing may cause damage to the gear, the Bayesian network can calculate the fault propagation probability from the bearing to the gear. In this way, the propagation path of the fault inside the device can be predicted to provide a basis for fault location. Next, the hot spot area of ​​the target device is identified based on the fault propagation probability distribution through the preset density space clustering algorithm to obtain the fault hot spot area. The density space clustering algorithm (such as DBSCAN) can cluster data points into different clusters, each cluster representing a fault hot spot area. In this method, the area with a higher probability of fault propagation can be regarded as a fault hot spot area. For example, if the Bayesian network calculates that the fault propagation probability of a certain bearing is high, the density space clustering algorithm can identify the area where the bearing is located as a fault hot spot area. The fault location of the target device is determined based on the fault hotspot area, where the fault location is the location with the highest fault probability in the fault hotspot area. The fault hotspot areas identified by the density space clustering algorithm can be further analyzed to find the location with the highest fault probability. For example, if the area where a bearing is located is identified as a fault hotspot area, and the failure probability of the bearing in this area is the highest, then the bearing can be determined as the fault location. Finally, in the fault hotspot area corresponding to the fault location, the fault type of the target device is judged based on the abnormal signal to obtain the fault type corresponding to the fault location.Fault type judgment can be achieved through a preset fault mode library and machine learning model. For example, the layered abnormal signal can be compared with the known fault mode to find the most matching fault type. If the characteristics of the vibration signal are most similar to the characteristics of bearing wear, then it can be determined that the fault type of the bearing is wear. In this way, the industrial Internet of Things system can accurately locate the fault location and determine the fault type, providing strong support for the rapid diagnosis and processing of faults. For example, on the production line of an automobile manufacturing plant, once the system detects an abnormal vibration signal at a joint of an assembly robot, and determines that the bearing at the joint is the fault location and the fault type is wear through the above method, the system will generate a detailed fault report to notify the operator to inspect and repair. This fault type analysis method based on hidden Markov model, Bayesian network and density space clustering algorithm can not only improve the accuracy of fault diagnosis, but also reduce the false alarm rate, providing strong technical support for the intelligent management of industrial equipment.

[0100] In a specific embodiment, the performing a failover prediction on the target device based on the abnormal signal, the fault location, and the fault type corresponding to the fault location to obtain a predicted failover path includes:

[0101] By using a preset fault tree analysis method, a fault logic structure of the target device is constructed based on the fault location and fault type;

[0102] Calling historical fault data in the database, and performing fault probability condition analysis on the fault logic structure based on the historical fault data to obtain a fault condition probability table; wherein the fault condition probability table includes a fault type identifier, a historical fault count, and a fault impact factor, and the fault type identifier includes a node identifier and a parent node identifier;

[0103] By using a preset Markov chain technology, a fault state transition analysis is performed on the fault type based on the fault condition probability table to obtain a state combination space and a fault transition probability;

[0104] A dynamic programming algorithm is used to predict the failure transfer of the target device based on the abnormal signal, the state combination space and the failure transfer probability to obtain a predicted failure transfer path.

[0105] Specifically, firstly, it is necessary to construct the fault logic structure of the target device based on the fault location and fault type through the preset fault tree analysis method. Fault tree analysis (FTA) is a system safety analysis method that represents various possible causes of system failures and their logical relationships by constructing a fault tree. In this method, a detailed fault logic structure can be constructed based on the determined fault location and fault type. For example, on the production line of an automobile manufacturing plant, if the bearing at a joint of an assembly robot is determined to be the fault location and the fault type is wear, a fault tree can be constructed, in which the root node is "bearing wear" and the child nodes may include "insufficient lubrication", "overload operation", "material defects", etc. Each child node represents a possible cause of failure, which is connected through logic gates (such as AND gates, OR gates, etc.) to form a complete fault logic structure. Next, the historical fault data in the database is called, and the fault probability condition analysis of the fault logic structure is performed based on the historical fault data to obtain a fault condition probability table. Historical fault data can provide rich information to help us understand the probability of occurrence of different fault causes and their influencing factors. In this method, relevant data on bearing wear of assembly robots in the past few years can be retrieved from the database, including the specific cause, number of occurrences and influencing factors of each fault. Then, based on these data, the probability of occurrence of each fault cause is calculated, and a fault condition probability table is generated. The fault condition probability table includes the identification of the fault type, the number of historical faults and the fault influencing factor, and the identification of the fault type includes the node identification and the parent node identification. For example, the identification of the fault type "insufficient lubrication" can be "L1", and its parent node identification can be "bearing wear" (B1), the number of historical faults is 10 times, and the fault influencing factor is 0.8. Then, through the preset Markov chain technology, the fault state transition analysis is performed on the fault type based on the fault condition probability table to obtain the state combination space and fault transition probability. Markov chain is a method for modeling random processes, which can be used to analyze the transition probability of a system between different states. In this method, each fault type in the fault tree can be regarded as a state, and the data in the fault condition probability table can be used to calculate the probability of transitioning from one state to another. For example, the probability of transitioning from the "insufficient lubrication" state to the "bearing wear" state can be calculated, thereby constructing a state combination space and fault transition probability matrix. The state combination space represents all possible combinations of fault states, and the fault transition probability matrix represents the probability of transitioning from one state to another. Using a dynamic programming algorithm, the target device is predicted to perform a fault transition based on the abnormal signal, the state combination space, and the fault transition probability, and a predicted fault transition path is obtained. The dynamic programming algorithm is a method for solving multi-stage decision problems and can be used to find the optimal fault transition path. In this method, the fault transition process can be regarded as a multi-stage decision problem, and the state transition of each stage is affected by the abnormal signal.Through the dynamic programming algorithm, the optimal path from the current fault state to the possible future fault state can be searched in the state combination space to obtain the predicted fault transfer path. For example, assuming that the current state is "insufficient lubrication", the dynamic programming algorithm can search for the optimal path from "insufficient lubrication" to "bearing wear" and then to "gear damage" to predict the development trend of the fault. In this way, the industrial Internet of Things system can accurately predict the fault development process of the equipment, take preventive measures in advance, and avoid the further spread of the fault. For example, on the production line of an automobile manufacturing plant, once the system detects an abnormal vibration signal at a joint of the assembly robot, and constructs a detailed fault logic structure through the above method, calls historical fault data, and performs fault state transfer analysis, and finally obtains the predicted fault transfer path, the system can generate a detailed fault report and notify the operator to take corresponding preventive measures, such as increasing the lubrication frequency, adjusting the load, or replacing the bearing in advance, so as to avoid a larger fault and ensure the continuity and safety of production. This fault transfer prediction method based on fault tree analysis, Markov chain technology and dynamic programming algorithm can not only improve the accuracy of fault prediction, but also provide scientific decision support for equipment maintenance and management.

[0106] In a specific embodiment, the fault tree analysis method is used to construct a fault logic structure of the target device based on the fault location and the fault type, including:

[0107] By using the hierarchical analysis method, the fault location is divided into regions to obtain a fault region hierarchy structure; wherein the fault region hierarchy structure includes fault region boundaries, connection relationships between fault region boundaries, and weights of fault region boundaries;

[0108] By using a preset fault tree analysis method, the fault factor is determined based on the fault area hierarchy and the fault type, and the minimum cut set analysis is performed on the fault factor to obtain a set of fault basic events; wherein the set of fault basic events includes fault basic events and fault basic event combinations;

[0109] Performing causal correlation analysis on the set of basic fault events to obtain a causal chain of fault events;

[0110] Through Boolean algebra operations, based on the causal chain of the fault events, logic gate analysis is performed on the combination of the fault basic events to obtain a fault event logic relationship table;

[0111] The fault event logic relationship table is dynamically modeled by a preset petri net modeling method to obtain a fault propagation network structure; wherein the fault propagation network structure includes a place node and a transition node;

[0112] Based on the fault propagation network structure, a fault logic structure is obtained.

[0113] Specifically, firstly, the fault location needs to be divided into regions by the analytic hierarchy process (AHP) to obtain the fault region hierarchy. The analytic hierarchy process is a multi-criteria decision analysis method that can be used to decompose complex decision problems into multiple levels, each level containing several elements. In this method, the target device can be divided into multiple fault regions by the analytic hierarchy process, and the boundary, connection relationship and weight of each fault region can be determined. For example, on the production line of an automobile manufacturing plant, if the bearing at a joint of an assembly robot is determined as the fault location, the joint can be divided into multiple fault regions, such as a bearing region, a gear region and a motor region. The boundary, connection relationship and weight of each region can be determined by expert experience and historical data. Next, the fault factor is determined based on the fault region hierarchy and the fault type by the preset fault tree analysis method (FTA), and the minimum cut set analysis is performed on the fault factor to obtain the set of fault basic events. The fault tree analysis method is a technology used for system safety analysis, which represents various possible causes of system failures and their logical relationships by constructing a fault tree. In this method, the basic event (fault factor) causing the failure can be determined according to the fault region hierarchy and the fault type. For example, for bearing wear faults, it can be determined that the fault factors include "insufficient lubrication", "overload operation" and "material defects". Then, through the minimum cut set analysis, the minimum combination that causes the fault to occur is found to obtain the set of fault basic events. Minimum cut set analysis can help identify which combination of fault factors is most likely to cause the fault. The set of fault basic events is subjected to causal association analysis to obtain the causal chain of fault events. Causal association analysis refers to constructing the causal chain of fault events by analyzing the causal relationship between fault basic events. For example, it can be analyzed whether "insufficient lubrication" will lead to "overload operation" and whether "overload operation" will lead to "material defects". Through this causal relationship analysis, the process and mechanism of the fault occurrence can be more clearly understood. Through Boolean algebra operations, based on the causal chain of the fault event, the combination of the fault basic events is subjected to logic gate analysis to obtain a fault event logical relationship table. Boolean algebra operations are basic mathematical tools for logical operations, and the logical relationship between the fault basic events can be represented by logic gates (such as AND gates, OR gates, NOT gates, etc.). In this method, a logic gate model can be constructed according to the causal chain of the fault event to generate a fault event logical relationship table. For example, a logical expression can be constructed to indicate that the combined effect of "insufficient lubrication" and "overload operation" will lead to "bearing wear". The fault event logical relationship table is dynamically modeled through the preset Petri net modeling method to obtain the fault propagation network structure. Petri net is a graphical mathematical modeling tool that can be used to represent and analyze the dynamic behavior of the system.In this method, the fault event logical relationship table can be converted into a Petri net model, in which the place node represents the fault event and the transition node represents the logical relationship between the fault events. Through Petri net modeling, the dynamic propagation process between fault events can be more intuitively displayed. For example, a Petri net model can be constructed, in which "insufficient lubrication" and "overload operation" are respectively used as two place nodes, and "bearing wear" is used as another place node, and the logical relationship between them is represented by the transition node. Based on the fault propagation network structure, the fault logic structure is obtained. Through the above steps, a detailed fault propagation network structure has been constructed, which contains the logical relationship and dynamic propagation process between fault events. Based on this network structure, the fault logic structure of the target device can be obtained. For example, on the production line of an automobile manufacturing plant, if a fault propagation network structure about bearing wear at the joint of an assembly robot is constructed by the above method, a complete fault logic structure can be obtained, which includes not only the direct cause of bearing wear, but also the indirect cause and propagation path that may cause bearing wear. Through this method, the industrial Internet of Things system can analyze and model the fault development process of the equipment in detail, providing a scientific basis for fault prediction and prevention. For example, on the production line of an automobile manufacturer, once the system detects an abnormal vibration signal at a joint of an assembly robot and constructs a detailed fault logic structure through the above method, the system can predict the development trend of the fault, generate a detailed fault report, and notify the operator to take corresponding preventive measures, such as increasing the lubrication frequency, adjusting the load, or replacing the bearing in advance, thereby avoiding a larger fault and ensuring the continuity and safety of production. This fault logic structure construction method based on hierarchical analysis method, fault tree analysis method, Boolean algebra operation and Petri net modeling can not only improve the accuracy of fault analysis, but also provide scientific decision support for equipment maintenance and management.

[0114] In a specific embodiment, generating a warning instruction based on the predicted fault transfer path, and performing a fault alarm based on the warning instruction, includes:

[0115] Based on the predicted fault transfer path, a quantitative assessment is performed on the risk level in the fault propagation process to obtain a risk assessment value sequence;

[0116] Determining a warning level according to the risk assessment value sequence, and constructing warning information content according to the warning level and the predicted fault transfer path;

[0117] Based on the message queue technology, a transmission queue is constructed for the warning information content to obtain a warning information queue to be sent;

[0118] Based on a multi-channel communication protocol, the early warning information queue is sent to a receiving end of the target device to obtain an early warning instruction of sending completion;

[0119] By using visualization technology, a fault alarm is issued at the receiving end of the target device according to the early warning instruction.

[0120] Specifically, firstly, it is necessary to quantitatively evaluate the risk level in the fault propagation process based on the predicted fault transfer path to obtain a risk assessment value sequence. Risk assessment is to evaluate the potential risk level by analyzing each node and path segment in the predicted fault transfer path. For example, on the production line of an automobile manufacturing plant, if it is predicted that the bearing wear at a joint of an assembly robot may cause gear damage, the risk of each fault node and path segment can be quantitatively evaluated. The evaluation indicators may include the probability of fault occurrence, the scope of fault impact, the difficulty of repair, etc. Through these indicators, a risk assessment value sequence can be generated, each value corresponding to a node or path segment in the fault propagation path. Next, the warning level is determined according to the risk assessment value sequence, and the warning information content is constructed according to the warning level and the predicted fault transfer path. The warning level is usually divided into multiple levels, such as low risk, medium risk and high risk. The warning level can be determined according to the maximum value or average value in the risk assessment value sequence. For example, if the maximum value in the risk assessment value sequence exceeds a preset threshold, the warning level can be determined as high risk. The warning information content should include the fault location, fault type, risk level, predicted fault transfer path and recommended countermeasures. For example, the content of the warning information may be: "Warning: The second joint bearing of the left arm of the assembly robot is worn, the risk level is high, and it is expected to cause gear damage within the next 24 hours. Please check and replace it immediately." Then, based on the message queue technology, the transmission queue of the warning information content is constructed to obtain the warning information queue to be sent. Message queue technology is a message transmission mechanism for asynchronous communication, which can ensure the reliable transmission and processing of messages. In this method, the generated warning information content can be put into the message queue and wait for transmission. The message queue can be managed according to the first-in-first-out principle to ensure that the warning information is sent in order. For example, message queue technologies such as RabbitMQ or Kafka can be used to construct the warning information content into a queue to be sent. Based on the multi-channel communication protocol, the warning information queue is sent to the receiving end of the target device to obtain the warning instruction of the sending completion. The multi-channel communication protocol refers to the technology of information transmission through multiple communication channels (such as wireless network, wired network, local area network, etc.). In this method, a suitable communication protocol can be selected according to the actual situation of the target device, and the content in the warning information queue is sent to the receiving end of the target device. For example, the warning information can be sent to the operator's mobile device through a wireless network, to the monitoring system in the control room through a wired network, or to the alarm system in the workshop through a local area network. After the sending is completed, a warning instruction is generated to ensure that the warning information is successfully delivered. Finally, visualization technology is used to perform a fault alarm at the receiving end of the target device according to the warning instruction. Visualization technology can display the warning information in a graphical way, so that the operator can intuitively understand the fault situation.For example, on the monitoring screen in the control room, the fault location, fault type, risk level and predicted fault transfer path can be displayed in the form of charts, animations, etc. In addition, an alarm can be sounded through a sound alarm to alert the operator. In this way, the operator can quickly take corresponding measures, such as suspending the operation of the equipment, conducting inspections and repairs, or adjusting the operating parameters of the equipment, so as to avoid the further development of the fault and ensure the continuity and safety of production. In this way, the industrial Internet of Things system can realize real-time monitoring and early warning of equipment failures, take preventive measures in advance, and avoid further spread of the fault. For example, on the production line of an automobile manufacturing plant, once the system predicts that the bearing wear at a joint of the assembly robot may cause gear damage, the system will generate detailed early warning information and send the early warning information to the operator's mobile device, the monitoring system of the control room, and the alarm system of the workshop through multiple communication channels. The operator can take measures quickly based on the early warning information to avoid the occurrence of faults and ensure the smooth progress of production. This early warning instruction generation and fault alarm method based on risk assessment, message queue technology and multi-channel communication protocol can not only improve the efficiency of fault management, but also provide scientific decision support for equipment maintenance and management.

[0121] The above describes the device failure prediction method based on the industrial Internet of Things in the embodiment of the present invention. The following describes the device failure prediction device based on the industrial Internet of Things in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, an equipment failure prediction device based on industrial Internet of Things includes:

[0122] The acquisition module 21 is used to collect physical signals from the target device through a preset industrial Internet of Things to obtain an original sequence signal;

[0123] The identification module 22 is used to identify whether there is an abnormal signal in the original sequence signal through a preset abnormality detection algorithm to obtain an identification result;

[0124] An analysis module 23 is configured to, if the identification result shows that an abnormal signal exists, analyze the fault type of the target device based on the abnormal signal to obtain a fault location and a fault type corresponding to the fault location;

[0125] A prediction module 24, configured to perform a fault transfer prediction on the target device based on the abnormal signal, the fault location and the fault type corresponding to the fault location, to obtain a predicted fault transfer path;

[0126] The generating module 25 is used to generate an early warning instruction based on the predicted fault transfer path, and to issue a fault alarm based on the early warning instruction.

[0127] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0128] Reference Figure 3 The present invention also provides a computer device in an embodiment, wherein the internal structure of the computer device can be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0129] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0130] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0131] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0132] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0133] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for predicting equipment failure based on industrial Internet of Things, characterized in that: The following steps are involved: The physical signal of the target device is collected through the preset industrial Internet of Things to obtain the original sequence signal; Identify whether there is an abnormal signal in the original sequence signal by using a preset abnormality detection algorithm to obtain an identification result; If the identification result is that an abnormal signal exists, a fault type analysis is performed on the target device based on the abnormal signal to obtain a fault location and a fault type corresponding to the fault location; Performing a fault transfer prediction on the target device based on the abnormal signal, the fault location and the fault type corresponding to the fault location to obtain a predicted fault transfer path; generating an early warning instruction based on the predicted fault transfer path, and performing a fault alarm based on the early warning instruction; Performing signal layering processing on the abnormal signal through a preset hidden Markov model to obtain a layered abnormal signal; Performing physical structure mapping on the target device based on the layered abnormal signal to obtain a signal-structure mapping relationship; By using a preset Bayesian network, based on the signal-structure mapping relationship, the probability of the fault propagation path of the target device is analyzed to obtain a fault propagation probability distribution; Using a preset density space clustering algorithm, hot spot areas of the target device are identified based on the fault propagation probability distribution to obtain fault hot spot areas; Determine the fault location of the target device based on the fault hotspot area; wherein the fault location is the location with the highest probability of fault propagation in the fault hotspot area; In a fault hotspot area corresponding to the fault location, judging the fault type of the target device based on the abnormal signal to obtain the fault type corresponding to the fault location; By using a preset fault tree analysis method, a fault logic structure of the target device is constructed based on the fault location and fault type; Calling historical fault data in the database, and performing fault probability condition analysis on the fault logic structure based on the historical fault data to obtain a fault condition probability table; wherein the fault condition probability table includes a fault type identifier, a historical fault count, and a fault impact factor, and the fault type identifier includes a node identifier and a parent node identifier; By using a preset Markov chain technology, a fault state transition analysis is performed on the fault type based on the fault condition probability table to obtain a state combination space and a fault transition probability; A dynamic programming algorithm is used to predict the failure transfer of the target device based on the abnormal signal, the state combination space and the failure transfer probability to obtain a predicted failure transfer path.

2. The equipment failure prediction method based on industrial Internet of Things according to claim 1 is characterized in that: The physical signal acquisition of the target device through the preset industrial Internet of Things to obtain the original sequence signal includes: Physical signals of the device are collected through multiple types of sensors deployed on the target device to obtain preliminary collected data; Performing denoising processing on the preliminary collected data to obtain denoised preliminary collected data; The denoised preliminary collected data is segmented by a preset sliding window mechanism to generate data segments within a plurality of time windows, wherein each of the data segments contains a continuous signal within a certain time period; Using wavelet transform technology, each of the data segments is decomposed into multiple scales to extract signal features in different frequency ranges to obtain multi-scale feature signals; Performing dimensionality reduction processing on the multi-scale feature signal to extract the most representative feature signal to obtain a feature signal after dimensionality reduction; The characteristic signals after dimension reduction are combined in time sequence to obtain original sequence signals; wherein the original sequence signals include vibration signals, temperature signals and electrical signals.

3. The equipment failure prediction method based on industrial Internet of Things according to claim 1 is characterized in that: The method of identifying whether there is an abnormal signal in the original sequence signal by using a preset abnormality detection algorithm to obtain an identification result includes: The original sequence signal is subjected to feature compression and reconstruction by a preset deep learning autoencoder to obtain a compressed feature vector; Performing radius format conversion on the compressed feature vector using a preset hypersphere description model to obtain a conversion format radius; Coinciding one end of the conversion format radius with the center of the ball in the hypersphere description model to obtain a center point coincidence radius; wherein the center point coincidence radius is a conversion format radius that coincides with the center of the ball; Using a preset anomaly detection algorithm, identifying whether the other end of the center point coincidence radius reaches the surface boundary of the hypersphere description model; If so, the original sequence signal corresponding to the center point coincidence radius that does not reach the surface boundary of the hypersphere description model is marked as abnormal to obtain a recognition result.

4. The equipment failure prediction method based on industrial Internet of Things according to claim 1 is characterized in that: The method of constructing a fault logic structure of the target device based on the fault location and the fault type by using a preset fault tree analysis method includes: By using the hierarchical analysis method, the fault location is divided into regions to obtain a fault region hierarchy structure; wherein the fault region hierarchy structure includes fault region boundaries, connection relationships between fault region boundaries, and weights of fault region boundaries; By using a preset fault tree analysis method, the fault factor is determined based on the fault area hierarchy and the fault type, and the minimum cut set analysis is performed on the fault factor to obtain a set of fault basic events; wherein the set of fault basic events includes fault basic events and fault basic event combinations; Performing causal correlation analysis on the set of basic fault events to obtain a causal chain of fault events; Through Boolean algebra operations, based on the causal chain of the fault events, logic gate analysis is performed on the combination of the fault basic events to obtain a fault event logic relationship table; The fault event logic relationship table is dynamically modeled by a preset petri net modeling method to obtain a fault propagation network structure; wherein the fault propagation network structure includes a place node and a transition node; Based on the fault propagation network structure, a fault logic structure is obtained.

5. The method for predicting equipment failure based on industrial Internet of Things according to claim 1, characterized in that: The generating of an early warning instruction based on the predicted fault transfer path, and performing a fault alarm based on the early warning instruction, comprises: Based on the predicted fault transfer path, a quantitative assessment is performed on the risk level in the fault propagation process to obtain a risk assessment value sequence; Determining a warning level according to the risk assessment value sequence, and constructing warning information content according to the warning level and the predicted fault transfer path; Based on the message queue technology, a transmission queue is constructed for the warning information content to obtain a warning information queue to be sent; Based on a multi-channel communication protocol, the early warning information queue is sent to a receiving end of the target device to obtain an early warning instruction of sending completion; By using visualization technology, a fault alarm is issued at the receiving end of the target device according to the early warning instruction.

6. An equipment failure prediction device based on industrial Internet of Things, characterized in that: include: The acquisition module is used to collect physical signals from the target device through the preset industrial Internet of Things to obtain the original sequence signal; An identification module is used to identify whether there is an abnormal signal in the original sequence signal through a preset abnormality detection algorithm to obtain an identification result; An analysis module, configured to, if the identification result indicates that an abnormal signal exists, analyze the fault type of the target device based on the abnormal signal to obtain a fault location and a fault type corresponding to the fault location; A prediction module, configured to predict a fault transfer path for the target device based on the abnormal signal, the fault location and the fault type corresponding to the fault location, and obtain a predicted fault transfer path; A generating module, configured to generate an early warning instruction based on the predicted fault transfer path, and to issue a fault alarm based on the early warning instruction; Performing signal layering processing on the abnormal signal through a preset hidden Markov model to obtain a layered abnormal signal; Performing physical structure mapping on the target device based on the layered abnormal signal to obtain a signal-structure mapping relationship; By using a preset Bayesian network, based on the signal-structure mapping relationship, the probability of the fault propagation path of the target device is analyzed to obtain a fault propagation probability distribution; Using a preset density space clustering algorithm, hot spot areas of the target device are identified based on the fault propagation probability distribution to obtain fault hot spot areas; Determine the fault location of the target device based on the fault hotspot area; wherein the fault location is the location with the highest probability of fault propagation in the fault hotspot area; In a fault hotspot area corresponding to the fault location, judging the fault type of the target device based on the abnormal signal to obtain the fault type corresponding to the fault location; By using a preset fault tree analysis method, a fault logic structure of the target device is constructed based on the fault location and fault type; Calling historical fault data in the database, and performing fault probability condition analysis on the fault logic structure based on the historical fault data to obtain a fault condition probability table; wherein the fault condition probability table includes a fault type identifier, a historical fault count, and a fault impact factor, and the fault type identifier includes a node identifier and a parent node identifier; By using a preset Markov chain technology, a fault state transition analysis is performed on the fault type based on the fault condition probability table to obtain a state combination space and a fault transition probability; A dynamic programming algorithm is used to predict the failure transfer of the target device based on the abnormal signal, the state combination space and the failure transfer probability to obtain a predicted failure transfer path.

7. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images