Hybrid fault-tolerant control method and system for simultaneous multi-sensor failure scenarios

Abstract

The present invention provides a hybrid fault-tolerant control method and system for scenarios where multiple sensors fail simultaneously. The method comprises: inputting historical time-series feature data of sensors into a long-short-term memory network and a stacking model, respectively, to obtain predicted values ​​of the sensor's output data at the current moment output by the long-short-term memory network and the stacking model; determining whether a sensor failure has occurred using a sequential probability ratio test based on the difference between the predicted value of the sensor's output data at the current moment output by the long-short-term memory network and the actual value of the output data; and, if a sensor failure has occurred, determining a reconstructed value of the sensor's output data at the current moment based on the predicted value of the sensor's output data at the current moment output by the long-short-term memory network and the stacking model. The present invention can perform real-time sensor fault detection and output data reconstruction, improving the stability and reliability of the system in the event of multi-point observation failure.

Description

Technical Field

[0001] The present invention relates to the technical field of sensor fault diagnosis and fault-tolerant control, and in particular to a hybrid fault-tolerant control method and system for scenarios where multiple sensors fail simultaneously. Background Art

[0002] As industrial systems rapidly develop towards high automation and intelligence, multi-sensor systems are widely used in key fields such as aerospace, intelligent manufacturing, nuclear power, shipping and rail transportation because they can achieve comprehensive perception and precise monitoring of complex operating conditions.

[0003] However, although the multi-sensor collaborative mechanism has played a positive role in improving the robustness of the system, once multiple sensors experience drift, failure, short circuit or data anomaly within the same time period, the original redundant design will be destroyed, which may not only cause confusion or failure of the fault detection system, leading to data fusion misjudgment and failure of the redundant mechanism, but also may cause control system execution errors, and in severe cases even cause paralysis of the entire system or catastrophic consequences, thus putting higher requirements on system safety and stability.

[0004] At present, the research on fault-tolerant control strategies is still limited to single sensor fault tolerance strategies. There is still no strategy that can ensure the stable operation of the system when multiple sensors fail at the same time. Summary of the Invention

[0005] The present invention provides a hybrid fault-tolerant control method and system for scenarios where multiple sensors fail simultaneously, to address the defect in the prior art that there is no strategy that can ensure the robust operation of the system when multiple sensors fail simultaneously, to implement hybrid fault-tolerant control for scenarios where multiple sensors fail simultaneously, and to improve the stability and reliability of the system in the event of multi-point observation failure.

[0006] The present invention provides a hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously, comprising:

[0007] Inputting the historical time series feature data of the sensor into the long short-term memory network and the stacking model respectively, and obtaining the output data prediction value of the sensor at the current moment output by the long short-term memory network and the stacking model;

[0008] Determining whether the sensor is faulty using a sequential probability ratio test based on a difference between a predicted value of the output data of the sensor at the current moment output by the long short-term memory network and an actual value of the output data;

[0009] In the event that the sensor fails, a reconstructed value of the output data of the sensor at the current moment is determined based on the predicted value of the output data of the sensor at the current moment output by the long short-term memory network and the stacking model.

[0010] According to a hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously, the present invention provides a method for determining whether a sensor fails using a sequential probability ratio test based on the difference between a predicted value of output data of the sensor at the current moment output by the long short-term memory network and the actual value of the output data, including:

[0011] If the difference is greater than a first preset threshold, determining that the sensor is faulty;

[0012] When the difference is less than a second preset threshold, it is determined that the sensor is normal, and the second preset threshold is less than the first preset threshold;

[0013] When the difference is less than or equal to the first preset threshold and greater than or equal to the second preset threshold, a sequential probability ratio test is used to determine whether the sensor is faulty based on the difference between the current moment and the next moment of the sensor.

[0014] According to a hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously, the present invention provides a method for inputting historical time series feature data of the sensors into a stacking model to obtain a predicted value of the output data of the sensors at the current moment output by the stacking model, including:

[0015] Inputting the historical time series feature data of the sensor into multiple basic learners in the Stacking model to obtain the output data prediction value of the sensor at the current moment output by each basic learner;

[0016] The output data prediction value of the sensor at the current moment output by the basic learner is input into the meta-learner in the Stacking model to obtain the output data prediction value of the sensor at the current moment output by the meta-learner.

[0017] According to a hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously, the present invention provides a basic learner including an optical gradient boosting machine, an extreme gradient boosting machine, a gradient boosting decision tree, a K-nearest neighbor, and a random forest, and the meta-learner is a minimum absolute shrinkage selection operator.

[0018] According to a hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously, the method comprises: determining a reconstructed value of the output data of the sensor at the current moment based on the predicted value of the output data of the sensor at the current moment output by the long short-term memory network and the stacking model; and

[0019] The output data prediction values ​​of the sensor at the current moment output by the long short-term memory network and the stacking model are weightedly fused as the output data reconstructed value of the sensor at the current moment.

[0020] According to a hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously, the output data prediction values ​​of the sensors at the current moment output by the long short-term memory network and the stacking model are weightedly fused using the following formula to obtain a reconstructed output data value of the sensors at the current moment:

[0021]

[0022] ,and

[0023] in, Reconstruct the output data value of the sensor at the current moment, and are the output data prediction values ​​of the sensor at the current moment output by the long short-term memory network and the stacking model respectively, and is a weighting coefficient, which is adaptively adjusted according to the prediction accuracy of the long short-term memory network and the stacking model.

[0024] According to the present invention, a hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously is provided. The calculation formula is:

[0025]

[0026] in, is the adjustment coefficient, and are the residual standard deviations of the prediction results of the long short-term memory network and the stacking model respectively.

[0027] The present invention also provides a hybrid fault-tolerant control system for a scenario where multiple sensors fail simultaneously, comprising:

[0028] A prediction module, configured to input the historical time series feature data of the sensor into the long short-term memory network and the stacking model respectively, and obtain the output data prediction value of the sensor at the current moment output by the long short-term memory network and the stacking model;

[0029] a detection module, configured to determine whether the sensor has failed by using a sequential probability ratio test based on a difference between a predicted value of output data of the sensor at a current moment output by the long short-term memory network and an actual value of the output data;

[0030] The reconstruction module is used to determine the reconstructed value of the output data of the sensor at the current moment according to the output data prediction value of the sensor at the current moment output by the long short-term memory network and the stacking model when the sensor fails.

[0031] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously as described above is implemented.

[0032] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously as described above is implemented.

[0033] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-described hybrid fault-tolerant control methods for a scenario where multiple sensors fail simultaneously.

[0034] The present invention provides a hybrid fault-tolerant control method and system for scenarios where multiple sensors fail simultaneously. Aiming at the problem of concurrent failure of multiple sensors in an industrial system within the same time period, a "fault diagnosis + information reconstruction" collaborative architecture is designed. A long short-term memory network (LSTM) combined with a sequential probability ratio test (SPRT) algorithm model is adopted to achieve high-precision online fault detection and rapid positioning of sensors. Based on the diagnosis results, a stacked ensemble learning model and LSTM are further used to construct a virtual sensor, and the observation data of the failed channel is intelligently reconstructed and compensated, realizing dual-time and space drive, and performing real-time sensor fault detection and output data reconstruction, thereby improving the stability and reliability of the system in the case of multi-point observation failure. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0036] Figure 1 This is one of the flow charts of the hybrid fault-tolerant control method provided by the present invention for a scenario where multiple sensors fail simultaneously;

[0037] Figure 2 This is the second flow chart of the hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously, provided by the present invention;

[0038] Figure 3 This is a structural diagram of the LSTM model in the hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously, provided by the present invention;

[0039] Figure 4 This is a schematic diagram of the principle of sequential probability ratio in the hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously, provided by the present invention;

[0040] Figure 5 It is a structural diagram of the Stacking model in the hybrid fault-tolerant control method for the scenario of simultaneous failure of multiple sensors provided by the present invention.

[0041] Figure 6 This is a schematic diagram of the structure of a hybrid fault-tolerant control system provided by the present invention for a scenario where multiple sensors fail simultaneously;

[0042] Figure 7 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0043] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0044] Aiming at complex scenarios where multiple sensors fail simultaneously within the same time period, this paper proposes a hybrid fault-tolerant control method that integrates deep learning and statistical inference to address the system monitoring blind spots, control failures, and misjudgment and miscontrol problems caused by simultaneous multi-sensor failures. The method mainly includes the following aspects:

[0045] 1. Identification of simultaneous failures of multiple sensors:

[0046] Because multiple sensors can drift, distort, or fail simultaneously within a short period of time, traditional single-sensor diagnostic methods cannot effectively locate failed sensors. This embodiment uses a Long Short-Term Memory (LSTM) network combined with a Sequential Probability Ratio Test (SPRT) to jointly model and statistically infer the time series output by multi-channel sensors, accurately identifying faulty sensor channels.

[0047] 2. Information reconstruction problem after system observation is lost:

[0048] Multiple sensor failures can lead to the loss of critical status information, severely impacting system operational status assessment and the correct execution of control instructions. This embodiment uses a stacking ensemble learning model to construct virtual sensors, providing real-time complementation and dynamic prediction of observations from failed channels, restoring the system's observational integrity.

[0049] 3. False alarm and miscontrol problems caused by multiple faults:

[0050] Simultaneous anomalies in multi-channel data can trigger false alarms and even mislead the control system into responding incorrectly. This invention effectively reduces false alarm rates and control errors through a collaborative judgment mechanism that discriminates model outputs and reconstruction results, thereby improving the overall safety and stability of the system.

[0051] 4. Lack of a unified fault-tolerance strategy for scenarios where multiple sensors fail simultaneously:

[0052] Most existing methods are designed for single-point failures and are unable to cope with the concurrent failure of multiple sensors. This paper proposes an end-to-end fault-tolerant framework that integrates diagnosis, reconstruction, and decision-making. This framework establishes a generalizable, system-level multi-fault response mechanism with strong versatility and engineering application value.

[0053] The following combination Figure 1 The present invention describes a hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously, comprising:

[0054] Step 101: Input historical time series feature data of the sensor into a long short-term memory network and a stacking model respectively, and obtain output data prediction values ​​of the sensor at the current moment output by the long short-term memory network and the stacking model;

[0055] Step 102, determining whether the sensor is faulty using a sequential probability ratio test based on the difference between the predicted output data value of the sensor at the current moment output by the long short-term memory network and the actual output data value;

[0056] Step 103 , when the sensor fails, determining a reconstructed output data value of the sensor at the current moment based on the output data prediction value of the sensor at the current moment output by the long short-term memory network and the stacking model.

[0057] This embodiment provides a multi-sensor fault-tolerant hybrid control strategy, which mainly includes a fault diagnosis (FDI) module and a reconstruction control module. The system can maintain system stability and expected performance when multiple sensors fail at the same time. The work flow diagram is as follows: Figure 2 As shown. The fault diagnosis (FDI) module adopts an online fault detection method that combines the long short-term memory network (LSTM) and the sequential probability ratio test (SPRT). First, the target feature (such as the actual value of the sensor's output data) is selected and isolated from other features, and only this feature is used for prediction. The feature data is first normalized to convert it into a format suitable for supervised learning and divided into training and test sets. By constructing a neural network model containing a single LSTM layer and a Dense output layer, the mean square error loss function and the Adam optimizer are used for training. The structure of the LSTM model is shown in Figure 3 shown.

[0058] After model training is complete, SPRT is used during the actual prediction process for fault detection. The likelihood ratio of the residual (the difference between the actual output data value and the predicted output data value) is calculated to determine whether the target sensor is faulty. This method can detect and locate the faulty sensor in real time when a fault occurs.

[0059] When a system or device fails, the fault diagnosis module (FDI) detects the target sensor failure. A virtual sensor quickly activates and replaces the faulty sensor to reconstruct data. The virtual sensor algorithm proposed in this embodiment uses the Stacking model, an ensemble learning method that combines two layers of learners to reconstruct information. The primary learners in the Stacking model can be homogeneous or heterogeneous. Basic learners are trained separately and then integrated through a meta-learner.

[0060] This embodiment addresses the problem of multiple sensors failing concurrently within the same time period in industrial systems by designing a collaborative "fault diagnosis + information reconstruction" architecture. This architecture employs a long short-term memory (LSTM) network combined with a sequential probability ratio test (SPRT) algorithm model to achieve high-precision online fault detection and rapid location of sensor faults. Based on the diagnostic results, a stacked ensemble learning model and LSTM are further utilized to construct virtual sensors. This intelligently reconstructs and compensates for observation data from failed channels, achieving dual-drive in both time and space. This allows for real-time sensor fault detection and output data reconstruction, thereby improving the system's stability and reliability in the event of multi-point observation failures.

[0061] Based on the above embodiment, this embodiment uses a sequential probability ratio test to determine whether the sensor is faulty based on the difference between the predicted value of the output data of the sensor at the current moment output by the long short-term memory network and the actual value of the output data, including:

[0062] If the difference is greater than a first preset threshold, determining that the sensor is faulty;

[0063] When the difference is less than a second preset threshold, it is determined that the sensor is normal, and the second preset threshold is less than the first preset threshold;

[0064] When the difference is less than or equal to the first preset threshold and greater than or equal to the second preset threshold, a sequential probability ratio test is used to determine whether the sensor is faulty based on the difference between the current moment and the next moment of the sensor.

[0065] The formula for constructing the LSTM prediction model is as follows:

[0066]

[0067]

[0068] Among them, X is the historical time series feature data, is the predicted value of the output data of the sensor at time t+1.

[0069] The residual is calculated as follows:

[0070]

[0071] in, is the actual value of the output data of the sensor at time t+1, It is the difference between the predicted value of the sensor output data at time t+1 and the actual value of the output data.

[0072] The discriminant function of SPRT is as follows:

[0073]

[0074] in, is the discriminant function value of SPRT, and are the distribution functions of the observation sequences.

[0075] The schematic diagram of the SPRT principle is as follows Figure 4 The judgment criteria are as follows:

[0076] ;

[0077] ;

[0078] .

[0079] Based on the above embodiment, the structural diagram of the Stacking model is as follows: Figure 5 As shown, in this embodiment, the historical time series feature data of the sensor is input into the Stacking model to obtain the output data prediction value of the sensor at the current moment output by the Stacking model, including:

[0080] Inputting the historical time series feature data of the sensor into multiple basic learners in the Stacking model to obtain the output data prediction value of the sensor at the current moment output by each basic learner;

[0081] The output data prediction value of the sensor at the current moment output by the basic learner is input into the meta-learner in the Stacking model to obtain the output data prediction value of the sensor at the current moment output by the meta-learner.

[0082] Based on the above embodiments, the basic learners in this embodiment include optical gradient boosting machine (LGBM), extreme gradient boosting machine (XGBoost), gradient boosting decision tree (GBDT), K-nearest neighbor (KNN) and random forest (RF), and the meta-learner is the least absolute shrinkage selection operator (LASSO).

[0083] In order to avoid overfitting, 5 grid search cross validations can be used to optimize the model parameters. Randomly divided into 5 data sets of equal size 、 、 、 and , use the basic learning algorithm to train the basic learner. Obtain data from the remaining data sets and finally generate the secondary training set , this dataset is used to train the meta-learner to generate the final fusion model. The specific processing flow is as follows:

[0084] Primary learners: LGBM, XGBoost, GBDT, KNN, Random Forest (RF);

[0085] Meta-learner: LASSO regression model;

[0086] The output of each primary model is:

[0087]

[0088] Among them, x is the input of each primary model, and M is the number of primary models.

[0089] Construct the training data matrix:

[0090]

[0091] Finally, the final output is obtained through the meta-learner g(·):

[0092] .

[0093] Based on the above embodiment, in this embodiment, the reconstructed value of the output data of the sensor at the current moment is determined based on the predicted value of the output data of the sensor at the current moment output by the long short-term memory network and the stacking model, including:

[0094] The output data prediction values ​​of the sensor at the current moment output by the long short-term memory network and the stacking model are weightedly fused as the output data reconstructed value of the sensor at the current moment.

[0095] This embodiment uses a stacking ensemble learning model (Stacking) and the LSTM algorithm to construct a spatiotemporal dual-drive virtual sensor, enabling the entire fault-tolerant control system to possess spatiotemporal dual-drive characteristics. This allows for real-time completion and dynamic prediction of the observation values ​​of failed channels. Based on the weights adaptively assigned by the model, the dynamic fusion of the LSTM and Stacking prediction outputs can be achieved, restoring the system's observation integrity and effectively ensuring the system's ability to operate continuously even when multi-channel information is missing.

[0096] Based on the above embodiment, in this embodiment, the output data prediction values ​​of the sensor at the current moment output by the long short-term memory network and the stacking model are weighted and fused using the following formula to serve as the reconstructed output data value of the sensor at the current moment:

[0097]

[0098] , and

[0099] in, Reconstruct the output data value of the sensor at the current moment, and are the output data prediction values ​​of the sensor at the current moment output by the long short-term memory network and the stacking model respectively, and is a weighting coefficient, which is adaptively adjusted according to the prediction accuracy of the long short-term memory network and the stacking model.

[0100] The working principle of the hybrid fault-tolerant control strategy for multi-sensor faults is as follows Figure 2 As shown in the figure, the system initially assumes all components are fully functional and monitors the status of target sensors in real time through the Fault Detection and Isolation (FDI) module. When a target sensor is operating normally, the system directly outputs its measurement value. If the FDI module detects a target sensor failure, the system immediately triggers an alarm and switches to virtual sensor fusion mode.

[0101] In this mode, the system uses virtual sensors (LSTM-based) and virtual sensors (Based on Stacking) Data reconstruction. Through weighted fusion formula , generate simulated values ​​to replace the data of the faulty sensor, realize information reconstruction, and ensure the continuous operation of the system. Weighting coefficient and It is adaptively adjusted according to the reliability of each sensor output through a learning algorithm.

[0102] Based on the above embodiment, in this embodiment The calculation formula is:

[0103]

[0104] in, is the adjustment coefficient, and are the residual standard deviations of the prediction results of the long short-term memory network and the stacking model respectively.

[0105] The residual of the prediction results of the long short-term memory network and the stacking model refers to the difference between the output data prediction value of the long short-term memory network and the stacking model each time and the corresponding output data actual value. The standard deviation of the difference is calculated. Calculation.

[0106] In addition, this system also uses the sequential probability ratio test (SPRT) method to monitor the status of multiple sensors to ensure that the system can still operate robustly and continue to provide high reliability and fault tolerance when multiple sensors fail simultaneously.

[0107] The design of this multi-sensor fault-tolerant hybrid control system ensures continued normal operation even when a single target sensor fails. It also maintains high reliability and fault tolerance even when multiple characteristic sensors fail simultaneously. The system demonstrates excellent adaptability and robustness in complex fault scenarios, effectively compensating for data bias and measurement errors caused by sensor failures, thereby ensuring the stability and accuracy of industrial systems in extreme environments.

[0108] Aiming at the extreme scenario where multiple sensors fail simultaneously within the same time period, this paper proposes a hybrid fault-tolerant control method that integrates fault detection and information reconstruction. This method offers significant advantages over existing technologies in terms of system robustness, detection accuracy, and operational efficiency. Specific technical benefits include:

[0109] 1. Realize high-precision real-time diagnosis under simultaneous faults of multiple sensors. This invention introduces the long short-term memory network (LSTM) model to model the historical time series data of the target sensor, and combines it with the sequential probability ratio test (SPRT) to perform dynamic statistical analysis on the model prediction residuals. It can quickly identify the faulty channel when multiple sensors experience abnormal fluctuations or failures at the same time, thereby improving the accuracy and real-time performance of online fault detection and avoiding the problems of missed detection and misjudgment of traditional methods when facing concurrent faults.

[0110] 2. Enhance the fault tolerance and robustness of complex systems. When the fault diagnosis module (FDI) detects a sensor failure, the system will immediately switch to a dual-channel virtual sensor fusion mechanism consisting of an LSTM model and a Stacking model. Through comprehensive analysis of multiple healthy channel information and historical sequences, it completes dynamic compensation and reconstruction of the failed variables, significantly improving the system's stable operation capability in the presence of severe information loss and effectively avoiding system interruption or loss of control.

[0111] 3. Reduce false alarm rate and control the risk of malfunction. The present invention performs likelihood ratio judgment and feature selection on the prediction residual, combined with a multi-source redundant information fusion reconstruction strategy, to effectively reduce false alarms or erroneous execution behaviors caused by environmental interference, instantaneous fluctuations, etc., improve the safety and reliability of system operation, and avoid resource waste and safety hazards caused by erroneous operations.

[0112] 4. Optimize resource allocation and reduce system operation and maintenance and hardware costs. Compared with the traditional method of achieving fault tolerance through redundant sensor deployment, the present invention relies solely on algorithmic means to maintain system performance in the event of multiple failures, reducing dependence on sensor hardware redundancy, reducing the system's initial construction cost and subsequent maintenance costs, and improving overall economic benefits.

[0113] The hybrid fault-tolerant control system for the scenario of simultaneous failure of multiple sensors provided by the present invention is described below. The hybrid fault-tolerant control system for the scenario of simultaneous failure of multiple sensors described below and the hybrid fault-tolerant control method for the scenario of simultaneous failure of multiple sensors described above can be referenced to each other.

[0114] like Figure 6 As shown, the system includes a prediction module 601, a detection module 602 and a reconstruction module 603, wherein:

[0115] The prediction module 601 is used to input the historical time series feature data of the sensor into the long short-term memory network and the stacking model respectively, and obtain the output data prediction value of the sensor at the current moment output by the long short-term memory network and the stacking model;

[0116] The detection module 602 is configured to determine whether the sensor is faulty using a sequential probability ratio test based on a difference between a predicted value of the output data of the sensor at a current moment output by the long short-term memory network and an actual value of the output data;

[0117] The reconstruction module 603 is used to determine the reconstructed value of the output data of the sensor at the current moment according to the output data prediction value of the sensor at the current moment output by the long short-term memory network and the stacking model when the sensor fails.

[0118] This embodiment addresses the problem of multiple sensors failing concurrently within the same time period in industrial systems by designing a collaborative "fault diagnosis + information reconstruction" architecture. This architecture employs a long short-term memory (LSTM) network combined with a sequential probability ratio test (SPRT) algorithm model to achieve high-precision online fault detection and rapid location of sensor faults. Based on the diagnostic results, a stacked ensemble learning model and LSTM are further utilized to construct virtual sensors. This intelligently reconstructs and compensates for observation data from failed channels, achieving dual-drive in both time and space. This allows for real-time sensor fault detection and output data reconstruction, thereby improving the system's stability and reliability in the event of multi-point observation failures.

[0119] Figure 7 An example of a physical structure diagram of an electronic device is shown below. Figure 7As shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730, and a communication bus 740, wherein the processor 710, the communication interface 720, and the memory 730 communicate with each other via the communication bus 740. The processor 710 may call logic instructions in the memory 730 to execute a hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously. The method includes: inputting historical time series feature data of the sensor into a long short-term memory network and a stacking model, respectively, to obtain output data prediction values ​​of the sensor at the current moment output by the long short-term memory network and the stacking model; determining whether the sensor has failed using a sequential probability ratio test based on the difference between the output data prediction value of the sensor at the current moment output by the long short-term memory network and the actual output data value; and, if the sensor has failed, determining a reconstructed output data value of the sensor at the current moment based on the output data prediction value of the sensor at the current moment output by the long short-term memory network and the stacking model.

[0120] Furthermore, the logic instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0121] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the hybrid fault-tolerant control method for the simultaneous failure scenario of multiple sensors provided by the above methods. The method includes: inputting the historical time series feature data of the sensor into the long short-term memory network and the Stacking model respectively to obtain the output data prediction value of the sensor at the current moment output by the long short-term memory network and the Stacking model; using a sequential probability ratio test to determine whether the sensor has failed based on the difference between the output data prediction value of the sensor at the current moment output by the long short-term memory network and the actual value of the output data; in the event that the sensor fails, determining the output data reconstruction value of the sensor at the current moment based on the output data prediction value of the sensor at the current moment output by the long short-term memory network and the Stacking model.

[0122] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the hybrid fault-tolerant control method for the simultaneous failure scenario of multiple sensors provided by the above-mentioned methods, the method comprising: inputting the historical time series feature data of the sensor into the long short-term memory network and the Stacking model respectively, to obtain the output data prediction value of the sensor at the current moment output by the long short-term memory network and the Stacking model; using a sequential probability ratio test to determine whether the sensor has failed based on the difference between the output data prediction value of the sensor at the current moment output by the long short-term memory network and the actual value of the output data; in the event that the sensor has failed, determining the output data reconstruction value of the sensor at the current moment based on the output data prediction value of the sensor at the current moment output by the long short-term memory network and the Stacking model.

[0123] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0124] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A hybrid fault-tolerant control method for scenarios where multiple sensors fail simultaneously, characterized in that: include: Inputting the historical time series feature data of the sensor into the long short-term memory network and the stacking model respectively, and obtaining the output data prediction value of the sensor at the current moment output by the long short-term memory network and the stacking model; Determining whether the sensor is faulty using a sequential probability ratio test based on a difference between a predicted value of the output data of the sensor at the current moment output by the long short-term memory network and an actual value of the output data; In the event that the sensor fails, determining a reconstructed output data value of the sensor at the current moment based on the output data prediction value of the sensor at the current moment output by the long short-term memory network and the stacking model; Determining a reconstructed value of the output data of the sensor at the current moment according to the predicted value of the output data of the sensor at the current moment output by the long short-term memory network and the stacking model includes: Performing weighted fusion on the output data prediction values ​​of the sensor at the current moment output by the long short-term memory network and the stacking model to obtain a reconstructed output data value of the sensor at the current moment; The output data prediction values ​​of the sensor at the current moment output by the long short-term memory network and the stacking model are weighted and fused using the following formula to serve as the output data reconstruction value of the sensor at the current moment: ； ,and ; in, Reconstruct the output data value of the sensor at the current moment, and are the output data prediction values ​​of the sensor at the current moment output by the long short-term memory network and the stacking model respectively, and is a weighting coefficient, which is adaptively adjusted according to the prediction accuracy of the long short-term memory network and the stacking model.

2. The hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously according to claim 1 is characterized in that: Determining whether the sensor is faulty using a sequential probability ratio test based on a difference between a predicted value of output data of the sensor at a current moment output by the long short-term memory network and an actual value of the output data includes: If the difference is greater than a first preset threshold, determining that the sensor is faulty; When the difference is less than a second preset threshold, it is determined that the sensor is normal, and the second preset threshold is less than the first preset threshold; When the difference is less than or equal to the first preset threshold and greater than or equal to the second preset threshold, a sequential probability ratio test is used to determine whether the sensor is faulty based on the difference between the current moment and the next moment of the sensor.

3. The hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously according to claim 1 is characterized in that: Inputting the historical time series feature data of the sensor into the Stacking model to obtain the output data prediction value of the sensor at the current moment output by the Stacking model includes: Inputting the historical time series feature data of the sensor into multiple basic learners in the Stacking model to obtain the output data prediction value of the sensor at the current moment output by each basic learner; The output data prediction value of the sensor at the current moment output by the basic learner is input into the meta-learner in the Stacking model to obtain the output data prediction value of the sensor at the current moment output by the meta-learner.

4. The hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously according to claim 3 is characterized in that: The basic learners include optical gradient boosting machine, extreme gradient boosting machine, gradient boosting decision tree, K nearest neighbor and random forest, and the meta-learner is the least absolute shrinkage selection operator.

5. The hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously according to claim 1 is characterized in that: The calculation formula is: ； in, is the adjustment coefficient, and are the residual standard deviations of the prediction results of the long short-term memory network and the stacking model respectively.

6. A hybrid fault-tolerant control system for scenarios where multiple sensors fail simultaneously, characterized in that: The hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously, as described in any one of claims 1 to 5, comprises: A prediction module is used to input the historical time series feature data of the sensor into the long short-term memory network and the stacking model respectively, and obtain the output data prediction value of the sensor at the current moment output by the long short-term memory network and the stacking model; a detection module, configured to determine whether the sensor has failed by using a sequential probability ratio test based on a difference between a predicted value of output data of the sensor at a current moment output by the long short-term memory network and an actual value of the output data; The reconstruction module is used to determine the reconstructed value of the output data of the sensor at the current moment based on the output data prediction value of the sensor at the current moment output by the long short-term memory network and the stacking model when the sensor fails.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously as described in any one of claims 1 to 5 is implemented.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the hybrid fault-tolerant control method for a scenario where multiple sensors fail simultaneously as described in any one of claims 1 to 5 is implemented.

Images