Primary helium fan state monitoring and early warning method and system based on multi-source heterogeneous data fusion

Abstract

The invention discloses a multi-source heterogeneous data fused primary helium fan state monitoring and early warning method, which comprises the following steps: collecting primary helium fan vibration data, motor parameter data and environment data, preprocessing the data, fusing multi-source heterogeneous data by adopting a deep neural network based on an attention mechanism, and obtaining a primary helium fan state monitoring and early warning result; outputting a comprehensive fault diagnosis feature vector; based on the fault diagnosis model of the fusion features, training a generator model for simulating future fault conditions by adopting a generative adversarial network, and carrying out adversarial training on the generator model and a fault prediction model based on LSTM (Long Short Term Memory); and a deep reinforcement learning algorithm is adopted to carry out adaptive optimization on the nuclear island layer diagnosis model and the prediction model, and fault early warning is triggered based on diagnosis and prediction results. According to the method, a layered industrial Internet of Things architecture is constructed, and intelligent monitoring and early warning of the primary helium fan are achieved.

Description

Technical Field

[0001] The present invention relates to the field of nuclear power technology, and in particular to a main helium blower state monitoring and early warning method and system with multi-source heterogeneous data fusion. Background Art

[0002] As a key equipment of a nuclear power plant, the main helium blower of a nuclear power plant plays a vital role in the entire operation process. The main function of the main helium blower is to compress helium in the primary circuit system[1] to ensure the safety and reliability of reactor operation. Helium, as the coolant of the nuclear reactor, circulates continuously in the primary circuit system to remove the heat generated by the reactor core. Therefore, it is very important to monitor the status of the main helium blower in real time and provide fault warning.

[0003] Existing main helium blower monitoring technologies mainly focus on a single data source, such as vibration monitoring or motor parameter monitoring. This single data source monitoring method is difficult to fully and accurately diagnose the operating status of the equipment, and there are problems of missed detection or false alarms, and it cannot effectively warn of impending failures. In addition, most existing monitoring systems use manual experience judgment, which makes it difficult to achieve adaptive fault diagnosis and prediction.

[0004] In view of the above problems, there is an urgent need for a main helium blower status monitoring and early warning system that can integrate multi-source heterogeneous data and use intelligent algorithms for fault diagnosis and prediction. The system should be able to fully perceive the operating status of the equipment, detect abnormalities in a timely manner, provide strong support for equipment maintenance, and improve the safety and reliability of nuclear power plants. The present invention aims to provide a multi-source heterogeneous data fusion method based on deep learning and reinforcement learning, build a layered industrial Internet of Things architecture, and realize intelligent monitoring and early warning of the main helium blower. Summary of the invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the present invention provides a main helium blower state monitoring and early warning method that integrates multi-source heterogeneous data, and a main helium blower state monitoring and early warning system that can integrate multi-source heterogeneous data and use intelligent algorithms to perform fault diagnosis and prediction.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: a main helium blower state monitoring and early warning method with multi-source heterogeneous data fusion, comprising:

[0008] The vibration data of the main helium fan, motor parameter data and environmental data are collected and preprocessed. A deep neural network based on the attention mechanism is used to fuse multi-source heterogeneous data to output a comprehensive fault diagnosis feature vector. Based on the fault diagnosis model of the fused features, a generative adversarial network is used to train a generator model that simulates future fault conditions, and the generator model is adversarially trained with the LSTM-based fault prediction model. A deep reinforcement learning algorithm is used to adaptively optimize the nuclear island layer diagnosis model and prediction model, and a fault warning is triggered based on the diagnosis and prediction results.

[0009] As a preferred solution of the main helium fan status monitoring and early warning method of multi-source heterogeneous data fusion described in the present invention, wherein: the main helium fan vibration data includes, by arranging an acceleration sensor in the main helium fan, collecting vibration signals in real time; the motor parameter data includes, installing voltage and current sensors at the motor end, and using a speed sensor to obtain speed, torque and power parameters; the environmental data includes monitoring and collecting temperature and humidity changes in the environment where the main helium fan is located.

[0010] As a preferred solution of the main helium blower state monitoring and early warning method of multi-source heterogeneous data fusion described in the present invention, the preprocessing includes filtering the collected main helium blower vibration data, motor parameter data and environmental data, using a Butterworth low-pass filter to remove measurement noise, and the transfer function is specifically:

[0011]

[0012] Among them, ω c is the cutoff frequency; n is the filter order; s is the complex frequency variable, s = jω, where j is the imaginary unit and ω is the angular frequency;

[0013] Using the 3σ criterion, outliers in the data sequence are detected. The determination formula for outliers is:

[0014] |x i -μ|>3s d

[0015] Among them, x i is the i-th data point, μ is the data mean, s d is the data standard deviation;

[0016] Extract the time domain and frequency domain features that may correspond to the normal state and potential faults of the equipment from the preprocessed data. Let the extracted time domain feature vector be x T , the frequency domain eigenvector is x F , then the eigenvector is expressed as:

[0017] x=[x T ,x F].

[0018] As a preferred solution of the main helium blower state monitoring and early warning method of multi-source heterogeneous data fusion described in the present invention, wherein: the fusion of multi-source heterogeneous data includes: inputting the feature vector x=[x T ,x F ],

[0019] Assume that the input multi-source heterogeneous data is X = [v1, v2, ..., v m ], where m is the number of data sources, v k represents the feature vector of the kth data source, k = 1, ..., m; the attention weight is α = [α1, α2, ..., α m ], then the fusion feature F is expressed as:

[0020]

[0021] Among them, v k is the feature vector of the kth data source; α k is the attention weight of the kth data source, and the attention weight α k Gain through learning, satisfaction

[0022] The fusion network is trained in an end-to-end manner and outputs a comprehensive fault diagnosis feature vector.

[0023] As a preferred solution of the main helium blower state monitoring and early warning method of multi-source heterogeneous data fusion described in the present invention, wherein: the fault diagnosis model includes establishing a fault diagnosis model based on the fusion feature F, and the fault diagnosis model objective function is specifically:

[0024]

[0025] sty j (w T *F j +b)≥1-ξ j ,ξ j ≥0,j=1,2,...,N

[0026] Among them, w is the normal vector of the classification hyperplane, w T represents the result after the vector w is transposed, b is the bias term, C is the penalty coefficient, ξ j is the slack variable, y j is the label of the jth sample, F j is the fused feature vector of the jth sample, and N is the total number of samples.

[0027] As a preferred solution of the main helium blower state monitoring and early warning method of multi-source heterogeneous data fusion described in the present invention, the generator model includes a generator model that simulates future fault conditions trained using a generative adversarial network GAN, the GAN consists of a generator G and a discriminator D, and the objective function is:

[0028]

[0029] Among them, p data is the actual data distribution, p z is the noise distribution, x is the real sample, and z is the random noise input;

[0030] The generator model G is trained against the LSTM-based fault prediction model. The specific process is as follows:

[0031] Generator G generates simulated future fault data, predicts the real historical data and the generated simulated data through the LSTM prediction model, determines whether the LSTM prediction result is accurate based on the discriminator D, and optimizes the generator G and LSTM prediction model through the feedback of the discriminator D;

[0032] The hidden state h of LSTM t Calculated recursively as follows:

[0033] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0034] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0035] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0036] c t =f t ⊙c t-1 +i t ⊙tanh(W c ·[h t-1 ,x t ]+b c )

[0037] h t =o t ⊙tanh(ct )

[0038] Among them, f t 、i t , o t They are respectively the forget gate, input gate, and output gate, c t is the memory unit, c t-1 is the memory unit at time t-1, x t is the input at time t, h t is the hidden state at time t, h t-1 is the hidden state at time t-1, W and b are the weight matrix and bias vector, σ is the sigmoid function, and ⊙ represents element-by-element multiplication.

[0039] As a preferred solution of the main helium blower state monitoring and early warning method of multi-source heterogeneous data fusion described in the present invention, the adaptive optimization includes using a deep Q network DQN of a deep reinforcement learning algorithm to adaptively optimize the nuclear island layer diagnosis model and prediction model, and the Q value update formula of DQN is:

[0040] Q(st,a)←Q(st,a)+λ[r+γmax a′ Q(st′,a′)-Q(st,a)]

[0041] Among them, st is the current state, a is the action, r is the reward, λ is the learning rate, γ is the discount factor, and st′ is the next state;

[0042] The DQN agent learns the best optimization strategy by interacting with the system operating environment and automatically adjusts the parameters of the fault diagnosis model and LSTM prediction model;

[0043] Based on the fault diagnosis and prediction results, the system evaluates the equipment status in real time and issues fault warnings. The warning mechanism is triggered by the dynamic threshold model. The triggering condition of the dynamic threshold model is expressed as:

[0044] P(f|F)>θ(t)

[0045] Where f represents the fault state, P(f|F) is the fault probability under the given fusion feature F, and θ(t) is the warning threshold dynamically adjusted over time t;

[0046] The system continuously receives the diagnosis results and prediction data of the nuclear island layer, calculates P(f|F) using the trained fault diagnosis model, and uses an adaptive algorithm to dynamically adjust θ(t) according to the system operating status:

[0047] θ(t)=g(θ(t-1),ε(t),η(t))

[0048] Among them, g is the adjustment function, the specific form is:

[0049] g(θ(t-1),ε(t),η(t))=θ(t-1)+β1(1-ε(t))+β2η(t)

[0050] Among them, ε(t) is the historical warning accuracy, η(t) is the current system load, β1 and β2 are adjustment coefficients, and the initial threshold θ(0) is set based on historical data;

[0051] When P(f|F)>θ(t), an early warning is triggered. According to the difference between P(f|F) and θ(t), Δ=P(f|F)-θ(t), a judgment is made. When 0<Δ≤0.1, it is judged as a slight early warning; when 0.1<Δ≤0.3, it is judged as a moderate early warning; when Δ>0.3, it is judged as a serious early warning;

[0052] The system adopts corresponding warning strategies according to different warning levels. The warning strategies include increasing the monitoring frequency, starting deep diagnosis, and stopping equipment operation. At the same time, the system distributes information and manages warnings, generates and distributes warning information, records and analyzes warning events, and continuously optimizes warning models and threshold adjustment strategies.

[0053] As a preferred solution of the main helium blower state monitoring and early warning system with multi-source heterogeneous data fusion described in the present invention, the system includes:

[0054] The equipment layer includes a variety of sensors for collecting vibration data of the main helium blower, motor parameter data and environmental data; the equipment layer uses a deep neural network based on an attention mechanism to fuse multi-source heterogeneous data and output a comprehensive fault diagnosis feature vector;

[0055] The nuclear island layer establishes a fault diagnosis model based on the fusion features, and uses a generative adversarial network to train a generator model that simulates future fault conditions. The generator model is adversarially trained with a LSTM-based fault prediction model to improve the generalization ability of the prediction model for complex faults.

[0056] The power plant layer adopts a deep reinforcement learning algorithm to achieve adaptive optimization of the nuclear island layer diagnosis model and prediction model, and triggers fault warning based on the diagnosis and prediction results; the power plant layer is designed with a Web / App-based graphical interface to display the operating status of the main helium fan, fault diagnosis results, and warning information in real time.

[0057] Beneficial effects of the invention: The main helium blower status monitoring and early warning system with multi-source heterogeneous data fusion provided by the invention can comprehensively and accurately monitor the status of the main helium blower, timely warn of possible failures, and provide reliable decision support for the operation and maintenance of key equipment in nuclear power plants, thereby improving the safety, reliability and economy of nuclear power plants. The system integrates cutting-edge technologies such as industrial Internet of Things, deep learning, and reinforcement learning, and represents the development direction of intelligent monitoring and operation and maintenance of nuclear power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0059] Figure 1 A schematic flow chart of a main helium blower status monitoring and early warning method using multi-source heterogeneous data fusion provided in accordance with an embodiment of the present invention.

[0060] Figure 2 A schematic diagram of the layered architecture of a main helium blower status monitoring and early warning system with multi-source heterogeneous data fusion provided by an embodiment of the present invention.

[0061] Figure 3 A flow chart of a main helium blower status monitoring and early warning system with multi-source heterogeneous data fusion provided by an embodiment of the present invention.

[0062] Figure 4 A relationship diagram of adaptive optimization, fault warning and visualization at the power plant layer of a main helium fan status monitoring and early warning method using multi-source heterogeneous data fusion provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0064] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0065] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0066] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0067] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0068] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0069] Example 1, reference Figure 1 , which is the first embodiment of the present invention, and provides a main helium blower state monitoring and early warning method by multi-source heterogeneous data fusion, such as Figure 1 As shown, including:

[0070] S1: Collect the vibration data of the main helium fan, motor parameter data and environmental data, pre-process the data, use a deep neural network based on the attention mechanism to fuse multi-source heterogeneous data, and output a comprehensive fault diagnosis feature vector.

[0071] S2: A fault diagnosis model based on fusion features uses a generative adversarial network to train a generator model that simulates future fault conditions, and the generator model is adversarially trained with a LSTM-based fault prediction model.

[0072] S3: Use deep reinforcement learning algorithms to adaptively optimize the nuclear island layer diagnostic model and prediction model, and trigger fault warnings based on the diagnosis and prediction results.

[0073] The vibration data of the main helium blower includes collecting vibration signals in real time by arranging an acceleration sensor in the main helium blower; the motor parameter data includes installing voltage and current sensors at the motor end and using a speed sensor to obtain speed, torque and power parameters; the environmental data includes monitoring and collecting changes in temperature and humidity in the environment where the main helium blower is located.

[0074] The preprocessing includes filtering the collected vibration data of the main helium blower, motor parameter data and environmental data, using a Butterworth low-pass filter to remove measurement noise, and the transfer function is specifically:

[0075]

[0076] Among them, ω c is the cutoff frequency; n is the filter order; s is the complex frequency variable, s=jω, where j is the imaginary unit and ω is the angular frequency.

[0077] Using the 3σ criterion, outliers in the data sequence are detected. The determination formula for outliers is:

[0078] |x i -μ|>3s d

[0079] Among them, x i is the i-th data point, μ is the data mean, s d is the standard deviation of the data.

[0080] Extract the time domain and frequency domain features that may correspond to the normal state and potential faults of the equipment from the preprocessed data. Let the extracted time domain feature vector be x T , the frequency domain eigenvector is x F , then the eigenvector is expressed as:

[0081] x=[x T ,x F ].

[0082] The fusion of multi-source heterogeneous data includes: inputting the feature vector x=[x T ,x F ].

[0083] Assume that the input multi-source heterogeneous data is X = [v1, v2, ..., v m ], where m is the number of data sources, v k represents the feature vector of the kth data source, k = 1, ..., m; the attention weight is α = [α1, α2, ..., α m ], then the fusion feature F is expressed as:

[0084]

[0085] Among them, v k is the feature vector of the kth data source; α k is the attention weight of the kth data source, and the attention weight α k Gain through learning, satisfaction

[0086] The fusion network is trained in an end-to-end manner and outputs a comprehensive fault diagnosis feature vector.

[0087] The fault diagnosis model includes establishing a fault diagnosis model based on the fusion feature F, and the fault diagnosis model objective function is specifically:

[0088]

[0089] sty j (w T *F j +b)≥1-ξ j ,ξ j ≥0,j=1,2,...,N

[0090] Among them, w is the normal vector of the classification hyperplane, w T represents the result after the vector w is transposed, b is the bias term, C is the penalty coefficient, ξ j is the slack variable, y j is the label of the jth sample, F j is the fused feature vector of the jth sample, and N is the total number of samples.

[0091] The generator model includes a generator model that simulates future fault conditions and is trained using a generative adversarial network (GAN). The GAN consists of a generator G and a discriminator D, and the objective function is:

[0092]

[0093] Among them, p data is the actual data distribution, p z is the noise distribution, x is the real sample, and z is the random noise input.

[0094] The generator model G is trained against the LSTM-based fault prediction model. The specific process is as follows:

[0095] Generator G generates simulated future fault data, predicts the real historical data and the generated simulated data through the LSTM prediction model, judges whether the prediction result of LSTM is accurate according to the discriminator D, and optimizes the generator G and LSTM prediction model through the feedback of the discriminator D.

[0096] The hidden state h of LSTM t Calculated recursively as follows:

[0097] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0098] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0099] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0100] c t =f t ⊙c t-1 +i t ⊙tanh(W c ·[h t-1 ,x t ]+b c )

[0101] h t =o t ⊙tanh(c t )

[0102] Among them, f t 、i t , o t They are respectively the forget gate, input gate, and output gate, c t is the memory unit, c t-1 is the memory unit at time t-1, x t is the input at time t, h t is the hidden state at time t, h t-1 is the hidden state at time t-1, W and b are the weight matrix and bias vector, σ is the sigmoid function, and ⊙ represents element-by-element multiplication.

[0103] The adaptive optimization includes using a deep Q network DQN of a deep reinforcement learning algorithm to adaptively optimize the nuclear island layer diagnosis model and the prediction model. The Q value update formula of DQN is:

[0104] Q(st,a)←Q(st,a)+λ[r+γmax a′ Q(st′,a′)-Q(st,a)]

[0105] Among them, st is the current state, a is the action, r is the reward, λ is the learning rate, γ is the discount factor, and st′ is the next state.

[0106] The DQN agent learns the best optimization strategy through interaction with the system operating environment and automatically adjusts the parameters of the fault diagnosis model and LSTM prediction model.

[0107] Based on the fault diagnosis and prediction results, the system evaluates the equipment status in real time and issues fault warnings. The warning mechanism is triggered by the dynamic threshold model. The triggering condition of the dynamic threshold model is expressed as:

[0108] P(f|F)>θ(t)

[0109] Where f represents the fault state, P(f|F) is the fault probability under the given fusion feature F, and θ(t) is the warning threshold dynamically adjusted over time t.

[0110] The system continuously receives the diagnosis results and prediction data of the nuclear island layer, calculates P(f|F) using the trained fault diagnosis model, and uses an adaptive algorithm to dynamically adjust θ(t) according to the system operating status:

[0111] θ(t)=g(θ(t-1),ε(t),η(t))

[0112] Among them, g is the adjustment function, the specific form is:

[0113] g(θ(t-1),ε(t),η(t))=θ(t-1)+β1(1-ε(t))+β2η(t)

[0114] Among them, ε(t) is the historical warning accuracy, η(t) is the current system load, β1 and β2 are adjustment coefficients, and the initial threshold θ(0) is set based on historical data.

[0115] When P(f|F)>θ(t), an early warning is triggered. According to the difference Δ between P(f|F) and θ(t), Δ=P(f|F)-θ(t), a judgment is made. When 0<Δ≤0.1, it is judged as a slight warning; when 0.1<Δ≤0.3, it is judged as a moderate warning; when Δ>0.3, it is judged as a serious warning. The system adopts corresponding early warning strategies according to different early warning levels. The early warning strategies include increasing the monitoring frequency, starting deep diagnosis, and stopping equipment operation. At the same time, the system distributes information and manages early warnings, generates and distributes early warning information, records and analyzes early warning events, and continuously optimizes the early warning model and threshold adjustment strategy.

[0116] Example 2

[0117] like Figure 2-Figure 4As shown, the main helium blower status monitoring and early warning system with multi-source heterogeneous data fusion of the present invention adopts a layered architecture, including an equipment layer, a nuclear island layer and a power plant layer:

[0118] The equipment layer includes various sensors for collecting vibration data, motor parameter data and environmental data of the main helium blower. At the same time, data preprocessing and multi-source heterogeneous data fusion are realized in this layer.

[0119] The nuclear island layer includes the establishment and training of fault diagnosis model and fault prediction model. The diagnosis model identifies faults based on fusion features, and the prediction model uses adversarial training of GAN and LSTM to improve the prediction ability of complex faults.

[0120] The power plant layer includes using a deep reinforcement learning algorithm to adaptively optimize the model of the nuclear island layer to achieve a fault warning function, and providing a Web / App-based graphical interface to display the operating status of the main helium fan, fault diagnosis results and warning information in real time.

[0121] The relationship between adaptive optimization, fault warning and visualization in the power station layer is as follows: Figure 4 As shown:

[0122] 1. The adaptive optimization module receives the diagnosis and prediction results from the nuclear island layer and uses the DQN algorithm to optimize the model parameters.

[0123] 2. The fault warning module issues warnings based on the optimized model results and dynamic thresholds.

[0124] 3. The visualization module displays the warning results and the real-time status of the main helium blower to the operator through the Web / App interface.

[0125] Through the above implementation methods, the present invention realizes intelligent monitoring and early warning of the main helium blower of the nuclear power plant, and improves the operational reliability and safety of the equipment. The method integrates multi-source heterogeneous data, adopts advanced deep learning and reinforcement learning algorithms, and has strong adaptability and generalization ability.

[0126] 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 preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0127] Embodiment 3, the third embodiment of the present invention, is different from the first two embodiments in that:

[0128] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0129] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0130] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0131] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0132] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0133] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A main helium blower state monitoring and early warning method based on multi-source heterogeneous data fusion, characterized by: include, The vibration data of the main helium blower, motor parameter data and environmental data are collected and preprocessed. A deep neural network based on the attention mechanism is used to fuse multi-source heterogeneous data and output a comprehensive fault diagnosis feature vector. A fault diagnosis model based on fusion features uses a generative adversarial network to train a generator model that simulates future fault conditions, and the generator model is adversarially trained with a fault prediction model based on LSTM; A deep reinforcement learning algorithm is used to adaptively optimize the nuclear island layer diagnosis model and prediction model, and trigger fault warnings based on the diagnosis and prediction results.

2. The main helium blower state monitoring and early warning method based on multi-source heterogeneous data fusion according to claim 1 is characterized by: The main helium blower vibration data includes collecting vibration signals in real time by arranging an acceleration sensor in the main helium blower; The motor parameter data includes installing voltage and current sensors at the motor end and using a speed sensor to obtain speed, torque and power parameters; the environmental data includes monitoring and collecting temperature and humidity changes in the environment where the main helium blower is located.

3. The main helium blower state monitoring and early warning method based on multi-source heterogeneous data fusion according to claim 2 is characterized by: The preprocessing includes filtering the collected vibration data of the main helium blower, motor parameter data and environmental data, using a Butterworth low-pass filter to remove measurement noise, and the transfer function is specifically: Among them, ω c is the cutoff frequency; n is the filter order; s is the complex frequency variable, s = jω, where j is the imaginary unit and ω is the angular frequency; Using the 3σ criterion, outliers in the data sequence are detected. The determination formula for outliers is: |x i -μ|>3s d Among them, x i is the i-th data point, μ is the data mean, s d is the data standard deviation; Extract the time domain and frequency domain features that may correspond to the normal state and potential faults of the equipment from the preprocessed data. Let the extracted time domain feature vector be x T , the frequency domain eigenvector is x F , then the eigenvector is expressed as: x=[x T ,x F ]。 4. The main helium blower state monitoring and early warning method based on multi-source heterogeneous data fusion according to claim 3 is characterized by: The fusion of multi-source heterogeneous data includes: inputting the feature vector x=[x T ,x F ], Assume that the input multi-source heterogeneous data is X = [v1, v2, ..., v m ], where m is the number of data sources, v k represents the feature vector of the kth data source, k = 1, ..., m; the attention weight is α = [α1, α2, ..., α m ], then the fusion feature F is expressed as: Among them, v k is the feature vector of the kth data source; α k is the attention weight of the kth data source, and the attention weight α k Gain through learning, satisfaction The fusion network is trained in an end-to-end manner and outputs a comprehensive fault diagnosis feature vector.

5. The main helium blower state monitoring and early warning method based on multi-source heterogeneous data fusion according to claim 4 is characterized by: The fault diagnosis model includes establishing a fault diagnosis model based on the fusion feature F, and the fault diagnosis model objective function is specifically: s.t.y j (w T *F j +b)≥1-ξ j ,ξ j ≥0,j=1,2,...,N Among them, w is the normal vector of the classification hyperplane, w T represents the result after the vector w is transposed, b is the bias term, C is the penalty coefficient, ξ j is the slack variable, y j is the label of the jth sample, F j is the fused feature vector of the jth sample, and N is the total number of samples.

6. The main helium blower state monitoring and early warning method based on multi-source heterogeneous data fusion according to claim 5 is characterized by: The generator model includes a generator model that simulates future fault conditions and is trained using a generative adversarial network (GAN). The GAN consists of a generator G and a discriminator D, and the objective function is: Among them, p data is the actual data distribution, p z is the noise distribution, x is the real sample, and z is the random noise input; The generator model G is trained against the LSTM-based fault prediction model. The specific process is as follows: Generator G generates simulated future fault data, predicts the real historical data and the generated simulated data through the LSTM prediction model, determines whether the LSTM prediction result is accurate based on the discriminator D, and optimizes the generator G and LSTM prediction model through the feedback of the discriminator D; The hidden state h of LSTM t Calculated recursively as follows: f t =σ(W f ·[h t-1 ,x t ]+b f ) i t =σ(W i ·[h t-1 ,x t ]+b i ) the t =σ(W o ·[h t-1 ,x t ]+b o ) c t =f t ⊙c t-1 +i t ⊙tanh(W c ·[h t-1 ,x t ]+b c ) h t =o t ⊙tanh(c t ) Among them, f t 、i t , o t They are respectively the forget gate, input gate, and output gate, c t is the memory unit, c t-1 is the memory unit at time t-1, x t is the input at time t, h t is the hidden state at time t, h t-1 is the hidden state at time t-1, W and b are the weight matrix and bias vector, σ is the sigmoid function, and ⊙ represents element-by-element multiplication.

7. The main helium blower state monitoring and early warning method based on multi-source heterogeneous data fusion according to claim 6 is characterized by: The adaptive optimization Including, using the deep Q network DQN of the deep reinforcement learning algorithm to adaptively optimize the nuclear island layer diagnosis model and prediction model. The Q value update formula of DQN is: Q(st,a)←Q(st,a)+λ[r+γmax a′ Q(st′,a′)-Q(st,a)] Among them, st is the current state, a is the action, r is the reward, λ is the learning rate, γ is the discount factor, and st′ is the next state; The DQN agent learns the best optimization strategy by interacting with the system operating environment and automatically adjusts the parameters of the fault diagnosis model and LSTM prediction model; Based on the fault diagnosis and prediction results, the system evaluates the equipment status in real time and issues fault warnings. The warning mechanism is triggered by the dynamic threshold model. The triggering condition of the dynamic threshold model is expressed as: P(f|F)>θ(t) Where f represents the fault state, P(f|F) is the fault probability under the given fusion feature F, and θ(t) is the warning threshold dynamically adjusted over time t; The system continuously receives the diagnosis results and prediction data of the nuclear island layer, calculates P(f|F) using the trained fault diagnosis model, and uses an adaptive algorithm to dynamically adjust θ(t) according to the system operating status: θ(t)=g(θ(t-1),ε(t),η(t)) Among them, g is the adjustment function, the specific form is: g(θ(t-1),ε(t),η(t))=θ(t-1)+β1(1-ε(t))+β2η(t) Among them, ε(t) is the historical warning accuracy, η(t) is the current system load, β1 and β2 are adjustment coefficients, and the initial threshold θ(0) is set based on historical data; When P(f|F)>θ(t), an early warning is triggered. According to the difference between P(f|F) and θ(t), Δ=P(f|F)-θ(t), a judgment is made. When 0<Δ≤0.1, it is judged as a slight early warning; when 0.1<Δ≤0.3, it is judged as a moderate early warning; when Δ>0.3, it is judged as a serious early warning; The system adopts corresponding warning strategies according to different warning levels. The warning strategies include increasing the monitoring frequency, starting deep diagnosis, and stopping equipment operation. At the same time, the system distributes information and manages warnings, generates and distributes warning information, records and analyzes warning events, and continuously optimizes warning models and threshold adjustment strategies.

8. A system for monitoring and early warning the main helium blower state using the multi-source heterogeneous data fusion method according to any one of claims 1 to 7, characterized in that: The system includes, The equipment layer includes a variety of sensors for collecting vibration data of the main helium blower, motor parameter data and environmental data; the equipment layer uses a deep neural network based on an attention mechanism to fuse multi-source heterogeneous data and output a comprehensive fault diagnosis feature vector; The nuclear island layer establishes a fault diagnosis model based on the fusion features, and uses a generative adversarial network to train a generator model that simulates future fault conditions. The generator model is adversarially trained with a LSTM-based fault prediction model to improve the generalization ability of the prediction model for complex faults. The power plant layer adopts a deep reinforcement learning algorithm to realize adaptive optimization of the nuclear island layer diagnosis model and prediction model, and triggers fault warning based on the diagnosis and prediction results; The power station layer is designed with a Web / App-based graphical interface to display the operating status, fault diagnosis results, and early warning information of the main helium blower in real time.

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

10. 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 7 are implemented.

Images