Nuclear equipment reliability evaluation method and system based on intelligent identification of degradation state

Abstract

The application discloses a nuclear equipment reliability evaluation method and system based on intelligent identification of degradation states. The method comprises the following steps: acquiring key statistical information of different types of monitoring data, and establishing identification labels corresponding to different degradation states; constructing a fusion deep network model to intelligently identify the equipment performance degradation state, and performing information fusion on different modal data at the network layer; establishing a corresponding reliability evaluation model; establishing a mapping relationship between the degradation state identification labels and the different reliability evaluation models; combining the intelligent identification result of the equipment performance degradation state, and calculating the reliability and the residual life of the equipment under the current state according to the corresponding reliability evaluation model; evaluating the current nuclear equipment reliability and outputting the result. The application fully utilizes the advantages of intelligent algorithms to couple the nuclear equipment degradation state identification, the reliability model modeling and the reliability evaluation process, so that the complexity of the nuclear equipment reliability evaluation under different operation stages is reduced.

Description

Technical Field

[0001] This invention relates to the field of equipment reliability assessment technology, and specifically to a method and system for assessing the reliability of nuclear equipment based on intelligent identification of degradation status. Background Technology

[0002] Nuclear facilities have complex structures, and various types of malfunctions are inevitable during system operation. These malfunctions can range from minor issues affecting system stability and increasing maintenance costs to serious accidents posing radioactive hazards and severely threatening public safety and property. As nuclear facilities age, the likelihood of malfunctions in related nuclear equipment such as steam generators, control rod drive mechanisms, pressurizers, and main pumps gradually increases. Current research on nuclear equipment reliability assessment mainly falls into two categories: mechanistic model-based methods and data-driven methods. Mechanistic model-based methods (physical, chemical, etc.) generally start from the internal operating mechanisms of the target engineering system, establishing mathematical models that reflect equipment reliability and predicting degradation trends, thus obtaining relatively accurate prediction results by delving into the essence of the target system. However, this approach is often contradictory to the difficulty of establishing accurate mathematical models for dynamic systems, especially when the degradation process is complex and the mechanisms are not yet clear, such as in the operating environment of nuclear facilities under multi-physics coupling conditions. Establishing effective physical models is challenging, limiting their engineering application prospects in critical nuclear facility equipment. With the continuous development of artificial intelligence technology, data-driven methods, which do not require the establishment of detailed models, have become a hot topic in the field and are widely studied and applied. However, nuclear facilities operate in a multi-physics coupled environment for extended periods. The complexity of their data structures and the limited number of samples hinder their development. Summary of the Invention

[0003] The present invention proposes a method, system, and device for nuclear equipment reliability assessment based on intelligent identification of degradation state, which can at least solve one of the technical problems in the background art.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] A method for assessing the reliability of nuclear equipment based on intelligent identification of degradation states involves the following steps performed using computer equipment:

[0006] Acquire real-time monitoring data on the performance degradation status of nuclear equipment;

[0007] Based on the acquired monitoring data, the corresponding key statistical information is determined, and the number and range of degradation states are determined using the key statistical information. Corresponding identification category labels for different degradation states are established.

[0008] A fusion deep network model is constructed to intelligently identify the performance degradation state of equipment. The original performance degradation data of the core equipment is used as input, and information fusion of different modal data is performed in the network layer, with degradation state labels as output.

[0009] Establish corresponding reliability assessment models based on different performance degradation states, and establish a mapping relationship between degradation state identification tags and reliability assessment models;

[0010] By combining the intelligent identification results of equipment performance degradation status and calculating the reliability and remaining life of the equipment under the current state based on the corresponding reliability assessment model, the reliability of the current nuclear equipment is assessed and the results are output.

[0011] Furthermore, the real-time monitoring data on the performance degradation status of the nuclear equipment includes equipment vibration acceleration data, temperature data, flow rate data, pressure data, and current data.

[0012] Furthermore, the process of determining corresponding key statistical information based on the acquired monitoring data, using this key statistical information to determine the number and range of degradation states, and establishing corresponding identification category labels for different degradation states specifically includes:

[0013] The root mean square of steady-state data is obtained, and the probability interval (X-3σ, X+3σ) is obtained by calculating the mean X and standard deviation σ using the 3σ criterion. If the value exceeds this interval, the monitoring equipment is considered to have entered the initial degradation state.

[0014] By sequentially obtaining the range of different degradation states, the number of degradation states in the process leading to complete failure is determined;

[0015] The maximum information coefficient is used to measure the correlation between different types of data and degradation indicators;

[0016] By integrating the correlation calculation results of different data, the intervals of degradation state are divided through adaptive weights, and corresponding degradation state labels are set.

[0017] Furthermore, the method of using the maximum information coefficient to measure the correlation between different types of data and degradation indicators specifically includes:

[0018] Calculate mutual information between different variables Where a and b represent the performance degradation data and the data in the monitored health indicators, respectively, p(a,b) is the joint probability density, and p(a) and p(b) are the marginal probability densities, respectively.

[0019] Define a two-dimensional distribution partition δ(m,n), where the range of values ​​for performance degradation data and monitored health indicators is divided into intervals m and n. Finally, obtain the MI value corresponding to the two-dimensional distribution partition δ(m,n), and the normalized result is... The maximum information coefficient between different types of data and degradation indicators is calculated using max{θ(m,n)}.

[0020] Furthermore, the construction of a fusion deep network model for intelligent identification of device performance degradation specifically includes,

[0021] A one-dimensional convolutional neural network, including a one-dimensional convolutional kernel, a one-dimensional pooling layer, and a ReLU activation function channel, is constructed to perform feature extraction operations on all one-dimensional performance degradation data.

[0022] A two-dimensional convolutional neural network is constructed, which includes a two-dimensional pooling layer and a ReLU activation function channel to perform feature extraction operations on all two-dimensional performance degradation data.

[0023] A weighted connection layer is constructed to perform weighted fusion of high-dimensional sparse features extracted from different channels in the final network layer.

[0024] Construct a classification loss function with cross-entropy structure; use the Softmax function to evaluate the output.

[0025] Furthermore, the construction of the fused deep network model for intelligent identification of device performance degradation status specifically includes the following fused deep network model optimization process:

[0026] Network structure layer optimization involves testing the performance of different combinations of layer numbers and convolutional kernel sizes in the designed fusion deep network model, and selecting the optimal combination for initial training process optimization.

[0027] Batch standardization, which involves standardizing the data again after it is input into the hidden network, effectively prevents overfitting during model training. Its form is as follows:

[0028]

[0029] Wherein, E(x) i Var(x) represents the average weight of the batch of neurons. i ) represents the weight variance of the batch neurons. Different adjustment parameters are input to the neurons for adjustment. These adjustment parameters are learned through training and are used to reverse the activation after transformation, thereby enhancing the network's expressive ability.

[0030] It also includes optimizing the traditional Softmax classifier using a marginal increment approach. The specific loss optimization process is as follows:

[0031]

[0032] Where N represents the total number of training samples, L i X represents the loss of a specific neuron. i For the input neuron vector, Represents the corresponding weight. represents the angle between the weight vector and the neuron, and b represents the fixed boundary bias term. This structure can more significantly highlight the decision boundary between different degenerate states.

[0033] This also includes learning rate optimization using the Adam optimizer with a self-decreasing learning rate, the optimization process of which is as follows:

[0034]

[0035] Where ω represents the gradient v and s is the exponentially weighted moving average operation process, dω and dω 2 α is the learning rate for updates, and ε and β are tuning parameters. The default values ​​for ε, β1, and β2 are 10, ... -8 , 0.9 and 0.999, where α0 is the initial learning rate, N is the training epoch, and λ is the decay coefficient.

[0036] Furthermore, the S400 establishes corresponding reliability assessment models based on different performance degradation states, specifically including:

[0037] Based on the assigned performance degradation status, a corresponding reliability assessment model is established according to the key parameters of nuclear equipment, fault diagnosis records, fault and maintenance records, including the equipment reliability function calculation model, the equipment lifetime cumulative distribution function model, and the probability density function model.

[0038] The reliability function calculation model, equipment lifetime cumulative distribution function model, probability density function model, and the calculation methods for the remaining lifetime distribution function and probability density function are as follows:

[0039]

[0040]

[0041] Where R(t), F(t), and f(t) represent the reliability function calculation model, the lifetime cumulative distribution function model, and the probability density function model, respectively, and L = T - τ represents the remaining lifetime after time τ.

[0042] On the other hand, the present invention also discloses a nuclear equipment reliability assessment system based on intelligent identification of degradation state, comprising the following units,

[0043] The condition monitoring data collection module is used to monitor and acquire the degradation status information of nuclear equipment in real time from data from different sensors.

[0044] The intelligent degradation state identification module constructs a fusion deep network model to intelligently identify the device performance degradation state, using the original performance degradation data of the core device as input; inputting the performance degradation data of the core device into the model in the form of small batch data, fusing information on different modal data at the network layer, and using the established degradation state label as output;

[0045] The degradation state reliability assessment and remaining lifetime calculation module is used to establish corresponding reliability assessment models based on different performance degradation states, establish the mapping relationship between degradation state identification tags and reliability assessment models, and calculate the remaining lifetime.

[0046] The equipment maintenance decision support module is used to provide maintenance decision support for the current equipment based on the calculation results of the remaining life of the equipment.

[0047] Furthermore, the status monitoring data collection module specifically includes:

[0048] The data acquisition unit is used to collect different types of monitoring data information through sensors;

[0049] The sliding window processing unit is used to store window monitoring data information according to a set length, including the mean and variance information of the distribution characteristics of the data contained in the window.

[0050] The data storage unit is used to store data of different types of degradation states.

[0051] Furthermore, the intelligent degradation state identification module specifically includes:

[0052] The data normalization unit is used to normalize and preprocess input monitoring data with different types.

[0053] The data augmentation unit is used to divide long-term continuous time-domain samples into different samples and make full use of the same type of data samples from different devices.

[0054] The model training optimization unit optimizes the network structure layers by testing the performance of different combinations of layer numbers and convolutional kernel sizes in the designed fusion deep network model, and selecting the optimal combination for initial training process optimization; it performs batch standardization by standardizing the data again after inputting it into the hidden network to effectively prevent overfitting during model training; it optimizes the traditional Softmax classifier using a marginal increment method; and it optimizes the learning rate using the Adam optimizer with a self-decreasing learning rate.

[0055] The degradation state identification output unit is used to output the degradation state identification results to facilitate the corresponding reliability assessment model.

[0056] Furthermore, the degradation state reliability assessment and remaining lifetime prediction module specifically includes:

[0057] The reliability assessment model storage unit is used to establish and store corresponding reliability assessment models based on key parameters of nuclear equipment, fault diagnosis records, fault and maintenance records under different tagged degradation states.

[0058] The equipment reliability model includes the equipment reliability function calculation model, the product lifetime cumulative distribution function model, and the probability density function model;

[0059] The reliability assessment unit is used to assess the reliability of nuclear equipment under its current condition using the corresponding model.

[0060] The remaining lifetime calculation unit is used to calculate the remaining lifetime based on the reliability function model, the product lifetime cumulative distribution function model, and the probability density function model. The calculation methods for the remaining lifetime distribution function and probability density function are as follows:

[0061]

[0062]

[0063] Where R(t), F(t), and f(t) represent the reliability function calculation model, the product lifetime cumulative distribution function model, and the probability density function model, respectively, and L = T - τ represents the remaining lifetime of the core device after time τ.

[0064] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0065] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0066] As can be seen from the above technical solution, the present invention provides a nuclear equipment reliability assessment method and system based on intelligent identification of degradation state. It makes full use of the advantages of intelligent algorithms to couple the nuclear equipment degradation state identification, reliability model modeling and reliability assessment process, thereby reducing the complexity of using different reliability models for assessment at different degradation stages.

[0067] In summary, this invention innovates in the "data source" and "model source" aspects of the reliability assessment method. It fully utilizes the advantages of intelligent algorithms to couple the identification of nuclear equipment degradation status, reliability model modeling, and reliability assessment process, thereby reducing the complexity of using different reliability models for assessment at different degradation stages.

[0068] Specifically, this invention aims to innovate in terms of data and model sources for nuclear equipment reliability assessment methods. It combines intelligent identification results of equipment performance degradation states with corresponding reliability assessment models to calculate the reliability and remaining lifetime of the equipment in its current state, thereby assessing and outputting the reliability of the current nuclear equipment. Its advantages lie in rapidly establishing a mapping relationship between degradation state identification tags and reliability assessment models using intelligent algorithms; and in deeply coupling reliability model building and reliability assessment processes, thereby improving the efficiency of using reliability models for assessment under complex degradation processes. Attached Figure Description

[0069] Figure 1 A flowchart of a nuclear equipment reliability assessment method based on intelligent identification of degradation state provided in an embodiment of the present invention;

[0070] Figure 2 The flowchart illustrates the construction of a deep intelligent algorithm for identifying degradation states in the nuclear equipment reliability assessment method based on intelligent identification of degradation states provided by this invention.

[0071] Figure 3 This is a structural diagram of the overall implementation of the fusion intelligent recognition model provided by the present invention;

[0072] Figure 4 A schematic diagram of the structure of the nuclear equipment reliability assessment system based on deep intelligent identification of degradation state provided in this embodiment of the invention. Detailed Implementation

[0073] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0074] like Figure 1 As shown in this embodiment, the nuclear equipment reliability assessment method based on intelligent identification of degradation state includes the following steps:

[0075] Acquire real-time monitoring data on the performance degradation status of nuclear equipment;

[0076] Based on the acquired monitoring data, the corresponding key statistical information is determined, and the number and range of degradation states are determined using the key statistical information. Corresponding identification category labels for different degradation states are established.

[0077] A fusion deep network model is constructed to intelligently identify the performance degradation state of equipment. The original performance degradation data of the core equipment is used as input, and information fusion of different modal data is performed in the network layer, with degradation state labels as output.

[0078] Establish corresponding reliability assessment models based on different performance degradation states, and establish a mapping relationship between degradation state identification tags and reliability assessment models;

[0079] By combining the intelligent identification results of equipment performance degradation status and calculating the reliability and remaining life of the equipment under the current state based on the corresponding reliability assessment model, the reliability of the current nuclear equipment is assessed and the results are output.

[0080] Specifically, it includes the following steps:

[0081] S100 acquires real-time monitoring data on the performance degradation status of nuclear equipment, including but not limited to one-dimensional monitoring data such as vibration acceleration data, thermal parameter data (temperature, flow, pressure parameters, etc.), and two-dimensional monitoring data such as thermal imaging information of nuclear equipment.

[0082] S200: Obtain key statistical information from different types of monitoring data, use key statistical information to determine the number and range of degradation states, and establish corresponding identification category labels for different degradation states;

[0083] S200 specifically includes:

[0084] Step 201: Obtain the root mean square of the steady-state data. Use the 3σ criterion to calculate the probability interval (X-3σ, X+3σ) by calculating the mean X and standard deviation σ. If the value exceeds this interval, the monitoring equipment is considered to have entered a degraded state.

[0085] Step 202: Calculate the root mean square of different types of monitoring input data.

[0086] Step 203: Calculate the mutual information between different variables. Where a and b represent the performance degradation data and the data in the monitored health indicators, respectively, p(a,b) is the joint probability density, and p(a) and p(b) are the marginal probability densities, respectively.

[0087] Step 204: Define a two-dimensional distribution partition δ(m,n), where the range of values ​​for performance degradation data and monitored health indicators is divided into intervals m and n. Finally, obtain the MI value corresponding to the two-dimensional distribution partition δ(m,n), and the normalized result is...

[0088] Step 205 involves calculating the maximum information coefficient to measure the correlation between different types of data and degradation indicators.

[0089] Step 206: Integrate the correlation calculation results of different data, and divide the interval of degradation state by adaptively setting the correlation weight, and set the corresponding degradation state identification category label.

[0090] Step 207: Based on the initial stage degradation data, continue to use the 3σ criterion to iteratively calculate the mean X and standard deviation σ, successively obtaining the range of different degradation states until the equipment completely fails. Determine the number of identification category tags for each degradation state during the process leading to complete failure based on the calculation results.

[0091] S300, construct a fusion deep network model to intelligently identify the performance degradation state of the device, using the original performance degradation data of the core device as input; input the performance degradation data of the core device into the model in the form of small batch data, perform information fusion on the data of different modes in the network layer, and use the degradation state label as output;

[0092] The overall process of S300 is as follows: Figure 2 As shown, it mainly includes:

[0093] Step 301: Collect different types of monitoring data information through sensors. Using the original performance degradation data of the nuclear equipment as input, normalize the input monitoring data with different fluctuation ranges to facilitate intelligent algorithm analysis and retrieval; divide long-term continuous time-domain samples into different samples and make full use of data samples of the same type from different devices.

[0094] Step 302: Data processing using a sliding window model. Normalized data within a certain degradation state interval is truncated according to a specific time length. A translation operation is performed based on the truncated window to obtain key information from different monitoring data samples, including but not limited to the mean, variance, and other key information regarding the distribution characteristics of the data within the window. Data processed using different sliding window models is stored according to corresponding degradation state identification category labels. Based on actual equipment, the nuclear equipment performance degradation data is input into the model in small batches for subsequent fusion deep network training. This small batch size is equivalent to the batch size in traditional deep networks. Since the available data volume for nuclear facility monitoring is smaller compared to traditional data, the network setup is based on a small batch size.

[0095] Step 303: In the network layer, information fusion is performed on different modal data (one-dimensional vibration acceleration monitoring data, thermal monitoring data, two-dimensional thermal imaging monitoring data, etc.) to construct a network structure such as... Figure 3The system uses a fusion deep network model to identify performance status category labels; it constructs preprocessing operations for monitoring information of different types of performance degradation states; it constructs a one-dimensional convolutional neural network, including one-dimensional convolutional kernels, one-dimensional pooling layers, and ReLU activation function channels, to extract features from all one-dimensional performance degradation data; it constructs a two-dimensional convolutional neural network, including two-dimensional pooling layers and ReLU activation function channels, to extract features from all two-dimensional performance degradation data; it constructs a weighted connection layer to perform weight fusion on the high-dimensional sparse features finally extracted from different channels at the terminal network layer; the weight calculation method after fusion is as follows: w = σ[W(F 2d-max (x); F 1d-average (x))], where F 2d-max (x) represents the sparse features obtained through 2d convolutional max pooling, F 1d-average The sparse features are obtained through one-dimensional convolutional mean pooling, where w is the new fused feature. A classification loss function with a cross-entropy structure is constructed; the output is processed using the Softmax function; the network structure layers are optimized by testing the performance of different combinations of layer numbers and convolutional kernel sizes in the designed fused deep network model, selecting the optimal combination for initial training optimization; and batch normalization of information within neurons is performed. Wherein, E(x) i Var(x) represents the average weight of the batch of neurons. i ) represents the weight variance of the batch of neurons. Different adjustment parameters are input to the neurons for adjustment; these adjustment parameters are learned through training and used to inversely transform the activations after transformation, thereby enhancing the network's expressive power. Data is standardized again after being input into the hidden network to effectively prevent overfitting during model training; a classification loss function with cross-entropy is constructed; the Softmax function is used to optimize the output; and the traditional Softmax classifier is optimized using a marginal increment approach, in the form of:

[0096]

[0097] Where N represents the total number of training samples, L i X represents the loss of a specific neuron. i For the input neuron vector, Represents the corresponding weight. represents the angle between the weight vector and the neuron, and b represents the fixed boundary bias term. This structure can more significantly highlight the decision boundary between different degenerate states; it uses an Adam optimizer with a self-decreasing learning rate. Optimize the learning rate. Here, ω represents the gradient v, and s is an exponentially weighted moving average operation. dω and dω 2α is the learning rate for updates, and ε and β are tuning parameters. The default values ​​for ε, β1, and β2 are 10. -8 α0 is the initial learning rate, N is the training epoch, and λ is the decay coefficient.

[0098] Step 304: The performance of the deep convolutional model in identifying degraded state category labels is tested using 10-fold cross-validation. A preliminary performance degradation state identification model is trained based on the identification performance requirements. The trained model, combined with actual core device test data, outputs the category labels corresponding to the intelligent state identification results.

[0099] S400 establishes a corresponding reliability assessment model based on monitoring data in different performance degradation state category labels, thereby establishing a correspondence between degradation state identification labels and reliability assessment models;

[0100] The S400 specifically includes the following steps:

[0101] Step 401: Obtain the raw monitoring data from the different performance degradation status category labels, and ensure that the data does not contain any outliers and that there is no missing or imperfect data.

[0102] The S500 combines the intelligent identification results of equipment performance degradation status with the corresponding reliability assessment model to calculate the reliability and remaining life of the equipment in the current state, evaluate the reliability of the current nuclear equipment, and output the results.

[0103] The S500 specifically includes the following steps:

[0104] Step 501: Based on different tagged degradation states, establish corresponding reliability assessment models according to, but not limited to, key parameters of nuclear equipment, fault diagnosis records, fault and maintenance records, etc. Equipment reliability assessment models include, but are not limited to, equipment reliability function calculation models, product lifetime cumulative distribution function models, and probability density function models.

[0105] Step 502: Based on the reliability function calculation model, the product lifetime cumulative distribution function model, and the probability density function model, according to the following formula... and Calculate the distribution function and probability density function of the remaining lifetime, respectively.

[0106] Where R(t), F(t), and f(t) represent the reliability function calculation model, the product lifetime cumulative distribution function model, and the probability density function model, respectively. L = T - τ represents the remaining lifetime of the nuclear equipment after time τ.

[0107] This invention also provides a nuclear equipment reliability assessment system based on intelligent identification of degradation state. Figure 4 The nuclear equipment reliability assessment system based on intelligent identification of degradation state provided in this embodiment mainly includes:

[0108] The condition monitoring data collection module 201 is used to monitor and acquire the degradation status information of nuclear equipment in real time from data from different sensors.

[0109] The intelligent degradation state identification module 202 constructs a fusion deep network model to intelligently identify the device performance degradation state, using the original performance degradation multi-type data of the core device as input; inputting the performance degradation data of the core device into the model in the form of small batch data, performing information fusion on different modal data at the network layer, and using the established degradation state label as output;

[0110] The degradation state reliability assessment and remaining lifetime calculation module 203 is used to establish corresponding reliability assessment models based on different performance degradation states, establish the mapping relationship between degradation state identification tags and reliability assessment models, and calculate the remaining lifetime.

[0111] The equipment maintenance decision support module 204 is used to provide maintenance decision support for the current equipment based on the calculation results of the remaining life of the equipment.

[0112] As an optional implementation, the status monitoring data collection module 201 specifically includes:

[0113] The data acquisition unit is used to collect different types of monitoring data through sensors. This includes, but is not limited to, one-dimensional monitoring data such as vibration acceleration data, thermal parameter data (temperature, flow rate, pressure parameters, etc.), and two-dimensional monitoring data such as thermal imaging information of nuclear equipment.

[0114] The sliding window processing unit is used to store window monitoring data information according to a certain length, including but not limited to key information such as the mean and variance of the data contained in the window and the distribution characteristics of the data.

[0115] The data storage unit is used to store data in different degradation states.

[0116] As an optional implementation, the intelligent degradation state identification module 202 specifically includes:

[0117] The data normalization unit is used to normalize input monitoring data with different fluctuation ranges, making it convenient for intelligent algorithms to analyze and call upon.

[0118] The data augmentation unit is used to divide long-term continuous time-domain samples into different samples and make full use of the same type of data samples from different devices.

[0119] The model training optimization unit optimizes the network structure layers by testing the performance of different combinations of layer numbers and convolutional kernel sizes in the designed fusion deep network model, and selects the optimal combination for initial training process optimization; it performs batch standardization by standardizing the data again after inputting it into the hidden network to effectively prevent overfitting in model training; it optimizes the traditional Softmax classifier using a marginal increment method; and it optimizes the learning rate using the Adam optimizer with adaptive learning rate descent.

[0120] The degradation state identification output unit outputs the degradation state identification results to facilitate the corresponding reliability assessment model.

[0121] As an optional implementation, the degradation state reliability assessment and remaining lifetime prediction module 203 specifically includes:

[0122] The reliability assessment model storage unit is used to establish and store corresponding reliability assessment models based on different tagged degradation states, and on, but not limited to, key parameters of nuclear equipment, fault diagnosis records, fault and maintenance records. Equipment reliability models include, but are not limited to, equipment reliability function calculation models, product lifetime cumulative distribution function models, and probability density function models.

[0123] The reliability assessment unit is used to assess the reliability of nuclear equipment under its current condition based on the corresponding model.

[0124] The remaining useful life calculation unit is used to calculate the product's lifetime based on the reliability function model, the cumulative product lifetime distribution function model, and the probability density function model, according to the formula. and Calculate the distribution function and probability density function of the remaining lifetime, respectively.

[0125] Where R(t), F(t), and f(t) represent the reliability function calculation model, the product lifetime cumulative distribution function model, and the probability density function model, respectively. L = T - τ represents the remaining lifetime of the core device after time τ.

[0126] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0127] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0128] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the nuclear device reliability assessment methods based on intelligent identification of degradation states in the above embodiments.

[0129] It is understood that the system provided in the embodiments of the present invention corresponds to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant content can be referred to the corresponding parts of the above methods.

[0130] This application also provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, communication interface, and memory communicate with each other via the communication bus.

[0131] Memory, used to store computer programs;

[0132] When a processor executes a program stored in memory, it implements the aforementioned nuclear device reliability assessment method based on intelligent identification of degradation states, the method comprising:

[0133] Acquire real-time monitoring data on the performance degradation status of nuclear equipment;

[0134] Based on the acquired monitoring data, the corresponding key statistical information is determined, and the number and range of degradation states are determined using the key statistical information. Corresponding identification category labels for different degradation states are established.

[0135] A fusion deep network model is constructed to intelligently identify the performance degradation state of equipment. The original performance degradation data of the core equipment is used as input, and information fusion of different modal data is performed in the network layer, with degradation state labels as output.

[0136] Establish corresponding reliability assessment models based on different performance degradation states, and establish a mapping relationship between degradation state identification tags and reliability assessment models;

[0137] By combining the intelligent identification results of equipment performance degradation status and calculating the reliability and remaining life of the equipment under the current state based on the corresponding reliability assessment model, the reliability of the current nuclear equipment is assessed and the results are output.

[0138] The communication bus mentioned in the aforementioned electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc.

[0139] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0140] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0141] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0142] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid state disk (SSD)).

[0143] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0144] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0145] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing the reliability of nuclear equipment based on intelligent identification of degradation state, characterized in that, Includes the following steps, S1. Acquire real-time monitoring data on the performance degradation status of nuclear equipment; the real-time monitoring data on the performance degradation status of nuclear equipment includes equipment vibration acceleration data, temperature data, flow rate data, pressure data, and current data. S2. Based on the acquired monitoring data, determine the corresponding key statistical information, use the key statistical information to determine the number and range of degradation states, and establish corresponding identification category labels for different degradation states; specifically including, The root mean square of steady-state data is obtained, and the probability interval (X-3σ, X+3σ) is obtained by calculating the mean X and standard deviation σ using the 3σ criterion. If the value exceeds this interval, the monitoring equipment is considered to have entered the initial degradation state. By sequentially obtaining the range of different degradation states, the number of degradation states in the process leading to complete failure is determined; The maximum information coefficient is used to measure the correlation between different types of data and degradation indicators; By integrating the correlation calculation results of different data, the intervals of degradation state are divided through adaptive weights, and corresponding degradation state labels are set. S3. Construct a fusion deep network model for intelligent identification of equipment performance degradation status. Take the original performance degradation data of the core equipment as input in the form of small batch data, perform information fusion on different modal data in the network layer, and take the degradation status label as output. S4. Establish corresponding reliability assessment models based on different performance degradation states, and establish a mapping relationship between degradation state identification tags and reliability assessment models; the reliability assessment models include equipment reliability function calculation models, equipment lifetime cumulative distribution function models, and probability density function models; S5. Based on the intelligent identification results of equipment performance degradation status, and according to the corresponding reliability assessment model, calculate the reliability and remaining life of the equipment under the current state, assess the reliability of the current nuclear equipment, and output the results. Specifically, the construction of a fusion deep network model for intelligent identification of device performance degradation includes: A one-dimensional convolutional neural network, including a one-dimensional convolutional kernel, a one-dimensional pooling layer, and a ReLU activation function channel, is constructed to perform feature extraction operations on all one-dimensional performance degradation data. A two-dimensional convolutional neural network is constructed, which includes a two-dimensional pooling layer and a ReLU activation function channel to perform feature extraction operations on all two-dimensional performance degradation data. A weighted connection layer is constructed to fuse the high-dimensional sparse features extracted from different channels in the final network layer; the weights after fusion are calculated as follows: ,in These are sparse features obtained through 2D convolutional max pooling. It is a sparse feature obtained by one-dimensional convolutional mean pooling, where w is the new feature after fusion; Construct a classification loss function with cross-entropy structure; use the Softmax function to evaluate the output. The construction of a fusion deep network model for intelligent identification of device performance degradation also includes optimizing the traditional Softmax classifier using a marginal incremental approach.

2. The nuclear equipment reliability assessment method based on intelligent identification of degradation state according to claim 1, characterized in that: The method of using the maximum information coefficient to measure the correlation between different types of data and degradation indicators specifically includes, Calculate mutual information between different variables Where a and b represent performance degradation data and data from the monitored health indicators, respectively. Yes, the joint probability density. and These are the marginal probability densities; Define a two-dimensional distribution partition This involves dividing the performance degradation data and the range of values ​​for the monitored health indicators into intervals of m and n, ultimately yielding the corresponding two-dimensional distribution. The MI value, the normalized result is pass Calculate the maximum information coefficient between different types of data and degradation indicators.

3. The nuclear equipment reliability assessment method based on intelligent identification of degradation state according to claim 2, characterized in that: The construction of a fusion deep network model for intelligent identification of device performance degradation states specifically includes the following fusion deep network model optimization process: Network structure layer optimization involves testing the performance of different combinations of layer numbers and convolutional kernel sizes in the designed fusion deep network model, and selecting the optimal combination for initial training process optimization. Batch standardization, which involves standardizing the data again after it is input into the hidden network, effectively prevents overfitting during model training. Its form is as follows: in, The average weight of the batch of neurons. The weight variance of a batch of neurons is used to adjust different regulation parameters by inputting them into the neurons. These regulation parameters are learned through training and are used to reverse the activation after transformation, thereby enhancing the network's expressive ability. It also includes learning rate optimization using the Adam optimizer with a self-decreasing learning rate.

4. The nuclear equipment reliability assessment method based on intelligent identification of degradation state according to claim 1, characterized in that: The establishment of corresponding reliability assessment models based on different performance degradation states specifically includes... Based on the assigned performance degradation status, a corresponding reliability assessment model is established using information such as key parameters of nuclear equipment, fault diagnosis records, and fault and maintenance records. Based on the reliability function calculation model, the equipment lifetime cumulative distribution function model, and the probability density function model, the calculation methods for the remaining lifetime distribution function and probability density function are as follows: in, , , These represent the reliability function calculation model, the lifetime cumulative distribution function model, and the probability density function model, respectively. express The remaining lifespan after a certain point in time.

5. A nuclear equipment reliability assessment system based on intelligent identification of degradation state, used to implement the nuclear equipment reliability assessment method based on intelligent identification of degradation state as described in any one of claims 1-4, characterized in that: Includes the following units, The condition monitoring data collection module is used to monitor and acquire the degradation status information of nuclear equipment in real time from data from different sensors. The intelligent degradation state identification module constructs a fusion deep network model to intelligently identify the device performance degradation state. It takes the original performance degradation data of the core device as input in the form of small batch data, performs information fusion on different modal data in the network layer, and takes the degradation state label as output. The degradation state reliability assessment and remaining lifetime calculation module is used to establish corresponding reliability assessment models based on different performance degradation states, establish the mapping relationship between degradation state identification tags and reliability assessment models, and calculate the remaining lifetime. The equipment maintenance decision support module is used to provide maintenance decision support for the current equipment based on the calculation results of the equipment's remaining life.

6. The nuclear equipment reliability assessment system based on intelligent identification of degradation state according to claim 5, characterized in that: The status monitoring data collection module specifically includes: The data acquisition unit is used to collect different types of monitoring data information through sensors; The sliding window processing unit is used to window the monitoring data according to a set length and calculate the key information of the monitoring data within the window, including mean and variance information. The data storage unit is used to store data of different types of degradation states.

7. The nuclear equipment reliability assessment system based on intelligent identification of degradation state according to claim 6, characterized in that: The intelligent degradation state identification module specifically includes: The data normalization unit is used to normalize and preprocess input monitoring data with different types. The data augmentation unit is used to divide long-term continuous time-domain samples into different samples and make full use of the same type of data samples from different devices. The model training optimization unit optimizes the network structure layers by testing the performance of different combinations of layer numbers and convolutional kernel sizes in the designed fusion deep network model, and selecting the optimal combination for initial training process optimization; it performs batch standardization by standardizing the data again after inputting it into the hidden network to effectively prevent overfitting during model training; it optimizes the traditional Softmax classifier using a marginal increment method; and it optimizes the learning rate using the Adam optimizer with a self-decreasing learning rate. The degradation state identification output unit is used to output the degradation state identification results to facilitate the corresponding reliability assessment model.

8. The nuclear equipment reliability assessment system based on intelligent identification of degradation state according to claim 7, characterized in that: The degradation state reliability assessment and remaining lifetime prediction module specifically includes: The reliability assessment model storage unit is used to establish and store corresponding reliability assessment models based on key parameters of nuclear equipment, fault diagnosis records, fault and maintenance records under different tagged degradation states. The equipment reliability model includes the equipment reliability function calculation model, the product lifetime cumulative distribution function model, and the probability density function model; The reliability assessment unit is used to assess the reliability of nuclear equipment under its current condition using the corresponding model. The remaining lifetime calculation unit is used to calculate the remaining lifetime based on the reliability function model, the product lifetime cumulative distribution function model, and the probability density function model. The calculation methods for the remaining lifetime distribution function and probability density function are as follows: in, , , These represent the reliability function calculation model, the product lifetime cumulative distribution function model, and the probability density function model, respectively. express The remaining lifespan of the nuclear equipment after a certain time.

9. A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of the method as claimed in any one of claims 1 to 4.

10. A computer device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method as claimed in any one of claims 1 to 4.

Images