Intelligent operation and maintenance system for large pulsed power device

Abstract

The application discloses a kind of large-scale pulse power device intelligent operation and maintenance system, comprising: data acquisition layer, for real-time acquisition of the operation parameter of multiple sampling points on device at each time pulse discharge, constructs the data sequence of each sampling point independently;Data analysis processing layer is used to respectively establish grey model and solve each sequence, obtain the predicted life and model accuracy of each sampling point, and based on the double criterion early warning of predicted life and accuracy, when triggering early warning, locate fault area according to sampling point position information.The application realizes the on-line monitoring, trend prediction and accurate positioning of large-scale pulse power device.

Description

Technical Field

[0001] This invention relates to the field of operation and maintenance management of large pulse power devices, and specifically to an intelligent operation and maintenance system for large pulse power devices, used for online monitoring, evaluation, early warning and fault location of the operating status of large pulse power devices. Background Technology

[0002] Large pulse sources possess extremely high instantaneous power, and large pulse power systems (such as linear transformer drivers and Max generators) are typically composed of multiple pulse power modules connected in series. To improve output amplitude and suppress pulse wave reflection, these modules often employ precise timing discharge techniques. However, as system scale increases and operating time accumulates, the performance degradation of devices / materials leads to a deterioration in the overall system output characteristics (such as total voltage drop, waveform distortion, and timing irregularities), which can severely trigger cascading failures, causing significant losses or even safety accidents.

[0003] In large-scale high-energy pulsed power devices, especially those at the 60mA level, diagnostic systems are a core component. These systems are diverse, including optical, particle, and electromagnetic diagnostics, and are mostly non-standard, small-batch customized products. The high technical requirements and development challenges of these systems, coupled with a single supply chain, pose significant challenges to integration and control. How to effectively integrate highly customized equipment and components from multiple manufacturers and models, and how to design multi-functional control systems, have become urgent problems to be solved.

[0004] Furthermore, these devices involve data from multiple dimensions, including near-field radiation monitoring, target systems, and protection systems. This data is diverse, encompassing optical diagnostics, particle diagnostics, and electromagnetic diagnostics, and is largely non-standard. The integration and control of equipment from different manufacturers, of different models, and with highly customized components present significant challenges. Traditional control and management methods are insufficient to meet the needs of effective monitoring, data processing, fault early warning, and fault location for such complex systems.

[0005] Currently, there is no mature technical solution for the operation of large pulse power devices. Most methods involve setting fixed warning thresholds for test parameters and derating operation schemes. This results in the device's output power failing to reach the design power, making full-power operation difficult, and inadequate timeliness and sensitivity of warnings. Fixed warning thresholds cannot capture gradual performance trends, and in application scenarios where pulse power output data fluctuates, they offer little practical guidance.

[0006] Therefore, there is a need for an intelligent operation and maintenance system that can effectively integrate multiple devices, comprehensively process multi-dimensional data, predict performance degradation trends, and accurately locate faults, so as to intelligently assess, monitor, and predict the health status of large pulse power devices. Summary of the Invention

[0007] The purpose of this invention is to propose an intelligent operation and maintenance system for large-scale pulse power devices to solve the problems existing in the prior art, such as the difficulty in integrating customized equipment from multiple manufacturers and models, the inability of fixed early warning thresholds to capture gradual performance changes, and the lack of accurate fault location methods.

[0008] To achieve the above objectives, this invention proposes an intelligent operation and maintenance system for large-scale pulsed power devices, comprising: The data acquisition layer is used to collect the operating parameters of multiple sampling points on the device in real time during each pulse discharge of the large pulse power device, and to construct an independent measured data sequence of operating parameters for each sampling point. The data acquisition layer includes a device access adapter for connecting various devices from different manufacturers, models and communication protocols. The adapter implements protocol parsing and data format conversion based on the configuration file, and converts non-standard data into a format that the platform can process. The data analysis and processing layer is used to establish and solve gray models for each of the measured data sequences to obtain the development coefficient and gray action quantity of each measured data sequence. Based on the development coefficient, gray action quantity and preset performance parameter failure threshold, the layer calculates the number of predicted pulse discharges required for each sampling point to reach the critical state, thereby obtaining the predicted lifetime of each sampling point and calculating the average relative accuracy of each gray model. The data analysis and processing layer is also used to perform dual-criteria early warning based on the predicted lifetime and average relative accuracy of each sampling point. When any sampling point triggers an early warning, the layer locates the fault area based on the monitoring location information of that sampling point.

[0009] Optionally, the operating parameters acquired by the data acquisition layer include at least one of voltage parameters, current parameters, and discharge cycle; the voltage parameters include at least one of maximum voltage, minimum voltage, voltage anti-peak coefficient, and the time when the discharge voltage first crosses the zero axis; the current parameters include at least one of maximum current, minimum current, current anti-peak coefficient, and the time when the discharge current first crosses the zero axis. The data acquisition layer organizes the data from each sampling point into an original data matrix. :

[0010] in, This represents the parameter value of the k-th discharge at the j-th sampling point in the pulse power device. J Where n is the total number of sampling points, and n is the total number of data points in each sequence; The data analysis and processing layer establishes a grey model for each row, i.e., each sampling point sequence, in the original data matrix, as an independent univariate data sequence.

[0011] Optionally, before establishing the grey model, the data analysis and processing layer first performs a level ratio test on each measured data sequence to determine the validity of the data: Calculate the stage ratio of the j-th sampling point sequence at the k-th discharge. : = ,2 ; If the grade ratio falls within the acceptable coverage range ( , If the sequence is within the range of ), then it is considered valid and can be used for gray modeling. When the measured data sequence at the same sampling point is determined to be invalid N times consecutively, the system issues a data acquisition link abnormality warning signal and suspends the lifetime prediction output of the sequence, where N≥5.

[0012] Optionally, the data analysis and processing layer establishes a GM(1,N) grey model in the following manner: For the original data matrix Each row in the sequence, i.e., each sampling point sequence, is processed by first-order accumulation to generate an independent accumulation sequence for each sampling point. The accumulation sequences of all sampling points are then arranged row by row in the original sampling order to construct an accumulation matrix. ; The accumulated sequence generated at the j-th sampling point is denoted as... The formula for calculating its k-th element is:

[0013] in, This represents the original parameter value of the i-th discharge at the j-th sampling point. This represents the cumulative sum of the original parameters of the k discharges before the j-th sampling point, that is, the value of the cumulative sequence at the k-th discharge. Construct the nearest neighbor mean sequence of two adjacent discharge data points for the j-th sampling point as the background value: = ,

[0014] in, This represents the background value of the j-th sampling point sequence during the k-th discharge; The following differential equation is established for the gray model:

[0015] Where a is the development coefficient reflecting the development trend of the time response function, b is the gray action quantity reflecting the data change relationship, and both a and b are undetermined parameters.

[0016] Optionally, the data analysis and processing layer solves for the development coefficient a and the gray action quantity b in the following manner: Construct the matrix B and vector Y corresponding to the j-th sampling point:

[0017] in, and Let J and J represent the construction matrix and vector corresponding to the k-th discharge at the j-th sampling point, respectively, where 1≤j≤J and 2≤k≤n; Solve for a and b using the least squares method:

[0018] The time response function of the model is obtained, i.e., the model's predicted lifetime matrix:

[0019] in, This represents the predicted value of the cumulative sequence; The predicted values ​​are then subtracted to restore the original sequence, resulting in the predicted values ​​for the original sequence.

[0020] in, This represents the predicted value of the original data sequence.

[0021] Optionally, the data analysis and processing layer calculates the predicted lifetime in the following manner: Set performance parameter failure threshold ; Calculate the predicted number of pulse discharges K required to reach the critical state. j :

[0022] Calculate the predicted remaining lifetime T of the sampling points j :

[0023] Where t is the pulse discharge interval period.

[0024] Optionally, the data analysis and processing layer calculates the average relative accuracy of the gray model in the following manner: Calculate residuals :

[0025] Calculate relative error :

[0026] The average relative error is calculated using the moving weighted average method. :

[0027] Calculate the average relative accuracy : .

[0028] Optionally, the data analysis and processing layer uses dual criteria for early warning: Lifespan criterion: When the ratio of the running time of the module monitored at the sampling point to the predicted lifespan is greater than the preset lifespan threshold, an early warning is triggered; Accuracy criterion: When the average relative accuracy A warning is triggered when the accuracy is less than a preset threshold. When any of the above criteria is triggered, the data analysis and processing layer determines that the pulse power module where the sampling point is located is in an abnormal state and issues a warning signal to stop using it or to check it immediately. The fault location is specifically defined as the location of the pulse power module corresponding to the sampling point number that triggered the early warning, which is the location of the fault area.

[0029] Optionally, the data analysis and processing layer performs modeling, prediction, and accuracy evaluation on multiple operating parameters for the same sampling point to obtain multiple sets of predicted lifetimes and corresponding model accuracies; The remaining lifetime obtained from each operating parameter is weighted and averaged or the minimum value is taken to obtain the final comprehensive predicted lifetime. The overall model accuracy is obtained by averaging the model accuracy obtained from each operating parameter.

[0030] Optionally, the system also includes: The upper management layer provides management functions for system management, test planning, and test operation. It uses a role-based access control module to implement permission authentication and operation management for different roles such as system administrators, test planners, and test operators. The service layer provides data monitoring services, file services, page computing services, and computing services, and supports data flow and interaction between different layers of the platform. The test data measurement and control layer is used to connect oscilloscopes, probes, power control devices, triggering devices, and diagnostic package measurement devices to achieve accurate measurement and real-time control of test data.

[0031] The beneficial effects of this invention are as follows: (1) High-efficiency integration: Through layered architecture design and standardized interface protocols, the effective integration of multi-manufacturer, multi-model, and highly customized equipment and components is achieved, which improves the overall collaborative working capability of the device.

[0032] (2) Comprehensive data processing: It can comprehensively collect, process and analyze data from various diagnostic technologies, monitoring systems and other dimensions, providing rich data support for the evaluation of the device's operating status and performance optimization.

[0033] (3) Timely early warning and precise location: The early warning and fault location mechanism based on data analysis can promptly detect abnormal situations in the operation of the device and accurately locate the fault point, which greatly shortens the fault troubleshooting time and improves the reliability and availability of the device. Strong trend prediction capability: By accumulating and generating data, the random noise and impulse interference in the monitoring data are weakened, and the inherent monotonic trend of performance degradation is effectively captured.

[0034] (4) The core algorithms are matrix operations and exponential calculations, which have a light computational burden and are easy to implement on embedded systems or industrial control computers, meeting real-time requirements.

[0035] (5) Flexible management: The design of the multi-functional platform makes the management of the device more flexible and convenient. Users with different roles can perform corresponding operations and management according to their own needs, which improves work efficiency.

[0036] (6) Supports multi-dimensional comprehensive evaluation: By integrating the evaluation results of multiple parameters such as voltage, current, and time, the overall health status of the device can be reflected more comprehensively and accurately.

[0037] (7) If any power module in a large pulse power device fails, the entire device will fail. This patent can realize the coupled analysis and calculation of data from multiple sampling points, provide early warning of the device's health status and fault location, and avoid causing huge losses.

[0038] The present invention has other features and advantages, which will be apparent from or will be set forth in detail in the accompanying drawings and the following detailed description, which together serve to explain the particular principles of the invention. Attached Figure Description

[0039] The above and other objects, features and advantages of the present invention will become more apparent from the accompanying drawings, in which like reference numerals generally denote like parts.

[0040] Figure 1 This is a schematic diagram of the layered architecture of an intelligent operation and maintenance system for a large-scale pulse power device according to an embodiment of the present invention.

[0041] Figure 2This is a schematic diagram of the functional modules of the test data acquisition layer in an intelligent operation and maintenance system for a large pulse power device according to an embodiment of the present invention. Detailed Implementation

[0042] The invention will now be described in more detail with reference to the accompanying drawings. While preferred embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0043] This invention provides an intelligent operation and maintenance system for large-scale pulse power devices, such as... Figure 1 As shown, the intelligent operation and maintenance system for this large-scale pulsed power device adopts a five-layer hierarchical architecture, including an upper management layer, a data analysis and processing layer, a test data acquisition layer (i.e., the data acquisition layer), a service layer, and a test data measurement and control layer, to achieve integrated control and multi-dimensional data processing for a 60MA-level large-scale pulsed power device. The following is a detailed description of each functional layer in the system.

[0044] In this embodiment, the test data measurement and control layer is directly connected to the various measurement and control devices of the apparatus. It connects to oscilloscopes, probes, power control devices, triggering devices, and diagnostic package measurement devices to achieve accurate measurement and real-time control of test data. In this embodiment, this layer uses a centralized control module to uniformly control devices such as oscilloscopes, high-voltage power supplies, and CCDs. Through interface connections with these devices, it enables parameter setting, status monitoring, and operational control of the devices, ensuring that the equipment operates according to predetermined requirements.

[0045] In this embodiment, the experimental data acquisition layer is the foundation of the platform's data source. It is responsible for collecting the operating parameters of multiple sampling points on the device in real time during each pulse discharge of the large pulse power device, and constructing an independent measured data sequence of operating parameters for each sampling point.

[0046] For equipment and components from multiple manufacturers, of multiple models, and highly customized, this layer adopts the following methods to achieve equipment integration and data acquisition: Device Access Adapter: The data acquisition layer includes a device access adapter for connecting various devices from different manufacturers, models, and using different communication protocols. The adapter performs protocol parsing and data format conversion based on a configuration file, uniformly converting non-standard data into a format that the platform can process. Specifically, for oscilloscopes using the SCPI protocol, high-voltage power supplies using the Modbus protocol, and other customized devices using proprietary serial port protocols, the adapter converts the data format of various devices into the platform's internal standard format according to a pre-configured protocol template.

[0047] Equipment Management Database: Establish an equipment management database to uniformly manage equipment parameters, performance, status, and other information, facilitating platform access and control.

[0048] Data preprocessing: For the diverse and non-standard data collected, data preprocessing is performed first, including data cleaning, format conversion, and other operations, to convert the data into a unified format and provide standardized data input for subsequent data analysis and processing.

[0049] The operating parameters acquired by the data acquisition layer include at least one of voltage parameters, current parameters, and discharge cycle; the voltage parameters include at least one of maximum voltage, minimum voltage, voltage anti-peak coefficient, and the time when the discharge voltage first crosses the zero axis; the current parameters include at least one of maximum current, minimum current, current anti-peak coefficient, and the time when the discharge current first crosses the zero axis.

[0050] The data acquisition layer organizes the data from each sampling point into an original data matrix. :

[0051] in, This represents the parameter value of the k-th discharge at the j-th sampling point in the pulse power device. J denoted as the total number of sampling points, and n as the total number of data points for each sequence. The data used to construct this matrix must be data collection points from the same discharge branch.

[0052] like Figure 2 As shown, in this embodiment, the experimental data acquisition layer further includes the following functional modules: Centralized control module: Responsible for unified control of devices such as oscilloscopes, high-voltage power supplies, and CCDs. Through interfaces with these devices, it enables parameter setting, status monitoring, and operational control, ensuring the equipment operates according to predetermined requirements.

[0053] Experimental Data Management Module: This module provides functions such as browsing, querying, and uploading experimental data. Users can easily view historical experimental data, query the data they need based on different conditions, and upload new experimental data to the platform for storage and analysis.

[0054] The target practice test management module encompasses functions such as test resource management, test task management, and test status management. It rationally allocates and manages the resources required for target practice tests, formulates test task plans, and monitors the execution status of tests in real time to ensure the smooth conduct of the tests.

[0055] On-site management module: Primarily responsible for on-site monitoring of the test range and equipment status. Through sensors and monitoring equipment installed on-site, it acquires real-time environmental information and equipment operating status, promptly detecting anomalies and issuing early warnings.

[0056] In this embodiment, the data analysis and processing layer is used to establish and solve gray models for each of the measured data sequences to obtain the development coefficient and gray action amount of each measured data sequence. Based on the development coefficient, gray action amount and preset performance parameter failure threshold, the layer calculates the number of predicted pulse discharges required for each sampling point to reach the critical state, thereby obtaining the predicted lifetime of each sampling point, and calculating the average relative accuracy of each gray model.

[0057] Specifically, the data analysis and processing layer is primarily responsible for in-depth analysis of various types of collected data. This includes neutron flight time analysis and physical analysis diagnosis, providing support for device operation status evaluation and performance optimization through data mining and analysis. In this invention, the data analysis and processing layer also undertakes core algorithmic functions such as gray model establishment, solution, lifetime prediction, and accuracy evaluation.

[0058] The data analysis and processing layer establishes a grey model for each row of the matrix, i.e., each sampling point sequence, as an independent univariate data sequence.

[0059] Before establishing the grey model, the data analysis and processing layer first performs a level ratio test on each measured data sequence to determine the validity of the data. Calculate the stage ratio of the j-th sampling point sequence at the k-th discharge. : = ,2 ; If the grade ratio falls within the acceptable coverage range ( , If the sequence is within the range of ), then it is considered valid and can be used for gray modeling. To prevent sporadic interference, if the measured data sequence from the same sampling point is deemed invalid N times consecutively (e.g., N≥5), the system issues a data acquisition link anomaly warning signal and suspends the lifetime prediction output for that sequence. Modeling will automatically restart once the data is restored to validity. Alternatively, the accuracy weight of this evaluation can be reduced.

[0060] The data analysis and processing layer establishes a GM(1,N) grey model in the following manner: Constructing the cumulative matrix , for The (Accumulated Generating Operation, 1-AGO) matrix (j=1,2,…,J), i.e. For matrix The cumulative generation, for The matrix undergoes first-order accumulation to reduce randomness and enhance trend. Specifically: the original data matrix... For each row (i.e., each sampling point sequence) in the dataset, perform first-order accumulation generation processing to obtain an independent accumulation generation sequence for each sampling point; then arrange the accumulation generation sequences of all sampling points row by row in the original sampling order to construct an accumulation matrix. ; The accumulated sequence generated at the j-th sampling point is denoted as... (A one-dimensional row vector), the formula for calculating the k-th element is:

[0061] in, This represents the original parameter value of the i-th discharge at the j-th sampling point. This represents the sum of the original parameters of the k discharges before the j-th sampling point, that is, the value of the accumulated sequence at the k-th discharge. Its length is the same as the original sequence, still n elements, corresponding to the sum of the parameters of the k discharges before the j-th sampling point. The cumulative matrix of all sampling points (Two-dimensional matrix): Generate a sequence by accumulating the J sampling points. , … The rows of the matrix are arranged in order and then concatenated to obtain the complete summation matrix:

[0062] Construct the nearest neighbor mean sequence of two adjacent discharge data points for the j-th sampling point as the background value: = ,

[0063] in, This represents the background value of the j-th sampling point sequence during the k-th discharge; The following differential equation is established for the gray model:

[0064] In this context, superscripts 0 and 1 distinguish between measured data and cumulative generated quantities, t represents a continuous time variable, a is the development coefficient reflecting the development trend of the time response function, b is the gray action quantity reflecting the data change relationship, and both a and b are undetermined parameters.

[0065] The data analysis and processing layer solves for the development coefficient a and the gray action quantity b in the following manner: Construct the matrix B and vector Y corresponding to the j-th sampling point:

[0066] in, and Let J and J represent the construction matrix and vector corresponding to the k-th discharge at the j-th sampling point, respectively, where 1≤j≤J and 2≤k≤n; Solve for a and b using the least squares method:

[0067] The time response function of the model is obtained, i.e., the model's predicted lifetime matrix:

[0068] in, This represents the predicted value of the cumulative sequence; The predicted values ​​are then subtracted to restore the original sequence, yielding the predicted values ​​for the original sequence:

[0069] in, This represents the predicted value of the original data sequence.

[0070] The data analysis and processing layer calculates the predicted lifetime in the following manner: Set performance parameter failure threshold (For example, 95% of the initial value is considered the failure threshold). Calculate the predicted number of pulse discharges K required to reach the critical state. j :

[0071] Calculate the predicted remaining lifetime T of the sampling points j :

[0072] Where t is the pulse discharge interval period.

[0073] To verify the reliability of the model's predictions, the data analysis and processing layer calculates the average relative accuracy of the gray model in the following manner: Calculate residuals :

[0074] Calculate relative error :

[0075] The average relative error is calculated using the moving weighted average method. :

[0076] Calculate the average relative accuracy : .

[0077] Preferably, in practical applications, the data analysis and processing layer performs modeling, prediction, and accuracy evaluation for multiple operating parameters at the same sampling point to obtain multiple sets of predicted lifetimes and corresponding model accuracies; the remaining lifetimes obtained from each operating parameter are weighted and averaged (e.g., weights are assigned according to the sensitivity of each parameter to performance degradation or equal weighted averages are used) or the minimum value is taken to obtain the final comprehensive predicted lifetime; the model accuracies obtained from each operating parameter are averaged to obtain the comprehensive model accuracy.

[0078] The data analysis and processing layer uses dual criteria for early warning: Lifetime criterion: When the ratio of the running time of the module monitored at the sampling point to the predicted lifetime is greater than the preset lifetime threshold (e.g., the preset lifetime threshold is 95%), an early warning is triggered. Accuracy criterion: When the average relative accuracy When the accuracy is less than the preset accuracy threshold (e.g., the preset accuracy threshold is 95%), an early warning is triggered; if the accuracy criterion triggers, it indicates that the model's reliability is low or the dispersion of the test data is increasing, and the prediction results are unreliable, which also triggers an early warning.

[0079] When any of the above criteria is triggered, the data analysis and processing layer determines that the pulse power module where the sampling point is located is in an abnormal state and issues a warning signal to stop using it or to check it immediately. Fault Location: The location of the pulse power module corresponding to the sampling point number that triggers the early warning is the location of the fault area. Since the modeling and evaluation of this invention are based on the independent data sequence of each sampling point, when an early warning is triggered at a certain sampling point, the fault location can be accurately pinpointed to the pulse power module near that sampling point, thereby guiding maintenance personnel to quickly troubleshoot.

[0080] Ultimately, for the entire 60mA large-scale pulsed power device, for all... Each sampling branch performs the aforementioned independent modeling, evaluation, and fusion calculation. When the comprehensive predicted lifetime or comprehensive accuracy of a branch triggers the early warning criterion, the system can achieve early warning and precise physical location of the pulse power modules under that branch.

[0081] In this embodiment, the service layer provides various service support to the platform, including neutron display service, data monitoring service, page calculation service, file service, etc. These services ensure the stable operation of the platform and the effective flow of data, providing reliable support for upper-layer management and data analysis.

[0082] In this embodiment, the upper management layer is responsible for macro-management and decision-making of the entire device, providing management functions for system management, test planning, and test operation. The role-based access control module enables permission authentication and operation management for different roles such as system administrator, test planner, and test operator.

[0083] Specifically, the system administrator is responsible for system configuration, user management, and security policy settings; the test planner is responsible for developing test task plans, allocating test resources, and arranging test procedures; and the test operator is responsible for executing test operations, monitoring the test process, and recording test data. All roles work together to ensure the normal operation of the device and the smooth conduct of the tests.

[0084] In summary, the intelligent operation and maintenance system for large-scale pulsed power devices provided in this invention effectively solves the integration challenges of multi-manufacturer, multi-model, and highly customized equipment through a five-layer hierarchical architecture design and the organic combination of equipment access adapters. By establishing grey models for multiple sampling points on the device and applying dual criteria of predicted lifetime and average relative accuracy for each model, independent health assessment and precise fault location of each pulsed power module in the large-scale pulsed power device are achieved. When any sampling point triggers an alarm, the location of the corresponding power module can be directly locked based on the physical number of that sampling point, overcoming the inherent limitation of traditional solutions that cannot locate the fault source due to data averaging. Furthermore, based on the small-sample adaptability of grey theory, this invention effectively weakens random noise and electromagnetic interference in the pulsed discharge process through first-order cumulative generation. It achieves computationally lightweight online prediction using grey differential equations with fixed parameters and the least squares method, involving only matrix operations and exponential calculations throughout the process. It requires no complex iterations or large-scale training data, making it easy to deploy on embedded systems or industrial control computers, thus meeting the stringent real-time requirements of large-scale pulsed power devices. Furthermore, by employing a multi-parameter accuracy weighted fusion strategy, evaluation information from multiple dimensions such as voltage, current, and time is integrated. This not only avoids the risk of misjudgment under specific operating conditions due to a single parameter, but also fully leverages the complementary advantages of each parameter, further enhancing the reliability and comprehensiveness of the evaluation results.

[0085] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.

Claims

1. A smart operation and maintenance system for a large-scale pulsed power device, characterized in that, include: The data acquisition layer is used to collect the operating parameters of multiple sampling points on the device in real time during each pulse discharge of the large pulse power device, and to construct an independent measured data sequence of operating parameters for each sampling point. The data acquisition layer includes a device access adapter for connecting various devices from different manufacturers, models and communication protocols. The adapter implements protocol parsing and data format conversion based on the configuration file, and converts non-standard data into a format that the platform can process. The data analysis and processing layer is used to establish and solve gray models for each of the measured data sequences to obtain the development coefficient and gray action quantity of each measured data sequence. Based on the development coefficient, gray action quantity and preset performance parameter failure threshold, the layer calculates the number of predicted pulse discharges required for each sampling point to reach the critical state, thereby obtaining the predicted lifetime of each sampling point and calculating the average relative accuracy of each gray model. The data analysis and processing layer is also used to perform dual-criteria early warning based on the predicted lifetime and average relative accuracy of each sampling point. When any sampling point triggers an early warning, the layer locates the fault area based on the monitoring location information of that sampling point.

2. The intelligent operation and maintenance system for large-scale pulse power devices according to claim 1, characterized in that, The operating parameters acquired by the data acquisition layer include at least one of voltage parameters, current parameters, and discharge cycle; the voltage parameters include at least one of maximum voltage, minimum voltage, voltage anti-peak coefficient, and the time when the discharge voltage first crosses the zero axis; the current parameters include at least one of maximum current, minimum current, current anti-peak coefficient, and the time when the discharge current first crosses the zero axis. The data acquisition layer organizes the data from each sampling point into an original data matrix. : in, This represents the parameter value of the k-th discharge at the j-th sampling point in the pulse power device. J Where n is the total number of sampling points, and n is the total number of data points in each sequence; The data analysis and processing layer establishes a grey model for each row, i.e., each sampling point sequence, in the original data matrix, as an independent univariate data sequence.

3. The intelligent operation and maintenance system for large-scale pulse power devices according to claim 2, characterized in that, Before establishing the grey model, the data analysis and processing layer first performs a level ratio test on each measured data sequence to determine the validity of the data. Calculate the stage ratio of the j-th sampling point sequence at the k-th discharge. : = ，2 ； If the grade ratio falls within the acceptable coverage range ( , If the sequence is within the range of ), then it is considered valid and can be used for gray modeling. When the measured data sequence at the same sampling point is determined to be invalid N times consecutively, the system issues a data acquisition link abnormality warning signal and suspends the lifetime prediction output of the sequence, where N≥5.

4. The intelligent operation and maintenance system for large-scale pulse power devices according to claim 3, characterized in that, The data analysis and processing layer establishes a grey model in the following manner: For the original data matrix Each row, i.e., each sampling point sequence, is processed by first-order accumulation to generate an independent accumulation sequence for each sampling point. The accumulation sequences of all sampling points are then arranged row by row in the original sampling order to construct an accumulation matrix. ; The accumulated sequence generated at the j-th sampling point is denoted as... The formula for calculating its k-th element is: in, This represents the original parameter value of the i-th discharge at the j-th sampling point. This represents the cumulative sum of the original parameters of the k discharges before the j-th sampling point, that is, the value of the cumulative sequence at the k-th discharge. Construct the nearest neighbor mean sequence of two adjacent discharge data points for the j-th sampling point as the background value: = ， in, This represents the background value of the j-th sampling point sequence during the k-th discharge; The following differential equation is established for the gray model: Where a is the development coefficient reflecting the development trend of the time response function, b is the gray action quantity reflecting the data change relationship, and both a and b are undetermined parameters.

5. The intelligent operation and maintenance system for large-scale pulse power devices according to claim 4, characterized in that, The data analysis and processing layer solves for the development coefficient a and the gray action quantity b in the following manner: Construct the matrix B and vector Y corresponding to the j-th sampling point: in, and Let J and J represent the construction matrix and vector corresponding to the k-th discharge at the j-th sampling point, respectively, where 1≤j≤J and 2≤k≤n; Solve for a and b using the least squares method: The time response function of the model is obtained, i.e., the model's predicted lifetime matrix: in, This represents the predicted value of the cumulative sequence; The predicted values ​​are then subtracted to restore the original sequence, resulting in the predicted values ​​for the original sequence. in, This represents the predicted value of the original data sequence.

6. The intelligent operation and maintenance system for large-scale pulse power devices according to claim 5, characterized in that, The data analysis and processing layer calculates the predicted lifetime in the following manner: Set performance parameter failure threshold ; Calculate the predicted number of pulse discharges K required to reach the critical state. j : Calculate the predicted remaining lifetime T of the sampling points j : Where t is the pulse discharge interval period.

7. The intelligent operation and maintenance system for large-scale pulse power devices according to claim 6, characterized in that, The data analysis and processing layer calculates the average relative accuracy of the gray model in the following manner: Calculate residuals : Calculate relative error : The average relative error is calculated using the moving weighted average method. : Calculate the average relative accuracy : 。 8. The intelligent operation and maintenance system for large-scale pulse power devices according to claim 7, characterized in that, The data analysis and processing layer uses dual criteria for early warning: Lifespan criterion: When the ratio of the running time of the module monitored at the sampling point to the predicted lifespan is greater than the preset lifespan threshold, an early warning is triggered; Accuracy criterion: When the average relative accuracy An alert is triggered when the accuracy is less than a preset threshold. When any of the above criteria is triggered, the data analysis and processing layer determines that the pulse power module where the sampling point is located is in an abnormal state and issues a warning signal to stop using it or to check it immediately. The fault location is specifically defined as the location of the pulse power module corresponding to the sampling point number that triggered the early warning, which is the location of the fault area.

9. The intelligent operation and maintenance system for large-scale pulse power devices according to claim 8, characterized in that, The data analysis and processing layer performs modeling, prediction, and accuracy evaluation on multiple operating parameters of the same sampling point to obtain multiple sets of predicted lifetimes and corresponding model accuracies. The remaining life obtained from each operating parameter is weighted and averaged or the minimum value is taken to obtain the final comprehensive predicted life. The overall model accuracy is obtained by averaging the model accuracy obtained from each operating parameter.

10. The intelligent operation and maintenance system for large-scale pulse power devices according to claim 1, characterized in that, The system also includes: The upper management layer provides management functions for system management, test planning, and test operation. It uses a role-based access control module to implement permission authentication and operation management for different roles such as system administrators, test planners, and test operators. The service layer provides data monitoring services, file services, page computing services, and computing services, and supports data flow and interaction between different layers of the platform. The test data measurement and control layer is used to connect oscilloscopes, probes, power control devices, triggering devices, and diagnostic package measurement devices to achieve accurate measurement and real-time control of test data.

Images