High-accuracy phased array radar subarray power supply unit health management method

Abstract

The invention provides a high-accuracy phased array radar subarray power supply unit health management method, which comprises hard fault detection and soft fault detection, and realizes power supply unit hard fault detection based on voltage, current and temperature detection circuits in order to detect hard faults such as overvoltage, overcurrent, short circuit and over-temperature of a subarray power supply unit; in order to detect a soft fault of a sub-array power supply unit caused by the change of basic parameters of a component, a core parameter waveform threshold of the component is determined based on Gaussian process regression taking a square exponential covariance function, a Matern covariance function and a periodic covariance function as kernel functions, and the threshold is dynamically adjusted according to the noisy condition of the waveform. And forming a soft fault safety threshold region, detecting a core parameter waveform of the basic component in real time, and comparing the core parameter waveform with the soft fault safety threshold region to realize soft fault detection. Through combination of soft and hard faults, the hard faults can be rapidly detected, and meanwhile, the failure problem of the device can be accurately predicted through a soft fault detection technology, so that intervention is performed in advance.

Description

Technical Field

[0001] The invention belongs to the field of radar technology and relates to a high-accuracy phased array radar subarray power supply unit health management method. Background Art

[0002] At present, the health management of the power supply unit of the phased array radar subarray is mainly carried out based on the hard fault detection method. This detection method will only alarm the processing unit when the power supply unit has failed, such as when the detection voltage, current, temperature, etc. exceed the set fault threshold, to indicate that the power supply unit has failed or the components are damaged. This fault detection method cannot accurately evaluate or predict the imminent failure of the power supply unit when the components in the subarray power supply unit are on the verge of failure or soft faults occur. Therefore, it cannot meet the requirements of high-accuracy health management of the subarray unit. Some literatures propose to add digital modeling of components in the power supply unit on the basis of the existing hard fault detection method, and introduce digital twin technology to realize the monitoring of the characteristic parameters of the components, thereby realizing the monitoring of the soft faults of the power supply unit. Although this method has high accuracy, its modeling is relatively complex and the engineering workload is large, which is not conducive to industrial application and promotion. Summary of the invention

[0003] In order to improve the fault detection accuracy of phased array radar subarray power supply units, especially to improve the soft fault detection capability of basic components of subarray power supply units, thereby improving the power supply stability of phased array radar subarrays, and ultimately achieving early prevention of faults, early repair and replacement of components, and significantly reducing maintenance costs, the present invention proposes a high-accuracy health management method for phased array radar subarray power supply units.

[0004] In order to solve the above technical problems, the present invention proposes a high-accuracy phased array radar subarray power supply unit health management method, the technical solution includes:

[0005] The hard fault detection and soft fault detection of the phased array sub-array power supply unit are combined to improve the health management level of the sub-array power supply unit. The hard fault detection mainly relies on the input voltage and current, output voltage and current, temperature detection circuit and data processing unit to realize hard fault detection such as input overvoltage, input overcurrent, input undervoltage, output overvoltage, output short circuit overcurrent and overtemperature. The soft fault detection relies on the voltage and current sampling circuit to sample the voltage and current of the basic components in the power supply unit at a fixed time, and send a total of n sampled values ​​to the data processing unit. The sampled values ​​of 0.6n are used as the training set of the Gaussian process regression model to observe the voltage and current output waveforms of different basic components, and determine the square exponential covariance based on the waveform. The combination of function, Matern covariance function and periodic covariance function is combined with the model to estimate the soft fault safety threshold area of ​​the normal output waveform, and statistical characteristic extreme value processing is performed to obtain the soft fault detection threshold of the basic components of the power supply unit. In the soft fault detection stage, the data processing unit compares the actual output waveform statistical characteristic extreme value with the threshold and makes a judgment. When the soft fault threshold is exceeded, it is determined that the basic component has a hidden fault risk, and it is necessary to pay special attention and continuously monitor the value change. When the value change exceeds the threshold by more than 15%, it is replaced in advance if necessary. If the statistical characteristic value of the sampling value is within the soft fault detection threshold range, it is determined that there is no soft fault risk for this part of the power supply.

[0006] Furthermore, the Gaussian process regression model in the soft fault detection is constructed based on the square exponential covariance function, the Matern covariance function and the periodic covariance function as the kernel function, and the modeling process is as follows:

[0007] Step 1: The voltage and current sampling values ​​of the basic components in the power supply unit are used as the Gaussian process regression training data set, and the corresponding sampled values ​​to be tested are used as the actual test values;

[0008] Step 2: According to the changing trends of the output voltage and current waveforms of the basic components, select a combination of the square exponential covariance function, the Matern covariance function, and the periodic covariance function to determine the covariance function and the original value of the hyperparameter;

[0009] Step 3: Establish a Gaussian process regression detection model;

[0010] Step 4: Train the training data set and optimize the hyperparameter values;

[0011] Step 5: Get the initial threshold domain for soft fault detection;

[0012] Step 6: After completing the Gaussian process regression modeling, perform statistical characteristic extreme value processing on the initial threshold domain of soft fault detection to facilitate soft fault judgment. The specific steps are as follows:

[0013] Step 7: Perform variance processing on the training data and the actual detection sampling values. The corresponding calculation formula is as follows, where x is the sampling value, m is the mean of the sampling value, and v is the variance:

[0014] v=sqr(E(xm) 2 )

[0015] Step 8: Perform skewness processing on the training data and the actual detection sampling values. The corresponding calculation formula is as follows, where s is the skewness:

[0016] s=E[((xm) / v) 3 ]

[0017] Step 9: Perform centroid processing on the training data and actual detection sampling values. The corresponding calculation formula is as follows, where n is the number of sampling points and c is the skewness:

[0018] c=(∑ n i＝1 n*x i ) / (∑ n i＝1 x i )

[0019] The present invention has the following beneficial effects and advantages:

[0020] 1. The present invention provides a high-accuracy health management method for a phased array radar subarray power supply unit, which combines hard fault detection with soft fault detection. Compared with traditional hard fault detection, the fault detection accuracy of the phased array radar subarray power supply unit is improved. Compared with digital twin technology, it is conducive to industrial promotion and application, and significantly reduces the health management cost.

[0021] 2. The soft fault detection technology proposed in the present invention proposes the soft fault detection threshold of basic components in the power supply unit based on the Gaussian regression equation, which provides a basis for the soft fault detection of components, prevents the occurrence of hard faults of components, and reduces the risk of hard faults in the power supply unit of the phased array radar subarray. Compared with the digital twin detection technology, it does not require modeling of components, reduces the detection cost, and improves the detection speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a flow chart of health management of power supply unit of phased array radar subarray;

[0023] Figure 2 This is the structural framework diagram of the phased array radar subarray power supply unit; DETAILED DESCRIPTION

[0024] The present invention will be further described below in conjunction with the accompanying drawings.

[0025] The present invention provides a high-accuracy health management method for a phased array radar subarray power supply unit. The phased array subarray power supply unit mainly includes various DC-DC voltage converters, LDOs, detection circuits and data processing unit modules. The method mainly manages the health of the phased array radar subarray power supply unit based on hard fault detection and soft fault detection. The hard fault detection part includes input voltage and current, output voltage and current, and temperature simulation detection circuits. The detection of different levels of voltage, current and temperature inside the phased array radar subarray power supply unit is completed through voltage division, precision resistor current detection, and temperature-sensitive resistors. The detection is compared with a preset reference source, and the comparison result is sent to a single-chip microcomputer for processing. When the detected voltage, current and temperature exceed a set threshold value, when the level sent to the single-chip microcomputer is a high level, it is determined that a hard fault occurs in the power supply of this part. At this time, the single-chip microcomputer issues an instruction to trigger an enable switch to turn off the power supply of the faulty part. If the level sent to the single-chip microcomputer is a low level, it is determined that no hard fault occurs in the power supply at this time. Soft fault detection includes the voltage and current sampling circuit and data processing unit of the basic components (inductors, resistors, capacitors, power transformers, power switching devices and drive devices) in the power supply unit. The voltage and current sampling circuit samples the voltage and current of the basic components at a fixed time and sends the sampled values ​​to the data processing unit. 3 / 5 of the total number of sampling points n is used as the training data set for Gaussian process regression. The voltage and current output waveforms of different basic components are observed, and the combination of square exponential covariance function, Matern covariance function and periodic covariance function is determined based on the waveform. The normal output is estimated in combination with the model. The waveform soft fault safety threshold area is located, and statistical characteristic extreme value processing is performed to obtain the soft fault detection threshold of the basic components of the power supply unit. In the soft fault detection stage, the sampling values ​​of the voltage and current sampling circuits of the basic components are processed with statistical characteristic extreme values, and their soft fault detection thresholds are compared. When the soft fault threshold is exceeded, it is determined that the basic component has a hidden fault risk, and it is necessary to pay special attention and continuously monitor the value change. When the value change exceeds the threshold by more than 15%, it should be replaced in advance if necessary. If the statistical characteristic value of the sampling value is within the soft fault detection threshold range, it is determined that there is no soft fault risk for this part of the power supply.

[0026] In this embodiment, the structural framework of the phased array radar subarray power supply unit is as follows: Figure 2 shown.

[0027] Each level of DC-DC power supply modules and LDO are responsible for converting the input voltage into a variety of required voltages in the phased array radar subarray power supply unit.

[0028] DC-DC power supply module, LDO includes inductor, resistor, capacitor, switching device, drive part, transformer, etc.

[0029] The detection circuit mainly includes voltage, current and temperature detection circuits, which mainly detect the input and output voltage and current of each level of DC-DC power supply module and LDO to determine whether a hard fault occurs in the circuit.

[0030] The sampling circuit mainly includes voltage and current sampling, which mainly realizes the sampling of voltage and current values ​​of basic components, and estimates the soft fault detection threshold based on Gaussian process regression and statistical characteristic values ​​to determine whether a soft fault occurs in the circuit.

[0031] The data processing unit mainly includes a single-chip microcomputer and an algorithm module. The single-chip microcomputer is mainly used to process hard fault data and take actions on hard faults, and the algorithm module is mainly used to process soft faults and take actions on soft faults.

[0032] like Figure 1 As shown in the figure, the health management of the phased array radar subarray power supply unit is achieved through hard fault and soft fault detection. The steps are as follows:

[0033] Step 1: The phased array radar subarray power supply unit is started;

[0034] Step 2: Hard fault detection detects the voltages of each level in the power supply unit at a fixed time to determine whether the voltage is overvoltage or undervoltage;

[0035] Step 3: Hard fault detection detects the current of each circuit in the power supply unit at a fixed time to determine whether the current is overcurrent or short circuit;

[0036] Step 4: Hard fault detection detects the temperature of key areas in the power supply unit at fixed time intervals to determine whether the temperature is overheated;

[0037] Step 5: When an overvoltage, undervoltage, overcurrent, or overtemperature fault occurs, the MCU will be activated to start the protection circuit and restart continuously until the fault is cleared. When a short circuit fault occurs, the MCU will be activated to disconnect the power module of the faulty part and restore power supply until the fault is cleared.

[0038] Step 6: Soft fault detection samples the voltage and current of basic components in the power supply unit at fixed time intervals;

[0039] Step 7 sends the sampled voltage and current values ​​to the data processing unit for training, and estimates the initial threshold domain for soft fault detection;

[0040] Step 8 performs statistical eigenvalue processing on the soft fault detection initial threshold domain to obtain the soft fault detection threshold, including variance, skewness and centroid;

[0041] Step 9: During the detection phase, statistical characteristic value processing is performed on the detection sampling point values ​​and compared with the soft fault detection threshold, including variance, skewness and centroid;

[0042] Step 10: When a soft fault occurs, if it exceeds the threshold by 15%, it is prompted to replace or maintain the device. When it exceeds the threshold by 0-15%, it is continuously monitored. When it is within the threshold, it is determined that there is no risk of soft fault.

[0043] Through the above measures and after verification by multiple project projects, the present invention can meet the requirements of a hard fault accuracy rate greater than 99%, a soft fault accuracy rate greater than 95%, and a false alarm rate less than 5%.

[0044] The detection contents of the phased array radar subarray power supply unit are shown in Table 1:

[0045] Table 1

[0046]

[0047]

[0048] Aiming at the characteristics of complex use environment and high reliability requirements of phased array radar subarray power supply units, the present invention proposes a health management method combining hard fault detection and soft fault detection. In the hard fault detection part, the focus is on monitoring faults such as overvoltage, undervoltage, overcurrent, short circuit and overtemperature. In the soft fault detection part, the voltage and current of basic components are sampled and trained mainly based on the Gaussian regression equation, the initial threshold domain of soft fault detection is estimated, and statistical characteristic value processing is performed on it, including variance, skewness and centroid, to obtain the soft fault detection threshold. In the detection stage, the voltage and current of basic components are sampled and statistical characteristic value processing is performed on them, including variance, skewness and centroid, which is compared with the soft fault detection threshold, so as to make a judgment on the current soft fault.

Claims

1. A high-accuracy phased array radar subarray power supply unit health management method, characterized by: The hard fault detection of the phased array sub-array power supply unit is combined with the soft fault detection, wherein the hard fault detection includes input overvoltage, input overcurrent, input undervoltage, output overvoltage, output short circuit overcurrent and overtemperature, which are realized by input voltage and current, output voltage and current, temperature detection circuit and data processing unit; the soft fault detection relies on the voltage and current sampling circuit to sample the voltage and current of the basic components in the power supply unit at a fixed time, and send a total of n sampled values ​​to the data processing unit, and use 0.6n sampled values ​​as the training set of the Gaussian process regression model, observe the voltage and current output waveforms of different basic components, and determine the square exponential covariance function according to the waveform, The combination of Matern covariance function and periodic covariance function is combined with the model to estimate the soft fault safety threshold area of ​​the normal output waveform, and statistical characteristic extreme value processing is performed to obtain the soft fault detection threshold of the basic components of the power supply unit. In the soft fault detection stage, the data processing unit compares the actual output waveform statistical characteristic extreme value with the threshold and makes a judgment. When the soft fault threshold is exceeded, it is determined that the basic component has a hidden fault. The value change is monitored. When the value change exceeds the threshold by more than 15%, it is replaced in advance. If the statistical characteristic value of the sampling value is within the soft fault detection threshold range, it is determined that there is no soft fault risk for this part of the power supply.

2. According to claim 1, a high-accuracy phased array radar subarray power supply unit health management method is characterized by: The Gaussian process regression model is constructed based on the square exponential covariance function, the Matern covariance function and the periodic covariance function as kernel functions, and the modeling process is as follows: Step 1: The voltage and current sampling values ​​of the basic components in the power supply unit are used as the Gaussian process regression training data set, and the corresponding sampled values ​​to be tested are used as the actual test values; Step 2: According to the output voltage and current waveform change trends of the basic components, select a combination of square exponential covariance function, Matern covariance function and periodic covariance function to determine the covariance function and the original value of the hyperparameter; Step 3: Establish a Gaussian process regression detection model; Step 4: Train the training data set and optimize the hyperparameter values; Step 5: Get the initial threshold domain for soft fault detection; Step 6: After completing the Gaussian process regression modeling, perform statistical characteristic extreme value processing on the initial threshold domain of soft fault detection to facilitate soft fault judgment. The specific steps are as follows: Step 7: Perform variance processing on the training data and the actual detection sampling values. The corresponding calculation formula is as follows, where x is the sampling value, m is the mean of the sampling value, and v is the variance: v=sqr(E(x-m) 2 ) Step 8: Perform skewness processing on the training data and the actual detection sampling values. The corresponding calculation formula is as follows, where s is the skewness: s=E[((x-m) / v) 3 ] Step 9: Perform centroid processing on the training data and actual detection sampling values. The corresponding calculation formula is as follows, where n is the number of sampling points and c is the skewness: c=(∑ n i＝1 n*x i ) / (∑ n i＝1 x i )。

Images