Layered evaluation method for performance of laser inertial measurement unit considering non-equal priorities of indexes

Abstract

The invention discloses a laser inertial measurement unit performance hierarchical evaluation method considering index non-equal priorities, and belongs to the field of supervision and evaluation of laser inertial measurement units in aerospace inertial systems, and the method comprises the following steps: obtaining a predetermined initial performance hierarchical evaluation model of the laser inertial measurement units; the method comprises the following steps: establishing a hierarchical evaluation model by analyzing indexes related to the performance of the laser inertial measurement unit, and establishing a BRB-NEP model considering non-equal priorities of the indexes for the problem of index priority inconsistency; acquiring a training sample of a to-be-evaluated laser inertial measurement unit; optimizing parameters in the initial performance hierarchical evaluation model by using a training sample of a to-be-evaluated laser inertial measurement unit and adopting a preset optimization algorithm to obtain an optimized performance hierarchical evaluation model for the to-be-evaluated laser inertial measurement unit; and evaluating the performance of the to-be-evaluated laser inertial measurement unit by using the optimized performance hierarchical evaluation model. The performance evaluation precision of the laser inertial measurement unit can be improved, and the problems of rule explosion, difficulty in embedding expert knowledge and the like caused by numerous indexes are solved.

Description

Technical Field

[0001] The present invention belongs to the field of supervision and prediction of laser inertial assemblies in aerospace inertial systems, and particularly relates to a method for hierarchical evaluation of laser inertial assembly performance considering non-equal priorities of indicators. Background Technique

[0002] Laser inertial assemblies have many advantages such as complete autonomy, full navigation information, continuous navigation information, high precision, and small volume, and are an indispensable basic navigation method in rockets and missiles. Since rockets, missiles, etc. all perform extremely important tasks, such as manned spaceflight, target strikes, etc., this poses extremely high reliability requirements for rockets and missiles. Therefore, it is of great significance to supervise and evaluate the performance of laser inertial assemblies.

[0003] The performance evaluation of laser inertial assemblies has the following characteristics: 1) The value of a single unit is extremely high and its working life is extremely limited, so the test data is limited and a data-driven evaluation method cannot be used; 2) Laser inertial assemblies are composed of multiple types of devices, with multi-format, complex and coupled signals, and a mechanism modeling-based evaluation method cannot be used; 3) Due to the complexity of the system, experts' understanding of the system is also fuzzy. Therefore, it is necessary to fuse the limited test data and fuzzy expert knowledge to obtain a high-precision evaluation model. The Belief Rule Base (BRB) is proposed on the basis of integrating IF-THEN rules, expert knowledge, and evidence reasoning algorithms, and can effectively fuse limited test data and expert knowledge, making full use of limited multi-source information.

[0004] Using BRB to solve the performance evaluation problem of laser inertial assemblies, the following problems need to be further solved: (1) There are many indicators for laser inertial assemblies. In the BRB model, a large number of indicators will lead to problems such as rule combination explosion and difficulty in embedding expert knowledge; (2) There is a problem of inconsistent priorities among the indicators of laser inertial assemblies, and the traditional BRB model is only applicable to the case of equal priorities of indicators.

[0005] Therefore, how to overcome the above defects and accurately evaluate the performance state of laser inertial assemblies is an urgent problem to be solved. If the performance state of laser inertial assemblies can be accurately evaluated, it can provide a basis for their daily maintenance and selection during important tasks, and will have great practical value. Summary of the Invention

[0006] In order to solve the above problems existing in the prior art, the present invention provides a method for hierarchical evaluation of laser inertial assembly performance considering non-equal priorities of indicators. The technical problems to be solved by the present invention are realized through the following technical solutions: Obtain the initial performance hierarchical evaluation model of a pre-determined laser inertial unit; wherein, the initial performance hierarchical evaluation model is obtained by analyzing the indicators related to the performance of the laser inertial unit to construct a hierarchical evaluation model, and for the problem of inconsistent indicator priorities therein, establishing a BRB-NEP model considering non-equal priorities of indicators; the BRB-NEP model is implemented based on the belief rule base method; Obtain the training samples of the laser inertial unit to be evaluated; wherein, the training samples contain the data of each indicator in the initial performance hierarchical evaluation model, and the training samples are marked with performance labels, and the performance labels contain the performance level of the laser inertial unit and the confidence level corresponding to the performance level; Utilize the training samples of the laser inertial unit to be evaluated and adopt a preset optimization algorithm to optimize the parameters in the initial performance hierarchical evaluation model to obtain an optimized performance hierarchical evaluation model for the laser inertial unit to be evaluated; Utilize the optimized performance hierarchical evaluation model to evaluate the performance of the laser inertial unit to be evaluated.

[0007] Advantages of the present invention: Regarding the problem of laser inertial unit performance evaluation, based on the belief rule base theory, the present invention establishes a laser inertial unit performance hierarchical evaluation model considering non-equal priorities of indicators, and further realizes a laser inertial unit performance hierarchical evaluation method considering non-equal priorities of indicators. This method first obtains the initial performance hierarchical evaluation model of a pre-determined laser inertial unit; wherein, the initial performance hierarchical evaluation model is obtained by analyzing the indicators related to the performance of the laser inertial unit to construct a hierarchical evaluation model, and for the problem of inconsistent indicator priorities therein, establishing a BRB-NEP model considering non-equal priorities of indicators; the BRB-NEP model is implemented based on the belief rule base method; secondly, obtain the training samples of the laser inertial unit to be evaluated; wherein, the training samples contain the data of each indicator in the initial performance hierarchical evaluation model, and the training samples are marked with performance labels, and the performance labels contain the performance level of the laser inertial unit and the confidence level corresponding to the performance level; then, since the initial performance hierarchical evaluation model is given by expert knowledge, the ambiguity of expert knowledge makes the initial performance hierarchical evaluation model have certain errors. Therefore, utilize the training samples of the laser inertial unit to be evaluated and adopt a preset optimization algorithm to optimize the parameters in the initial performance hierarchical evaluation model to obtain an optimized performance hierarchical evaluation model for the laser inertial unit to be evaluated; finally, utilize the optimized performance hierarchical evaluation model to evaluate the performance of the laser inertial unit to be evaluated.

[0008] The laser inertial unit performance hierarchical evaluation model considering non - equal priorities of indicators provided by the present invention is essentially a hybrid - driven model that can integrate fuzzy expert knowledge and limited test data, realizing the full utilization of limited test data and fuzzy expert knowledge, and effectively improving the performance evaluation accuracy of the laser inertial unit. According to the hierarchical evaluation idea, the present invention designs a performance evaluation method for the laser inertial unit, allocating numerous indicators to different sub - models. Each sub - model has limited input indicators, which can not only solve problems such as rule explosion caused by numerous indicators and difficulty in embedding expert knowledge, but also each sub - model gives the evaluation results of each component, realizing the support of component performance results for system performance results, and making the evaluation results of the laser inertial unit have better interpretability. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 It is a schematic flow chart of a laser strapdown inertial unit performance hierarchical evaluation method considering non - equal priorities of indicators provided by an embodiment of the present invention; Figure 2 It is a schematic structural diagram of a hierarchical index system provided by an embodiment of the present invention; Figure 3 It is a schematic flow chart of parameter optimization of a hierarchical evaluation model based on an artificial bee colony algorithm provided by an embodiment of the present invention; Figure 4 It is a result graph of the zero - bias stability of the X - axis laser gyroscope of a certain laser strapdown inertial unit in the case of an embodiment of the present invention; Figure 5 It is a result graph of the zero - bias stability of the Y - axis laser gyroscope of a certain laser strapdown inertial unit in the case of an embodiment of the present invention; Figure 6 It is a result graph of the zero - bias stability of the Z - axis laser gyroscope of a certain laser strapdown inertial unit in the case of an embodiment of the present invention; Figure 7 It is the change situation of the utility RMSE value during the training process of two models in the case of an embodiment of the present invention; Figure 8 It is the evaluation result of the optimized hierarchical evaluation model for 800 test samples in the case of an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0010] The following further describes the present invention in detail with specific embodiments, but the implementation manners of the present invention are not limited thereto.

[0011] An embodiment of the present invention provides a laser inertial unit performance hierarchical evaluation method considering non - equal priorities of indicators. As Figure 1 shown, the method may include the following steps: S1, obtain the initial performance hierarchical evaluation model of the laser inertial unit determined in advance; In the embodiments of the present invention, performance analysis and research are pre - carried out on laser inertial units, and an initial performance hierarchical evaluation model suitable for all laser inertial units is constructed. Among them, the initial performance hierarchical evaluation model is obtained by analyzing the indicators related to the performance of the laser inertial unit to construct a hierarchical evaluation model, and establishing a BRB - NEP model considering non - equal priorities of indicators for the problem of inconsistent indicator priorities; the BRB - NEP model is implemented based on the belief rule base method; Among them, the determination process of the initial performance hierarchical evaluation model of the laser inertial unit includes steps A1 to A2: Step A1, analyze multiple indicators related to the performance of the laser inertial unit, and construct a hierarchical indicator system through indicator selection; Specifically, step A1 may include the following steps A11 to A13: Step A11, for the laser strapdown inertial unit composed of a laser gyroscope and a quartz accelerometer, through performance analysis, select multiple indicators characterizing historical information and test information, where the historical information includes the number of historical transports and the cumulative transport mileage; the test information includes the self - calibration test result, the self - aiming test result, and the self - detection test result; and determine the expression form of the self - detection test result based on the 0 - 1 output form, determine the expression form of the self - aiming test result based on the 0 - 1 output form, and determine the calibration parameters corresponding to the laser gyroscope and the quartz accelerometer respectively by determining the expression of the output error model corresponding to the laser gyroscope and the quartz accelerometer respectively, which are used to characterize the self - calibration test result; First, analyze the historical information, which includes the number of historical transports, the cumulative transport mileage, etc. of the laser inertial unit.

[0012] The inertial navigation system for rockets (i.e., the laser inertial unit) contains many high - precision instruments, including gyroscopes, accelerometers, high - precision frames, angle sensors, servo motors, etc. These instruments have extremely high requirements for the transportation and storage environment, etc. In particular, random vibrations and sharp turns during transportation all cause certain damage to high - precision instruments. Therefore, grass - roots users will record the number of historical transports and the cumulative transport mileage and other information.

[0013] Second, analyze the test information, which includes the self - calibration test result, the self - aiming test result, and the self - detection test result.

[0014] Since rockets and missiles have extremely high requirements for system reliability, the test of the high - precision inertial navigation system is one of the important test objects before rocket flight. In the present invention, taking the laser strapdown inertial unit composed of a laser gyroscope and a quartz accelerometer as an example, the test items are implemented after the laser strapdown inertial unit is powered on and starts to stabilize, mainly including self - calibration test, self - aiming test, and self - detection test.

[0015] The self-test is to check the qualification of the laser strapdown inertial unit point by point through a preset program and multiple detection points. When a certain point passes the inspection, its identifier output is 0; when the inspection fails, its identifier output is 1. If all identifier outputs are 0, it indicates that the self-test is qualified; if a certain identifier output is 1, it indicates that the self-test is unqualified. Therefore, the self-test result is a 0-1 output, that is: (1); In the above formula (1), represents the self-test result of the laser strapdown inertial unit.

[0016] The self-aiming test is used to test whether the aiming performance of the laser strapdown inertial unit is normal. Generally, three self-aiming tests are carried out. If the differences between the three self-aiming results are all within the threshold range, the self-aiming performance is normal; otherwise, the self-aiming performance is abnormal. From the perspective of the test method, the output of the self-aiming test result can be regarded as a 0-1 output, which is: (2); In the above formula (2), represents the self-aiming test result of the laser strapdown inertial unit; , , are the three self-aiming results respectively; is the self-aiming error threshold; represents finding the intersection.

[0017] The purpose of the self-calibration test is to calibrate the error model of the inertial instruments in the laser strapdown inertial unit. Below, the laser gyroscope and the quartz accelerometer are distinguished and analyzed separately. For the convenience of description, the X, Y, and Z-axis laser gyroscopes are respectively denoted as , , , and the X, Y, and Z-axis quartz accelerometers are respectively denoted as , , .

[0018] For the three-axis laser gyroscopes, their output error model is: (3); In the above formula (3), , , are respectively the rotational angle increment pulse numbers output by the laser gyroscopes , , ; , , are respectively the laser gyroscopes , , scale factor; , , are respectively the zero biases of the laser gyroscopes , , ; , , are respectively the rotational angular velocities of the carrier in the X, Y, and Z directions; , are respectively the installation errors of the laser gyroscopes about the Y and Z axes; , are respectively the installation errors of the laser gyroscopes about the X and Z axes; , are respectively the installation errors of the laser gyroscopes about the X and Z axes; represents the time.

[0019] The output error models of the quartz accelerometers in three axial directions are: (4); In the above formula (4), , , are respectively the output voltages of the quartz accelerometers , , ; , , are respectively the scale factors of the quartz accelerometers , , ; , , are respectively the zero biases of the quartz accelerometers , , ; , , are respectively the acceleration values of the carrier along the X, Y, and Z directions; , are respectively the installation errors of the quartz accelerometers , with the X axis; , are respectively the installation errors of the quartz accelerometers , with the Y axis, , They are quartz accelerometers respectively , and the installation error with the Z-axis.

[0020] The self-calibration of the laser strapdown inertial unit is to use the gravitational acceleration and the ground speed as the excitation, change the input excitation by flipping the inner frame of the laser inertial unit at multiple positions, and collect the outputs of the laser gyroscopes and quartz accelerometers, and finally solve the error model parameters. Specifically, The calibration parameters of the laser gyroscope include: the laser gyroscope , , bias , , , the laser gyroscope , , scale factor , , , the laser gyroscope installation errors around the Y and Z axes , ; the laser gyroscope installation errors around the X and Z axes , the laser gyroscope installation errors around the X and Z axes , .

[0021] The calibration parameters of the quartz accelerometer include: the quartz accelerometer , , bias , , , the quartz accelerometer , , scale factor , , , the quartz accelerometer , installation error with the X-axis , ; the quartz accelerometer , installation error with the Y-axis , , the quartz accelerometer , installation error with the Z-axis , .

[0022] Step A12: Based on the information determined by analyzing multiple indicators, indicator selection is performed, and finally, the 28-dimensional laser strapdown inertial unit indicators are determined, expressed as: (5); where, represents the set of 28-dimensional laser strapdown inertial unit indicators; represents the self-test test result; represents the self-aiming test result; represents the historical transportation times; represents the cumulative transportation mileage; represents the zero-bias stability of the X-axis laser gyroscope; represents the zero-bias stability of the Y-axis laser gyroscope; represents the zero-bias stability of the Z-axis laser gyroscope; represents the installation error of the X-axis laser gyroscope around the Z-axis; represents the installation error of the X-axis laser gyroscope around the Y-axis; represents the installation error of the Y-axis laser gyroscope around the Z-axis; represents the installation error of the Y-axis laser gyroscope around the X-axis; represents the installation error of the Z-axis laser gyroscope around the Y-axis; represents the installation error of the Z-axis laser gyroscope around the X-axis; represents the scale factor non-linearity of the X-axis laser gyroscope; represents the scale factor non-linearity of the Y-axis laser gyroscope; represents the scale factor non-linearity of the Z-axis laser gyroscope; represents the zero-bias stability of the X-axis quartz accelerometer; represents the zero-bias stability of the Y-axis quartz accelerometer; represents the zero-bias stability of the Z-axis quartz accelerometer; represents the installation error between the Y-axis quartz accelerometer and the X-axis; represents the installation error between the Z-axis quartz accelerometer and the X-axis; represents the installation error between the X-axis quartz accelerometer and the Y-axis; represents the installation error between the Z-axis quartz accelerometer and the Y-axis; represents the installation error between the X-axis quartz accelerometer and the Z-axis; represents the installation error between the Y-axis quartz accelerometer and the Z-axis; represents the scale factor stability of the X-axis quartz accelerometer; represents the scale factor stability of the Y-axis quartz accelerometer; represents the scale factor stability of the Z-axis quartz accelerometer; For Step A12, during the process of indicator selection, the following analysis is carried out: For historical information, since the transportation process greatly affects the usage accuracy and lifespan of high-precision instruments, the historical transportation times and cumulative transportation mileage

[0023] are selected as the indicators for evaluating the performance of the laser strapdown inertial unit. For the self-test results and self-aiming test results in the test information, both are 0-1 output type indicators. When or the output is 1, the performance of the laser strapdown inertial unit can be directly determined as poor; when and the output is 0, other indicators need to be combined for further determination. Therefore,

[0024] In the self-calibration test, for a laser gyroscope, the zero bias refers to the output value of the laser gyroscope under no input conditions. If the zero bias value is stable, only constant compensation is required. Therefore, zero bias stability is the key to measuring this indicator. Based on the above analysis, zero bias stability and and are selected as the indicators for evaluating the performance of the laser strapdown inertial unit. Among them, zero bias stability and and are the standard deviations of the zero biases and and of the laser gyroscopes and and respectively, and the calculation method is the existing method, which will not be elaborated here.

[0025] From the approximation process of the installation error of the laser gyroscope, taking as an example, the smaller the value, the smaller the approximation process error. Therefore, installation error and and and and are selected as the indicators for evaluating the performance of the laser strapdown inertial unit. In addition, a laser gyroscope is a rate gyroscope. For a rate gyroscope, the field pays more attention to the scale factor nonlinearity rather than the scale factor itself. The scale factor nonlinearity refers to the ratio of the maximum deviation value of the gyroscope output relative to the least squares fitting line to the maximum output amount within the input angular rate range. Taking the X-axis laser gyroscope as an example, the calculation method is: (6); In the above formula (6), represents the scale factor nonlinearity of the X-axis laser gyroscope; represents the calculated laser gyroscope scale factor nonlinearity; is the th fitting value calculated for the fitting curve; is the th output value of the laser gyroscope; is the maximum output value of the laser gyroscope; represents taking the absolute value; represents taking the maximum value. The scale factor nonlinearity can represent the magnitude of the relative error in the measurement of the laser gyroscope. Therefore, the scale factor nonlinearity , , are selected as the indicators for evaluating the performance of the laser strapdown inertial assembly. It can be understood that , represent the scale factor nonlinearities of the Y-axis and Z-axis laser gyroscopes respectively. The calculation process of the scale factor nonlinearity can be understood with reference to the related technology and will not be elaborated here.

[0026] For quartz accelerometers, the indicators used to evaluate performance include: zero-bias stability , , , installation error , , , , , , scale factor stability , , . It should be emphasized that, different from the laser gyroscope, the quartz accelerometer focuses on the scale factor stability.

[0027] Among them, zero-bias stability , , are the mean square deviations of the zero biases , , of the quartz accelerometer , , respectively; scale factor stability , , are the scale factor stabilities of the quartz accelerometer , , Scale factor and and Standard deviation.

[0028] Regarding zero-bias stability and and , and scale factor stability and and The calculation process can be understood with reference to related technologies and will not be elaborated here.

[0029] Therefore, based on the above analysis, the performance indicators of the 28-axis laser strapdown inertial assembly are finally selected, as shown in the above formula (5).

[0030] Step A13, according to the 28-axis laser strapdown inertial assembly indicators, stratify by instrument type to construct a five-layer hierarchical indicator system; After the above analysis, the present invention has selected a total of 28 indicators to evaluate the performance of the laser strapdown inertial assembly. If the 28 indicators are directly used to construct the BRB model, assuming that each indicator is divided into 3 levels, 3^28 rules will be formed, which will inevitably cause the problem of rule explosion and impose great pressure on computing power. In addition, assuming that experts only formulate rules for 2 or 3 indicators, expert knowledge can be better embedded. However, when the number of indicators reaches as many as 28, expert knowledge cannot determine the output level and corresponding confidence according to the indicators.

[0031] The above analysis shows that when there are many indicators, it will not only cause the problem of rule explosion, but also lead to the inability to embed expert knowledge. To solve this problem, the present invention constructs a hierarchical indicator system according to the needs of the hierarchical evaluation model. The indicator system of the laser strapdown inertial assembly can be stratified based on various factors, such as by instrument type, by functional subsystem, or by sensitive direction. Considering that it is relatively easy to fuse indicators for the same type of instrument, the method of stratifying by the same type of instrument is adopted.

[0032] In the laser strapdown inertial assembly, the most core components are the three-axis laser gyroscopes and quartz accelerometers, and the indicators of the laser gyroscopes and quartz accelerometers can be obtained. Therefore, the laser gyroscopes and quartz accelerometers are respectively used as important evaluation objects. Then the construction of the indicator system of the laser strapdown inertial assembly is as Figure 2 shown. Figure 2 In, the ellipsis under the Y-axis and Z-axis laser gyroscopes indicates that their indicator systems are the same as those of the X-axis laser gyroscope; the ellipsis under the X-axis and Y-axis quartz accelerometers indicates that their indicator systems are the same as those of the Z-axis quartz accelerometer. To simplify the length, Figure 2 in, the laser gyroscope is abbreviated as gyro, and the quartz accelerometer is abbreviated as accelerometer.

[0033] SeeFigure 2 In the hierarchical index system of the laser strapdown inertial assembly, the indexes from the fifth layer to the first layer are set from bottom to top; The indexes of the fifth layer include: 、 、 、 、 、 , and, 、 、 、 、 、 ; The indexes of the fourth layer include: the bias stability, scale factor nonlinearity and installation error evaluation results of the X-axis laser gyroscope; the bias stability, scale factor nonlinearity and installation error evaluation results of the Y-axis laser gyroscope; the bias stability, scale factor nonlinearity and installation error evaluation results of the Z-axis laser gyroscope; and, the bias stability, scale factor stability and installation error evaluation results of the X-axis quartz accelerometer; the bias stability, scale factor stability and installation error evaluation results of the Y-axis quartz accelerometer; the bias stability, scale factor stability and installation error evaluation results of the Z-axis quartz accelerometer; among which, the installation error evaluation result of the X-axis laser gyroscope is calculated based on the indexes of the fifth layer 、 ; the installation error evaluation result of the Y-axis laser gyroscope is calculated based on the indexes of the fifth layer 、 ; the installation error evaluation result of the Z-axis laser gyroscope is calculated based on the indexes of the fifth layer 、 ; the installation error evaluation result of the X-axis quartz accelerometer is calculated based on the indexes of the fifth layer 、 ; the installation error evaluation result of the Y-axis quartz accelerometer is calculated based on the indexes of the fifth layer 、 ; the installation error evaluation result of the Z-axis quartz accelerometer is calculated based on the indexes of the fifth layer 、 ; The indicators of the third layer include: the evaluation results of the X-axis laser gyroscope, the evaluation results of the Y-axis laser gyroscope, the evaluation results of the Z-axis laser gyroscope, and the evaluation results of the X-axis quartz accelerometer, the evaluation results of the Y-axis quartz accelerometer, and the evaluation results of the Z-axis quartz accelerometer; among them, the evaluation result of the X-axis laser gyroscope is calculated based on the indicators of the X-axis laser gyroscope in the fourth layer, namely zero bias stability, scale factor nonlinearity, and installation error evaluation results; the evaluation result of the Y-axis laser gyroscope is calculated based on the indicators of the Y-axis laser gyroscope in the fourth layer, namely zero bias stability, scale factor nonlinearity, and installation error evaluation results; the evaluation result of the Z-axis laser gyroscope is calculated based on the indicators of the Z-axis laser gyroscope in the fourth layer, namely zero bias stability, scale factor nonlinearity, and installation error evaluation results; the evaluation result of the X-axis quartz accelerometer is calculated based on the indicators of the X-axis quartz accelerometer in the fourth layer, namely zero bias stability, scale factor stability, and installation error evaluation results; the evaluation result of the Y-axis quartz accelerometer is calculated based on the indicators of the Y-axis quartz accelerometer in the fourth layer, namely zero bias stability, scale factor stability, and installation error evaluation results; the evaluation result of the Z-axis quartz accelerometer is calculated based on the indicators of the Z-axis quartz accelerometer in the fourth layer, namely zero bias stability, scale factor stability, and installation error evaluation results; The indicators of the second layer include: self-detection test results, self-aiming test results, historical transportation times, cumulative transportation mileage, evaluation results of the laser gyroscope assembly, and evaluation results of the quartz accelerometer assembly; among them, the evaluation result of the laser gyroscope assembly is calculated based on the indicators of the laser gyroscope in the third layer; the evaluation result of the quartz accelerometer assembly is calculated based on the indicators of the quartz accelerometer in the third layer; The indicators of the first layer include: the evaluation result of the laser strapdown inertial assembly, which is calculated based on all the indicators of the second layer.

[0034] That is to say, the evaluation process from bottom to top is: 1) From the fifth layer to the fourth layer: Based on the installation error 、 give the installation error evaluation result of the X-axis laser gyroscope; based on the installation error 、 give the installation error evaluation result of the Y-axis laser gyroscope; based on 、 give the installation error evaluation result of the Z-axis laser gyroscope, and based on the installation error 、 give the installation error evaluation result of the X-axis quartz accelerometer; based on the installation error 、 give the installation error evaluation result of the Y-axis quartz accelerometer; based on the installation error 、 Give the installation error evaluation results of the Z-axis quartz accelerometer; 2) From the fourth layer to the third layer: Integrate the three indicators of the zero-bias stability, scale factor nonlinearity, and installation error evaluation results of the X-axis laser gyroscope to obtain the evaluation results of the X-axis laser gyroscope; Integrate the three indicators of the zero-bias stability, scale factor nonlinearity, and installation error evaluation results of the Y-axis laser gyroscope to obtain the evaluation results of the Y-axis laser gyroscope; Integrate the three indicators of the zero-bias stability, scale factor nonlinearity, and installation error evaluation results of the Z-axis laser gyroscope to obtain the evaluation results of the Z-axis laser gyroscope; Integrate the three indicators of the zero-bias stability, scale factor stability, and installation error evaluation results of the X-axis quartz accelerometer to obtain the evaluation results of the X-axis quartz accelerometer; Integrate the three indicators of the zero-bias stability, scale factor stability, and installation error evaluation results of the Y-axis quartz accelerometer to obtain the evaluation results of the Y-axis quartz accelerometer; Integrate the three indicators of the zero-bias stability, scale factor stability, and installation error evaluation results of the Z-axis quartz accelerometer to obtain the evaluation results of the Z-axis quartz accelerometer; 3) From the third layer to the second layer: Integrate the evaluation results of the laser gyroscopes in three axes to obtain the evaluation results of the laser gyroscope assembly; Integrate the evaluation results of the quartz accelerometers in three axes to obtain the evaluation results of the quartz accelerometer assembly; At the same time, the self-detection test results, self-aiming test results, historical transportation times, and cumulative transportation mileage can be calculated; 4) From the second layer to the first layer: Integrate the self-detection test results, self-aiming test results, historical transportation times, cumulative transportation mileage, evaluation results of the laser gyroscope assembly, and evaluation results of the quartz accelerometer assembly to obtain the evaluation results of the laser strapdown inertial assembly.

[0035] In step A2, according to the hierarchical index system, the traditional BRB model is used for the evaluation below the highest level, and in view of the problem of inconsistent priorities of the highest-level indicators, a BRB-NEP model considering non-equal priorities of indicators is established for evaluation, thereby constructing a hierarchical evaluation model.

[0036] In the embodiment of the present invention, a hierarchical index system is first constructed to determine the logic of hierarchical evaluation calculation from bottom to top, and then a hierarchical evaluation model is constructed based on this hierarchical index system, and the constructed hierarchical evaluation model is used to specifically implement the calculation and processing of the hierarchical index system.

[0037] Step A2 may include: For the process of the lower-layer indicators obtaining the upper-layer indicators from the fifth layer to the fourth layer, from the fourth layer to the third layer, and from the third layer to the second layer in the hierarchical index system, a traditional BRB model is set as the corresponding belief rule base model; and for the evaluation process from the second layer to the first layer, in view of the problem of inconsistent priorities of the first-layer indicators, a BRB-NEP model considering non-equal priorities of indicators is established as the corresponding belief rule base model, thereby constructing a hierarchical evaluation model; among them, in the hierarchical index system, the priorities of the self-detection test results and the self-aiming test results are the same and are higher than the priorities of the remaining indicators.

[0038] Specifically, in the embodiment of the present invention, according to the constructed hierarchical index system and the corresponding evaluation process, a hierarchical evaluation model is constructed with the BRB model as the sub-model. This hierarchical evaluation model includes a total of 15 belief rule base sub-models.

[0039] In this hierarchical evaluation model, the evaluation starts from the fifth-layer indicators and proceeds layer by layer upward until the performance evaluation result of the first layer "laser strapdown inertial unit" is obtained. This hierarchical evaluation model can not only solve the problem of rule explosion, but also give the evaluation results of the intermediate links, making the evaluation results of the laser strapdown inertial unit have better interpretability.

[0040] In addition, in the above evaluation process, the evaluation of the second layer and below can be realized by using the traditional BRB model, and its models are respectively denoted as BRB1~BRB14. However, there is a problem of inconsistent indicator priorities in the evaluation of "the second layer → the first layer", so a BRB-NEP model considering non-equal priorities of indicators is adopted.

[0041] According to Figure 2 the constructed hierarchical index system and the corresponding evaluation process, a hierarchical evaluation model is constructed. This hierarchical evaluation model includes a total of 15 belief rule base models, as shown in Table 1 (for the sake of simplifying the space, the text in Table 1 is omitted).

[0042] Table 1 Hierarchical Evaluation Model

[0043] Next, the establishment and reasoning problems of the BRB-NEP model need to be solved.

[0044] In the comprehensive evaluation of the laser strapdown inertial unit, for the sake of convenience of description, the 6 indicators of the self-detection test result, the self-aiming test result, the historical transportation times, the cumulative transportation mileage, the evaluation result of the laser gyroscope assembly, and the evaluation result of the quartz accelerometer assembly are respectively denoted as , where , , , . and As an intermediate evaluation result, it is given in the form of an evaluation level and a confidence level.

[0045] Among the indicators In, And Are 0-1 output indicators. Such indicators have the "right of veto". When Or The output is 1, the performance of the laser strapdown inertial unit can be directly judged as poor; when Or The output is 0, and the performance evaluation result is given by synthesizing other indicators. From the above analysis, it can be seen that And Should have a higher priority, while other indicators are of a lower priority. Generalize this problem as: Suppose there are Indicators , among which the first Indicators Are 0-1 output and have a higher priority, and the remaining Indicators Have a lower priority, and it is necessary to solve the BRB modeling problem under different indicator priorities.

[0046] The traditional BRB model requires that all indicators be at the same priority. To solve the modeling problem under different indicator priorities, the present invention adds an "or rule" on the basis of the "and rule" of the BRB model.

[0047] Specifically, in the embodiment of the present invention, the process of establishing a BRB-NEP model considering non-equal priorities of indicators may include the following steps B1 to B4: Step B1, for the modeling problem of different indicator priorities in the highest level, add rules on the basis of the traditional BRB model , described as: (7); Among them, assume that there are Indicators , among which the first Indicators Are 0-1 output and have a higher priority, and the remaining Indicators Have a lower priority; Is the set of the remaining Indicators ; Represents the or rule; Represents the and rule; Is the Performance levels of the laser inertial unit; Is the evaluation result; is the rule weight of the rule; is the index weight of the , and are natural numbers greater than 0; Step B2, construct the remaining rules by comprehensively using AND rules and OR rules. Then the th rule is expressed as: (8); where are respectively the index reference values of the th rule for the th index; is the confidence level for the th rule for the th performance level; is the rule weight of the rule; is the total number of rules, is a natural number greater than 0; Since the OR rule is used, in order not to affect the model inference process, force ; In addition, the rule and the remaining rules are mutually exclusive. Therefore, the rule fusion principle can be determined. See Step B3 for details.

[0048] Step B3, force , and determine the rule fusion principle as: ; Analyzing formulas (7) and (8), it can be seen that: 1) For formula (7), for example, if the laser strapdown inertial unit is divided into 4 performance levels: excellent, good, medium, and poor, then the rule indicates that if any one of the indexes is 1, directly determine the performance as the worst level, and the confidence level of the worst level is 1; 2) For formula (8), on the premise that are all 0, comprehensively consider the remaining indexes to evaluate the performance level and confidence level of the laser strapdown inertial unit.

[0049] Step B4, determine the inference process of the BRB-NEP model as three steps: index matching, rule activation, and rule fusion, and determine the corresponding processing methods for each step.

[0050] ① The processing process of index matching includes: The indicators of the BRB-NEP model may come from different sources, have different formats, and different dimensions, making it impossible to directly fuse the indicators. Therefore, they need to be converted into a unified format and dimension.

[0051] For non-0-1 output indicators, the reference value of the th indicator and the th rule is denoted as ; for , when , determine the matching degree of the th rule as which is expressed as: (9); where: is the test value of the th indicator at time ; is the matching degree of the th rule; is the total number of rules; represents the reference value of the th indicator and the th rule; For 0-1 output indicators, that is, the mentioned above, in order not to affect subsequent rule activation and rule fusion, directly set the matching degree of such indicators to 0.5, which is expressed as: (10); where, represents the matching degree of at time for the th rule.

[0052] ② The process of rule activation processing includes: According to the matching degree of the indicators, some rules will be activated and some rules will not be activated. Moreover, due to the different matching degrees, the degree of rule activation is also different.

[0053] Therefore, first, calculate the matching degree of all indicators for the th rule as: (11); (12); where, represents the indicator weight of the th indicator; represents the relative indicator weight of the th indicator; In the previous text, the weight of the 0-1 output index was forced to be 0, that is ; the matching degree of the 0-1 output index was set to 0.5, that is . Substituting the above parameters into Equation (11), we can get: (13); Equation (13) means that the 0-1 index does not affect the matching degree of the remaining indexes to the th rule , indicating that the setting of the weight and matching degree of the 0-1 output index is reasonable.

[0054] Due to the inconsistent index matching degrees, the activation degrees of different rules are also inconsistent. The activation degree of the th rule is represented by the activation weight . Therefore, the following processing is carried out.

[0055] Combining the rule weight and the matching degree of the th rule, the activation weight corresponding to the th rule is: (14); where is the activation weight corresponding to the th rule at time ; represents the rule weight of the th rule; represents the matching degree of all indexes to the th rule at time .

[0056] ③ The processing process of rule fusion includes: Based on the activation weight and the confidence of the th rule, calculate the basic probability mass, which is expressed as: (15); where is the basic probability mass that the output result of the th rule at time is ; represents the empty set; is the confidence for the th rule and the th performance level; is the complete set of all performance levels; when is true, is the remaining basic probability mass at a moment, representing global ignorance; , and are two components of global ignorance; Since the index matching degrees of each input are different, the activated rules are also different. By fusing all the activated rules, the final evaluation result of the laser strapdown inertial unit can be obtained. Generally, the method of Evidential Reasoning (ER) is used for rule fusion. The ER algorithm includes an analytical algorithm and a recursive algorithm. Considering that the recursive ER algorithm has good traceability, therefore: Use the recursive ER algorithm for rule fusion to obtain the fusion result of the first pieces of evidence, expressed as: (16); where is the normalized basic probability mass of the output result after fusing the first rules; is the normalized basic probability mass of the output result after fusing the first rules; is the normalization parameter of the first rules; is the normalization parameter of the first rules; is the abbreviation of ; is the abbreviation of , and is the basic probability mass that the output result of the th rule at the moment is the th performance level; is the remaining basic probability mass of the th rule; is the remaining basic probability mass after fusing the first rules; and are two components of ; and are the remaining basic probability mass of the th rule 's two components; Based on the normalized basic probability mass, the confidence degrees of the laser strapdown inertial unit at different performance levels are obtained, expressed as: (17); where is an abbreviation of and is the output result after the moment rule fusion confidence; in the BRB-NEP model inference process, since the empty set has no corresponding physical meaning in engineering applications, the basic probability mass and confidence assigned to the empty set are uniformly set to 0; represents the output result after fusing all rules is the normalized basic probability mass; represents one of the components and is the remaining basic probability mass after fusing all rules; According to the fusion results of all rules, the output of the BRB-NEP model is: (18); Assume that the utility value of the performance level of the laser strapdown inertial unit is , then the utility value of the evaluation result is: (19); wherein, is the utility value of the evaluation result.

[0057] S2, obtain the training samples of the laser inertial unit to be evaluated; wherein, the training samples contain the data of each index in the initial performance stratification evaluation model, and the training samples are marked with performance labels, and the performance labels contain the performance level of the laser inertial unit and the confidence corresponding to the performance level; Specifically, a training sample contains the specific values of 28-dimensional laser strapdown inertial unit indexes. The performance labels can be determined in combination with expert experience.

[0058] S3, use the training samples of the laser inertial unit to be evaluated and adopt a preset optimization algorithm to optimize the parameters in the initial performance stratification evaluation model to obtain an optimized performance stratification evaluation model for the laser inertial unit to be evaluated; The initial performance stratification evaluation model is set according to expert knowledge. Due to the limitations and ambiguity of expert experience, there are certain errors in the parameters of the initial performance stratification evaluation model, and further optimization is required to obtain an accurate stratification evaluation model.

[0059] Specifically, S3 may include the following steps: S31. Establish a parameter optimization model for the initial performance stratification evaluation model with the goal of minimizing the root mean square error between the output value of the evaluation model utility and the actual value. The parameters to be optimized include: the weights of each index, the weights of the rules, and the confidence levels of the performance levels corresponding to the th rules; (20); In formula (20), is the objective function; is the sample number; is the number of samples; is the st sample; is the model evaluation utility value of the th sample; is the true utility value of the th sample; is the index weight of the th index; is the rule weight of the

[0060] th (21); For the meanings of each parameter, please refer to the relevant descriptions above.

[0061] S32. According to the optimization goal, use the training sample data of the laser inertial unit to be evaluated, and adopt the artificial bee colony algorithm to optimize the parameters in the initial performance stratification evaluation model to obtain an optimized performance stratification evaluation model for the laser inertial unit to be evaluated.

[0062] The present invention combines the problem of optimizing the parameters of a hierarchical evaluation model with the Artificial Bee Colony (ABC) algorithm. First, the model parameters are encoded as the positions of artificial bees, and the problem of parameter optimization is transformed into the problem of searching for the optimal positions of artificial bees. Second, according to the parameter optimization objective, a fitness function is set to unify the evaluation criteria for artificial bees and model parameters.

[0063] Specifically, according to the artificial bee encoding method, the fitness function setting method, and in combination with the ABC algorithm principle, a parameter optimization process for the hierarchical evaluation model based on the ABC algorithm is formulated. Please refer to Figure 3 . Specifically, it includes: Step a1, encoding the positions of artificial bees and setting the fitness function; Encoding the position of the th artificial bee as the parameter to be optimized, that is: (22); wherein, is the position encoding of the th artificial bee; is the position dimension.

[0064] The fitness function is used to evaluate the quality of the positions of artificial bees. Combining with the background of parameter optimization for the hierarchical evaluation model, the fitness function is used to evaluate the quality of model parameters. According to the meaning of the fitness function, the larger the fitness value, the better the model parameters. Based on Equation (20), the fitness function is set as: (23); wherein, is the fitness function of the position of the artificial bee; Step a2, initializing the positions of artificial bees, that is, setting the initial positions of artificial bees as; (24); wherein, represents the position encoding of the th artificial bee in dimension ; represents the lower limit of the position interval of the artificial bee in dimension ; represents the upper limit of the position interval of the artificial bee in dimension ; represents a random number between (0, 1); Step a3, classifying leading bees and following bees according to the fitness function; After all artificial bees have been initialized in position, a fitness ranking is performed. The th artificial bees with higher fitness are defined as leading bees, and the remaining artificial bees are set as following bees; Step a4: The leading bee passes the fitness value of its own position to the follower bees through the waggle dance. According to the fitness function of the leading bee, calculate the selection probability of the follower bees for the leading bee; (25); where, represents the selection probability of the follower bee for the leading bee; represents the coding index of the leading bee; represents the number of leading bees; represents the fitness value of the leading bee; Step a5: After the follower bee selects a leading bee, the follower bee searches for a position in the same way as the leading bee.

[0065] The position search methods include the following 3 types, and one method is randomly selected from them each iteration; (26); where, represents the position after the th artificial bee searches in dimension ; represents the position of the th artificial bee in dimension ; represents a random number between (0, 1); represents a random number between (0, 1); represents a random number between (0, 1); represents a random number between (0, 1); represents the position of the population-optimal artificial bee in dimension ;

[0066] Step a6: Judge whether there exists a situation where the search times of a certain position reach the threshold and the fitness does not increase significantly; if it exists, execute Step a7, if not, execute Step a5; Step a7: The corresponding artificial bee is transformed into a scout bee, and its position is randomly initialized again, and then it is transformed into a follower bee; (27); Step a8: Judge whether the maximum search times have been reached; if so, execute Step a9, if not, execute Step a5; The maximum search times are generally determined through repeated attempts according to the convergence situation of the optimization process.

[0067] Step a9: Output the optimal position, and decode the optimal position into the optimal parameter value. The decoding method is the inverse process of the artificial bee position coding method; Through the processing of S3, the parameters of the initial performance hierarchical evaluation model pre-constructed for all laser inertial units can be optimized using the training samples of the known performance levels of the current laser inertial unit to be evaluated, and an optimized performance hierarchical evaluation model for the laser inertial unit to be evaluated can be obtained.

[0068] S4. Use the optimized performance hierarchical evaluation model to evaluate the performance of the laser inertial unit to be evaluated.

[0069] Specifically, S4 may include: S41. According to the indicators included in the initial performance hierarchical evaluation model, collect the corresponding indicator data of the laser inertial unit to be evaluated to obtain a sample to be measured; Similarly, it can be understood that the specific values of the 28-dimensional laser strapdown inertial unit indicators corresponding to a certain moment are collected here, so as to obtain the sample to be measured at this moment.

[0070] S42. Use the optimized performance hierarchical evaluation model to conduct hierarchical evaluation on the sample to be measured, and obtain the performance evaluation result of the laser inertial unit to be evaluated. The performance evaluation result includes a performance level and the corresponding confidence level.

[0071] Using the optimized performance hierarchical evaluation model for the laser inertial unit to be evaluated to conduct hierarchical evaluation on the sample to be measured, the performance level and the corresponding confidence level can be obtained as the performance evaluation result of the laser inertial unit to be evaluated. Among them, the level with the highest confidence level is the performance level corresponding to the laser inertial unit to be evaluated.

[0072] Regarding the problem of laser inertial unit performance evaluation, based on the theory of the belief rule base, the present invention establishes a laser inertial unit performance hierarchical evaluation model considering non-equal priorities of indicators, and further realizes a laser inertial unit performance hierarchical evaluation method considering non-equal priorities of indicators. The method first obtains the initial performance hierarchical evaluation model of the laser inertial unit determined in advance; among them, the initial performance hierarchical evaluation model is obtained by analyzing the indicators related to the laser inertial unit performance to construct a hierarchical evaluation model, and for the problem of inconsistent indicator priorities, establishing a BRB-NEP model considering non-equal priorities of indicators; the BRB-NEP model is implemented based on the belief rule base method; secondly, obtain the training samples of the laser inertial unit to be evaluated; among them, the training samples contain the data of each indicator in the initial performance hierarchical evaluation model, and the training samples are marked with performance labels, and the performance labels contain the performance level of the laser inertial unit and the confidence level corresponding to the performance level; then, since the initial performance hierarchical evaluation model is given by expert knowledge, the fuzziness of expert knowledge makes the initial performance hierarchical evaluation model have certain errors. Therefore, using the training samples of the laser inertial unit to be evaluated and adopting a preset optimization algorithm, the parameters in the initial performance hierarchical evaluation model are optimized to obtain an optimized performance hierarchical evaluation model for the laser inertial unit to be evaluated; finally, use the optimized performance hierarchical evaluation model to evaluate the performance of the laser inertial unit to be evaluated.

[0073] The laser inertial unit performance hierarchical evaluation model considering non - equal priorities of indicators provided by the present invention is essentially a hybrid - driven model that can integrate fuzzy expert knowledge and limited test data, realizing the full utilization of limited test data and fuzzy expert knowledge, and effectively improving the performance evaluation accuracy of the laser inertial unit. The present invention designs a performance evaluation method for the laser inertial unit according to the hierarchical evaluation idea, allocating numerous indicators to different sub - models. Each sub - model has limited input indicators, which can not only solve problems such as rule explosion caused by numerous indicators and difficulty in embedding expert knowledge, but also each sub - model gives the evaluation results of each component, realizing the support of component performance results for system performance results, making the evaluation results of the laser inertial unit have better interpretability.

[0074] Moreover, the present invention takes the lead in considering how to conduct evaluation in the case of inconsistent indicator priorities in the performance evaluation of the laser inertial unit. Compared with traditional technologies that mainly consider the impact of data reliability on the performance evaluation of the laser inertial unit (traditional technologies can only solve the performance evaluation problem when indicator priorities are consistent), the present invention can better handle the performance evaluation problem of complex laser inertial unit systems and has more practical value. At the same time, after fully analyzing the indicators related to the performance of the laser inertial unit, the present invention selects 28 - dimensional laser strapdown inertial unit indicators as evaluation indicators, making the indicators more comprehensive, so as to improve the accuracy of evaluation. In addition, the present invention uses a belief rule - base to implement the evaluation model. The belief rule - base (BRB) can handle multiple uncertainties (fuzziness, randomness, incompleteness) in the form of rules by integrating fuzzy logic, probability reasoning, and expert experience, supports quantitative / qualitative mixed input, and the parameters can be optimized. Compared with traditional evidence reasoning, it has stronger uncertainty modeling ability and adaptive optimization characteristics in the evaluation of laser inertial units, and can improve the evaluation accuracy and interpretability.

[0075] To facilitate the understanding of the embodiment scheme of the present invention and verify the performance evaluation accuracy of the hierarchical evaluation model of the present invention for the laser inertial unit, a specific experimental case is given below.

[0076] (1) The construction process of the initial performance hierarchical evaluation model; 1. Data collection Taking 8 sets of laser strapdown inertial units of the same batch and model as the test objects, these laser strapdown inertial units are used for the navigation of a certain type of medium- and long-range rocket. Since the laser strapdown inertial units and the rockets are only tested annually during storage, the sample size of each set of laser strapdown inertial units is very limited, only 30. Since the output frequency of the laser strapdown inertial unit is 50 Hz, the data volume of a single sample is very large. Therefore, in order to expand the sample size and meet the sample size requirements of the experiment, the data of a single sample is expanded into 20 by using an overlapping sliding window method. Then, each set of laser strapdown inertial units can obtain 600 samples after expansion, and 8 sets of laser strapdown inertial units can obtain 4800 samples.

[0077] Each sample consists of data of 28 laser strapdown inertial unit indicators, a total of 28 dimensions, and the vast majority of the data is calculated from the outputs of laser gyroscopes and quartz accelerometers. Taking a certain laser strapdown inertial unit as an example, the gyro zero-bias stability in the X, Y, and Z axes (randomly intercepting 200 samples) is as Figure 4 , Figure 5 and Figure 6 shown.

[0078] 2. Construction of the initial performance stratification evaluation model; According to the stratification evaluation model given in Table 1, three aspects need to be clarified, namely: 1) the evaluation process of the BRB model; 2) how the evaluation results of the lower layer are applied to the evaluation process of the higher layer; 3) the evaluation process of the BRB-NEP model. Here, while clarifying the above three aspects, the initial stratification evaluation model is given.

[0079] 1) The evaluation process of the BRB model; The stratification evaluation model shown in Table 1 includes a total of 14 evaluation models, namely BRB1 to BRB14. Taking the BRB1 model as an example, its evaluation process is introduced. First, the 2 indicators , of the BRB1 model are divided into 3 reference levels: large (denoted by Q1), medium (denoted by Q2), and small (denoted by Q3). According to expert experience and the output of the gyroscope, the index weights, reference levels, and reference values of the indicators , are shown in Table 2.

[0080] Table 2 Index reference levels and reference values of the BRB1 model

[0081] The BRB1 model contains 2 index inputs, and each index has 3 reference levels. According to the rule combination method, 9 confidence rules can be obtained. The fourth-level index "installation error (X-axis gyro)" is also divided into 3 reference levels: large (denoted by Q1), medium (denoted by Q2), and small (denoted by Q3). Then, according to expert experience, the initial BRB1 model is shown in Table 3.

[0082] Table 3 Initial BRB1 model

[0083] Referring to the corresponding processing procedures of index matching, rule activation, and rule fusion in the previous text, the index matching degrees are calculated according to Table 2, and then rule activation and rule fusion are performed. Finally, the evaluation results in the form of confidence distribution are obtained, which are: (26); Among them, is the output of the BRB1 model; and and are the confidence degrees of the three levels Q1, Q2, and Q3 respectively; The evaluation processes of the remaining BRB models are similar to that of the BRB1 model, and the only difference lies in the setting of the reference levels and reference values of the indexes. Therefore, the evaluation processes of the remaining BRB models are not introduced here.

[0084] 2) Method for applying the evaluation results of the lower layer to the evaluation of the higher layer; Taking the BRB7 model as an example, the application method of the evaluation results of the BRB1 model is introduced here. The BRB7 model has 3 input indexes, which are zero-bias stability and scale factor nonlinearity and installation error. Among them, and By setting the reference levels and reference values, and through relevant formulas, the index matching degrees can be obtained; the installation error is the evaluation result in the form of Equation (26). At this time, and and are regarded as the matching degrees of the three levels Q1, Q2, and Q3. Then, rule activation and rule fusion can be further performed, and the evaluation results of the BRB7 model can be obtained.

[0085] The application methods of the evaluation results of the remaining lower layers are the same as this, and are not introduced here.

[0086] 3) Evaluation process of the BRB-NEP model; The output of the BRB-NEP model includes self-detection test results and self-aiming test results and the number of historical transportation times , Cumulative transportation distance , Evaluation results of laser gyroscope assemblies and evaluation results of quartz accelerometer assemblies. Among them , is an index with higher priority, historical transportation times , Cumulative transportation distance , Evaluation results of laser gyroscope assemblies and evaluation results of quartz accelerometer assemblies are indexes with low priority; in addition , , The matching degree is obtained by setting the reference level and reference value. The evaluation confidence degrees of the evaluation results of laser gyroscope assemblies and quartz accelerometer assemblies are regarded as the matching degree. Set , to three reference levels and reference values of many (denoted by Q1), medium (denoted by Q2), and few (denoted by Q3), as shown in Table 4

[0087] Table 4 Index reference levels and reference values of BRB-NEP model

[0088] In the BRB-NEP model , is set to three reference levels. The evaluation results of gyroscope assemblies and accelerometer assemblies are four levels of excellent (denoted by W1), good (denoted by W2), medium (denoted by W3), and poor (denoted by W4). Then a total of A total of 144 confidence rules are combined. In addition, there are 3 rules for activation , for a total of 147 rules. The performance of the laser strapdown inertial unit is also divided into four levels of excellent (denoted by W1), good (denoted by W2), medium (denoted by W3), and poor (denoted by W4). To save space, only some rules are given here, as shown in Table 5. In the table, " " represents any value

[0089] Table 5 Initial BRB-NEP model (partial display)

[0090] The BRB-NEP model first follows the fusion principle given by Equation (15). If rule is activated, the evaluation result of rule is output; if rule is activated, the fusion result of the remaining rules is output. Finally, the evaluation result of the BRB-NEP model is obtained, which is the performance evaluation result of the laser strapdown inertial unit

[0091] (2) Hierarchical evaluation process of laser inertial unit performance considering non-equal priorities of indexes As described above, it will not be elaborated here.

[0092] To obtain the global optimal parameter set of the hierarchical evaluation model, it is necessary to optimize the 15 belief rule base models in the hierarchical evaluation model simultaneously. The statistics of the parameters to be optimized in the 15 belief rule base models are shown in Table 6, with a total of 2,426 parameters to be optimized.

[0093] Table 6 Parameters to be optimized in the hierarchical evaluation model

[0094] Randomly select 4,000 samples from 4,800 samples as training samples, and the remaining 800 samples as test samples. The performance labels of the samples are determined by the votes of multiple experts. The utility values of the four levels of the laser strapdown inertial unit (W1, W2, W3, W4) are set to (4, 3, 2, 1) respectively. The parameters of the artificial bee colony algorithm are set as follows: the number of artificial bees is set to 800, the maximum number of algorithm iterations is set to 5,000, the maximum search threshold for a single position is 10, and the number of employed bees is 50% of the total scale. For comparison, the hierarchical evaluation model and the traditional BRB model are trained simultaneously. At the same time, the above two models are optimized 20 times, and the root mean square error (RMSE) of the utility evaluation results and the actual values is shown in Table 7.

[0095] Table 7 Optimization results of the two models for 20 times

[0096] Taking a certain optimization process as an example, the change of the utility RMSE value during the training process of the two models is as Figure 7 shown. Combining Table 7 and Figure 7 it can be seen that: 1) In the 20 optimizations, comparing the hierarchical evaluation model with the traditional BRB model, the optimal RMSE value evaluated by the former is reduced by 1.42 times compared with the latter, and the average RMSE value is reduced by 1.20 times, indicating that the hierarchical evaluation model has higher evaluation accuracy; 2) From the search process, it can be seen that the RMSE value of the hierarchical evaluation model drops rapidly, while the RMSE value of the traditional BRB model drops slowly, indicating that the hierarchical evaluation model is easier to optimize; 3) The standard deviation of the RMSE of the hierarchical evaluation model is reduced by 3.38 times compared with the traditional BRB model, indicating that the hierarchical evaluation model has better evaluation stability.

[0097] 3. Evaluate using the optimized performance hierarchical evaluation model; Taking the BRB1 model and the BRB-NEP model as examples, the parameter optimization results of the hierarchical evaluation model incorporating prior information are shown in Table 8 and Table 9. Since the number of rules of the BRB-NEP model is too large, only some rules are given here.

[0098] Table 8 Optimized BRB1 Model

[0099] Table 9 Optimized BRB-NEP Model

[0100] 800 test samples were input into the optimized hierarchical evaluation model and the traditional BRB model, and the comparison of the evaluation utility values and the true utility values is shown in Table 10 as follows.

[0101] Table 10 20 Test Results of the Two Models

[0102] The evaluation results of the optimized hierarchical evaluation model for 800 test samples are as Figure 8 shown. After calculation, Figure 8 the RMSE of the evaluation utility value and the true utility value in Figure 8 is 0.1344. It can be seen from

[0103] The above are only the preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A hierarchical evaluation method for the performance of a laser inertial unit considering non-equal priorities of indicators, characterized in that Including: Obtain the initial performance hierarchical evaluation model of a pre-determined laser inertial unit; wherein, the initial performance hierarchical evaluation model is obtained by analyzing the indicators related to the performance of the laser inertial unit to construct a hierarchical evaluation model, and for the problem of inconsistent indicator priorities, establishing a BRB-NEP model considering non-equal priorities of indicators; the BRB-NEP model is implemented based on the belief rule base method; Obtain the training samples of the laser inertial unit to be evaluated; wherein, the training samples contain the data of each indicator in the initial performance hierarchical evaluation model, and the training samples are labeled with performance labels, and the performance labels contain the performance level of the laser inertial unit and the confidence level corresponding to the performance level; Using the training samples of the laser inertial unit to be evaluated, adopt a preset optimization algorithm to optimize the parameters in the initial performance hierarchical evaluation model to obtain an optimized performance hierarchical evaluation model for the laser inertial unit to be evaluated; Use the optimized performance hierarchical evaluation model to evaluate the performance of the laser inertial unit to be evaluated.

2. The method according to claim 1, wherein The determination process of the initial performance hierarchical evaluation model of the laser inertial unit includes: Analyze multiple indicators related to the performance of the laser inertial unit, and construct a hierarchical indicator system through indicator selection; According to the hierarchical indicator system, use the traditional BRB model for the evaluation below the highest level, and establish a BRB-NEP model considering non-equal priorities of indicators for the problem of inconsistent indicator priorities at the highest level for evaluation, so as to construct a hierarchical evaluation model.

3. The method according to claim 2, wherein The analysis of multiple indicators related to the performance of the laser inertial unit and the construction of a hierarchical indicator system through indicator selection include: For the laser strapdown inertial unit composed of a laser gyroscope and a quartz accelerometer, through performance analysis, select multiple indicators representing historical information and test information, where the historical information includes the number of historical transports and the cumulative transport mileage; the test information includes the self-calibration test result, the self-aiming test result, and the self-detection test result; and determine the expression form of the self-detection test result based on the 0-1 output form, determine the expression form of the self-aiming test result based on the 0-1 output form, and determine the calibration parameters corresponding to the laser gyroscope and the quartz accelerometer respectively by determining the expression of the output error model corresponding to the laser gyroscope and the quartz accelerometer respectively, for representing the self-calibration test result; According to the information determined by analyzing multiple indicators, conduct indicator selection, and finally determine the 28-dimensional laser strapdown inertial unit indicators, expressed as: ； Among them, represents the collection of 28 - dimensional laser strapdown inertial unit indicators; represents the self - detection test result; represents the self - aiming test result; represents the historical transportation times; represents the cumulative transportation mileage; represents the zero - bias stability of the X - axis laser gyroscope; represents the zero - bias stability of the Y - axis laser gyroscope; represents the zero - bias stability of the Z - axis laser gyroscope; represents the installation error of the X - axis laser gyroscope around the Z - axis; represents the installation error of the X - axis laser gyroscope around the Y - axis; represents the installation error of the Y - axis laser gyroscope around the Z - axis; represents the installation error of the Y - axis laser gyroscope around the X - axis; represents the installation error of the Z - axis laser gyroscope around the Y - axis; represents the installation error of the Z - axis laser gyroscope around the X - axis; represents the scale - factor non - linearity of the X - axis laser gyroscope; represents the scale - factor non - linearity of the Y - axis laser gyroscope; represents the scale - factor non - linearity of the Z - axis laser gyroscope; represents the zero - bias stability of the X - axis quartz accelerometer; represents the zero - bias stability of the Y - axis quartz accelerometer; represents the zero - bias stability of the Z - axis quartz accelerometer; represents the installation error between the Y - axis quartz accelerometer and the X - axis; represents the installation error between the Z - axis quartz accelerometer and the X - axis; represents the installation error between the X - axis quartz accelerometer and the Y - axis; represents the installation error between the Z - axis quartz accelerometer and the Y - axis; represents the installation error between the X - axis quartz accelerometer and the Z - axis; represents the installation error between the Y - axis quartz accelerometer and the Z - axis; represents the scale - factor stability of the X - axis quartz accelerometer; represents the scale - factor stability of the Y - axis quartz accelerometer; represents the scale - factor stability of the Z - axis quartz accelerometer; According to the 28-dimensional laser strapdown inertial unit indicators, layer by layer according to the instrument type, construct a hierarchical indicator system with a total of five layers; wherein, the indicators from the fifth layer to the first layer are set from bottom to top. The indicators of the fifth layer include: , , , , , , and, , , , , , ; The indicators of the fourth layer include: zero-bias stability, scale factor nonlinearity, and installation error evaluation results of the X-axis laser gyroscope; zero-bias stability, scale factor nonlinearity, and installation error evaluation results of the Y-axis laser gyroscope; zero-bias stability, scale factor nonlinearity, and installation error evaluation results of the Z-axis laser gyroscope; and zero-bias stability, scale factor stability, and installation error evaluation results of the X-axis quartz accelerometer; zero-bias stability, scale factor stability, and installation error evaluation results of the Y-axis quartz accelerometer; zero-bias stability, scale factor stability, and installation error evaluation results of the Z-axis quartz accelerometer; among which, the installation error evaluation result of the X-axis laser gyroscope is calculated based on the indicators of the fifth layer , The installation error evaluation result of the Y-axis laser gyroscope is calculated based on the indicators of the fifth layer , The installation error evaluation result of the Z-axis laser gyroscope is calculated based on the indicators of the fifth layer , The installation error evaluation result of the X-axis quartz accelerometer is calculated based on the indicators of the fifth layer , The installation error evaluation result of the Y-axis quartz accelerometer is calculated based on the indicators of the fifth layer , The installation error evaluation result of the Z-axis quartz accelerometer is calculated based on the indicators of the fifth layer , The installation error evaluation result of the Z-axis quartz accelerometer is calculated based on the indicators of the fifth layer The indicators of the third layer include: the evaluation results of the X-axis laser gyroscope, the evaluation results of the Y-axis laser gyroscope, the evaluation results of the Z-axis laser gyroscope, and the evaluation results of the X-axis quartz accelerometer, the evaluation results of the Y-axis quartz accelerometer, and the evaluation results of the Z-axis quartz accelerometer; among them, the evaluation result of the X-axis laser gyroscope is calculated based on the indicators of the fourth layer: the zero-bias stability, scale factor nonlinearity, and installation error evaluation results of the X-axis laser gyroscope; the evaluation result of the Y-axis laser gyroscope is calculated based on the indicators of the fourth layer: the zero-bias stability, scale factor nonlinearity, and installation error evaluation results of the Y-axis laser gyroscope; the evaluation result of the Z-axis laser gyroscope is calculated based on the indicators of the fourth layer: the zero-bias stability, scale factor nonlinearity, and installation error evaluation results of the Z-axis laser gyroscope; the evaluation result of the X-axis quartz accelerometer is calculated based on the indicators of the fourth layer: the zero-bias stability, scale factor stability, and installation error evaluation results of the X-axis quartz accelerometer; the evaluation result of the Y-axis quartz accelerometer is calculated based on the indicators of the fourth layer: the zero-bias stability, scale factor stability, and installation error evaluation results of the Y-axis quartz accelerometer; the evaluation result of the Z-axis quartz accelerometer is calculated based on the indicators of the fourth layer: the zero-bias stability, scale factor stability, and installation error evaluation results of the Z-axis quartz accelerometer; The indicators of the second layer include: self-detection test results, self-aiming test results, historical transportation times, cumulative transportation mileage, evaluation results of the laser gyroscope assembly, and evaluation results of the quartz accelerometer assembly; among them, the evaluation results of the laser gyroscope assembly are calculated based on the indicators of the laser gyroscope in the third layer; the evaluation results of the quartz accelerometer assembly are calculated based on the indicators of the quartz accelerometer in the third layer; The indicators of the first layer include: the evaluation results of the laser strapdown inertial assembly, which are calculated based on all the indicators of the second layer.

4. The method according to claim 3, wherein According to the hierarchical indicator system, the traditional BRB model is used for the evaluation below the highest level, and in view of the problem of inconsistent priorities of the highest-level indicators, a BRB-NEP model considering non-equal priorities of indicators is established for evaluation, thereby constructing a hierarchical evaluation model, including: For the process from the fifth layer to the fourth layer, from the fourth layer to the third layer, and from the third layer to the second layer in the hierarchical indicator system, the traditional BRB model is set as the corresponding belief rule base model for the process of obtaining the upper-level indicators from the lower-level indicators; and for the evaluation process from the second layer to the first layer, in view of the problem of inconsistent priorities of the first-layer indicators, a BRB-NEP model considering non-equal priorities of indicators is established as the corresponding belief rule base model, thereby constructing a hierarchical evaluation model; among them, in the hierarchical indicator system, the priorities of the self-detection test results and the self-aiming test results are the same and are both higher than the priorities of the other indicators.

5. The method according to claim 4, wherein The process of establishing a BRB-NEP model considering non-equal priorities of indicators includes: For the modeling problem with different indicator priorities at the highest level, rules are added based on the traditional BRB model , described as: ； Among them, it is assumed that there are indicators , among which the first indicators are 0-1 outputs and have higher priority, and the remaining indicators have a lower level of priority; is the complete set of the remaining indicators ; represents the OR rule; represents the AND rule; is the performance levels of the laser inertial unit; is the evaluation result; is the rule weight of the rule is the indicator weight of the , and are natural numbers greater than 0; Construct the remaining rules by comprehensively using the AND rule and the OR rule. Then, the rule is expressed as: ； wherein, are respectively the index reference values for the th index corresponding to the th rule; is the confidence level for the th performance level corresponding to the th rule; is the rule weight of rule ; is the total number of rules, is a natural number greater than 0; Injunction , the rule fusion principle is determined as follows: ； Determine the inference process of the BRB-NEP model as three steps: indicator matching, rule activation, and rule fusion, and determine the corresponding processing methods for each step.

6. The method according to claim 5, characterized in that, The processing process of the indicator matching includes: For non-0-1 output metrics, the reference value of the th metric and the th rule is denoted as ; for , when , determine the matching degree of the th rule is expressed as: ； Wherein: is the test value of the th index at time ; is the matching degree for the th rule; is the total number of rules; represents the th index and the index reference value of the th rule; For the 0-1 output metrics, directly set the matching degree of such metrics to 0.5, expressed as: ； Among them, represents the moment when the matching degree of the th rule.

7. The method according to claim 6, wherein The processing process activated by the rule includes: Calculate the matching degree of all metrics against the th rule as: ； ； Among them, represents the index weight of the th index; represents the relative index weight of the th index; Combined with the rule weight of Rule and the matching degree of Rule , the activation weight corresponding to Rule is as follows: ； Among them, is the activation weight corresponding to the th rule at the moment; represents the rule weight of the th rule; represents the matching degree of all indicators to the th rule at the moment .

8. The method according to claim 7, wherein The processing process of rule fusion includes: Based on the activation weight of Rule and confidence , the basic probability mass is calculated as follows: ； Among them, is the output result of the th rule at time is the basic probability mass; represents the empty set; is the confidence level for the th rule and the th performance level; is the complete set of all performance levels; when is the case, is the remaining basic probability mass at time, representing global ignorance; , , are two components of global ignorance; Using the recursive ER algorithm for rule fusion, the fusion result of the first pieces of evidence is obtained and expressed as: ； Among them, is the normalized basic probability mass before fusion after the rules; is the normalized basic probability mass before fusion after the rules; is the normalization parameter of the rules before fusion; is the normalization parameter of the rules before fusion; is the abbreviation of ; is the abbreviation of and is, at time , the basic probability mass that the output of the rule is the th performance level; ; is the remaining basic probability mass of the rule; is the remaining basic probability mass after the rules before fusion; , are the two components of ; , are the two components of the remaining basic probability mass of the rule; Based on the normalized basic probability mass, obtain the confidence level of the laser strapdown inertial unit at different performance levels, expressed as: ； Among them, is the abbreviation of and is the confidence of the output result after the fusion of the time rules ; represents the normalized basic probability mass that the output result is after fusing all rules; represents one of the components, which is the remaining basic probability mass after fusing all rules; According to the fusion results of all rules, the output of the BRB-NEP model is obtained as: ； Assume the performance level of the laser strapdown inertial unit has a utility value of , then the utility value of the evaluation result is: ； Among them, is the utility value of the evaluation result.

9. The method according to claim 8, wherein Using the training sample data of the laser inertial unit to be evaluated, and adopting a preset optimization algorithm, optimize the parameters in the initial performance hierarchical evaluation model to obtain an optimized performance hierarchical evaluation model for the laser inertial unit to be evaluated, including: Taking the minimum root mean square error between the output value of the evaluation model utility and the actual value as the optimization objective, a parameter optimization model of the initial performance stratification evaluation model is established, where the parameters to be optimized include: the weights of indicators , the weights of rules and the confidence levels of the performance levels corresponding to the th rule; According to the optimization objective, using the training sample data of the laser inertial unit to be evaluated, and adopting the artificial bee colony algorithm, optimize the parameters in the initial performance hierarchical evaluation model to obtain an optimized performance hierarchical evaluation model for the laser inertial unit to be evaluated.

10. The method according to any one of claims 1-9, characterized in that, Using the optimized performance hierarchical evaluation model to evaluate the performance of the laser inertial unit to be evaluated, including: According to the metrics included in the initial performance hierarchical evaluation model, collect the corresponding metric data of the laser inertial unit to be evaluated to obtain a sample to be measured; Using the optimized performance hierarchical evaluation model to conduct hierarchical evaluation on the sample to be measured, and obtain the performance evaluation result of the laser inertial unit to be evaluated, where the performance evaluation result includes the performance level and the corresponding confidence level.

Images