A three-self laser inertial measurement unit fault detection method based on interval value confidence rule base

Abstract

The application relates to a three-self laser inertial measurement unit fault detection method based on an interval value confidence rule base, which comprises the following steps: determining a plurality of key index information of a three-self laser inertial measurement unit, and setting a reference interval according to each key index information; activating corresponding rules in an optimized confidence rule base according to the reference interval to which each key index information belongs, so as to obtain a plurality of activated rules; fusing the plurality of activated rules according to the weight, confidence and availability of each activated rule, so as to obtain a rule synthesis result; performing utility conversion by using the rule synthesis result and the utility value of a detection result, so as to obtain a utility conversion result; judging whether the three-self laser inertial measurement unit is faulty according to the utility conversion result, so as to obtain a detection result. The fault detection method solves the problems of index combination rule quantity explosion and the uncertainty and unreliability of expert knowledge.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of inertial measurement unit fault detection, and particularly relates to a three-self laser inertial measurement unit fault detection method based on an interval value confidence rule base. BACKGROUND

[0002] The three-self laser inertial measurement unit is an important navigation device, can work without relying on external signals, has good concealment, and has been widely applied to military fields such as missile weapons, carrier rockets, and aircrafts. As a key single machine of a control system, the three-self laser inertial measurement unit plays an important role in the control system, and can be used for positioning, orientation, and navigation tracking by measuring the acceleration increments and angular velocity increments of three axes of a space coordinate system. The main components of the three-self laser inertial measurement unit include orthogonally distributed gyroscopes and accelerometers, which are used to measure the angular acceleration and linear acceleration information of a carrier, and the information will be used as the main source of control instructions of the control system.

[0003] In the use process of the laser inertial measurement unit carrier, the inertial measurement unit system is mainly used to perceive the speed and attitude changes of the carrier. However, due to the influence of factors such as the drastic change of the environment in different stages of flight, the complex environment, and the strong enemy interference, the performance of the inertial measurement unit is continuously in a high-load working state, and the performance degradation speed is accelerated. The influence of such performance degradation can be an environmental influence. Therefore, the performance state of the laser inertial measurement unit is degraded during the working process due to a series of environmental influences, which greatly increases the probability of the occurrence of faults of the laser inertial measurement unit, and further seriously restricts the overall reliability of the carrier rocket, missile, and other aircrafts. Therefore, timely and effective detection of the faults of the three-self laser inertial measurement unit is the key to ensuring the safe and reliable operation of the carrier rocket, missile, and other aircrafts.

[0004] Since the three-self laser inertial measurement unit belongs to a single machine with high reliability, the probability of failure in the use process is low, and most of the monitoring data that can be obtained are normal data, and the data in the fault state are lacking. The BRB modeling method is suitable for small sample data, and it is appropriate to establish a BRB model by using the data of the three-self laser inertial measurement unit. However, there are problems of explosive number of rule combinations of indexes and unreliability of indexes in the traditional BRB. SUMMARY

[0005] In order to solve the above problems in the prior art, the application provides a three-self laser inertial measurement unit fault detection method based on an interval value confidence rule base. The technical problem to be solved by the application is solved by the following technical scheme.

[0006] The application provides a three-self laser inertial measurement unit fault detection method based on an interval value confidence rule base, which comprises the following steps:

[0007] A plurality of key index information of the three-self laser inertial measurement unit is determined, and a reference interval is set according to each key index information.

[0008] According to the reference interval to which each of the key indicator information belongs, activate corresponding rules in the optimized confidence rule base to obtain a plurality of activated rules;

[0009] Fuse the plurality of activated rules according to the weight, confidence and availability of each of the activated rules to obtain a rule synthesis result;

[0010] Perform utility conversion using the rule synthesis result and the utility value of the detection result to obtain a utility conversion result;

[0011] Judge whether the three-self laser inertial measurement unit has a fault according to the utility conversion result to obtain a detection result.

[0012] In an embodiment of the present application, the plurality of key indicator information includes: X-axis cumulative pulse amount of a gyroscope, Y-axis cumulative pulse amount of the gyroscope, Z-axis cumulative pulse amount of the gyroscope, X-axis cumulative pulse amount of an accelerometer, Y-axis cumulative pulse amount of the accelerometer, Z-axis cumulative pulse amount of the accelerometer, and unit time cumulative pulse amount of the gyroscope and the accelerometer.

[0013] In an embodiment of the present application, the reference interval according to each of the key indicator information is set, including:

[0014] Set an initial reference value of each key indicator information;

[0015] According to the number of the initial reference values, equally divide the initial reference values into the reference intervals by an expert system.

[0016] In an embodiment of the present application, according to the reference interval to which each of the key indicator information belongs, activate corresponding rules in the optimized confidence rule base to obtain a plurality of activated rules, including:

[0017] Construct an initial confidence rule base based on the key indicator information and expert knowledge information;

[0018] Optimize the confidence of the rules, the weight of the rules and the reliability of the indicators in the initial rule base by using a projection covariance matrix adaptive strategy optimization method to obtain the optimized confidence rule base;

[0019] Judge the reference interval into which the key indicator information falls, and activate the rules corresponding to the reference interval in the optimized confidence rule base to obtain the plurality of activated rules.

[0020] In an embodiment of the present application, the kth rule in the initial confidence rule base is:

[0021] BR kIF x1∈[a1,b1] ∨x2∈[a2,b2] ∨... ∨x MK ∈[a MK ,b MK ]

[0022] THEN result is{(D1,β 1,k ),(D2,β 2,k ),...,(D N ,β N,k )}

[0023] WITH rule weightθ k

[0024] AND indicator reliability r Mk ,rule availabilityΔ k

[0025] k∈{1,...,L},

[0026] wherein x1,x2,…,x Mk represent key index information of three self-excited laser inertial measurement units, M k represents the total number of key index information, [a1,b1],[a2,b2],...,[a Mk ,b Mk ] represent interval reference values of key index information, D1,D2,…,D N represent N detection results, β 1,k ,β 2,k ,…,β N,k represent the confidence of each rule detection result, θ k represents the weight of the kth rule, r Mk represents the index reliability, L represents the number of rules in the confidence rule base, and Δ k represents the availability of the kth rule.

[0027] Δ k =(1-(1-r1)(1-r2)...(1-r Mk ))ψ

[0028] wherein ψ represents a confidence factor, and ψ is equal to the absolute value of the difference between the square of the confidence value.

[0029] In an embodiment of the present application, in the optimization process of the confidence of rules in the initial rule base, the weight of rules and the index reliability, the confidence of rules, the weight of rules and the index reliability satisfy the following constraint conditions:

[0030] min MSE(θ k ,β i,k ,r i )

[0031] st.

[0032] 0≤θ k ≤1,k=1,2,...L

[0033] 0≤β i,k ≤1,i=1,..,N,k=1,2,...L

[0034] 0≤r i ≤1,i=1,...,M k

[0035] where, T represents the number of training sample sets, output estimated and output acutal represent the predicted utility value and the true utility value, θ k represents the weight of the kth rule, L represents the number of rules in the confidence rule base, β i,k represents the confidence of each rule detection result, r i represents the index reliability, M k represents the total number of key index information.

[0036] In one embodiment of the present application, the plurality of active rules are fused according to the weight, confidence and availability of each of the active rules to obtain a rule synthesis result, comprising:

[0037] The basic probability mass of each active rule is calculated according to the weight and confidence of each active rule:

[0038]

[0039] wherein m n,k represents the basic probability mass of the kth active rule, θ k represents the weight of the kth rule, β n,k represents the confidence of the nth rule detection result, n∈{1,...,N}, Θ represents the identification framework, and Θ={D1,...,D N}, D1,D2,…,D N represent N detection results;

[0040] The plurality of active rules are fused according to the basic probability mass of each of the active rules and the availability of the rules to obtain a rule synthesis result:

[0041]

[0042]

[0043] wherein, p n,k represents the confidence of the synthesized rule, represents the joint probability mass, Δ i represents the availability of the ith rule, Δ j represents the availability of the jth rule, m n,j represents the basic probability mass of the jth rule, m n,i represents the basic probability mass of the ith rule, m A,j represents the basic probability mass of the jth rule with the evaluation result as A, m B,i represents the basic probability mass of the ith rule with the evaluation result as B, A represents one evaluation result case in Θ, and B represents one evaluation result case in Θ, represents the basic probability mass of the kth rule with the evaluation result as A.

[0044] In one embodiment of the present application, utility conversion is performed by using the utility value of the rule synthesis result and the detection result, so as to obtain a utility conversion result, which comprises:

[0045] setting the utility value of the detection result;

[0046] performing utility conversion by using the utility value of the rule synthesis result and the detection result, so as to obtain a utility conversion result:

[0047]

[0048] wherein, u(D n ) represents the utility value of the detection result as D n , and p n,k represents the confidence of the synthesized rule.

[0049] In one embodiment of the present application, whether the three-self laser inertial measurement unit is faulty is judged according to the utility conversion result, so as to obtain a detection result, which comprises:

[0050] when the utility conversion result is located in a first interval, the detection result is a first numerical value; and when the utility conversion result is located in a second interval, the detection result is a second numerical value;

[0051] whether the three-self laser inertial measurement unit is faulty is judged according to the first numerical value and the second numerical value, so as to obtain a detection result.

[0052] Compared with the prior art, the present application has the following beneficial effects:

[0053] The fault detection method of the present application sets a reference interval according to key index information, compared with a single index reference value, the interval reference value greatly reduces the error and uncertainty, thereby solving the problem of the number explosion of index combination rules; at the same time, by adopting an optimized confidence rule base, adding the availability parameter of the rule in rule fusion, and calculating the availability of the rule by using the index reliability, the uncertainty and unreliability of the expert knowledge are reduced, the reliability of the index information is improved, and the precision of the method is improved. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 A flowchart of a three-self laser inertial measurement unit fault detection method based on an interval value confidence rule base provided by the embodiment of the present application is shown in the figure.

[0055] Figure 2 A flowchart of another three-self laser inertial measurement unit fault detection method based on an interval value confidence rule base provided by the embodiment of the present application is shown in the figure.

[0056] Figure 3 A flowchart of a P-CMA-ES optimization method provided by the embodiment of the present application is shown in the figure.

[0057] Figure 4 An output result diagram of the three-self laser inertial measurement unit fault detection method based on the interval value confidence rule base provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0058] The present application will be further described in detail below in combination with specific embodiments, but the embodiments of the present application are not limited thereto.

[0059] Embodiment One

[0060] The present embodiment aims to solve the problems of the unreliability of expert knowledge and the number explosion of index combination rules in the three-self laser inertial measurement unit fault detection process, proposes a reference interval, a disjunction rule, an index reliability and a new availability calculation method of the rule, and at the same time, in order to further improve the detection precision, the model parameters given by the expert knowledge in the initial confidence rule base are optimized and learned by using an optimization algorithm in combination with the acquired historical data and model parameter physical significance.

[0061] Please refer to Figure 1 and Figure 2 , Figure 1 A flowchart of a three-self laser inertial measurement unit fault detection method based on an interval value confidence rule base provided by the embodiment of the present application is shown in the figure, Figure 2Another flowchart of a three-self laser inertial measurement unit fault detection method based on interval value confidence rule base provided by the embodiment of the present application. The three-self laser inertial measurement unit fault detection method based on the interval value confidence rule base (Belief rule base, BRB) comprises the following steps:

[0062] S1, determining a plurality of key index information of the three-self laser inertial measurement unit, and setting a reference interval according to each key index information.

[0063] Specifically, the key index information comprises an X-axis cumulative pulse amount of a gyroscope, a Y-axis cumulative pulse amount of the gyroscope, a Z-axis cumulative pulse amount of the gyroscope, an X-axis cumulative pulse amount of an accelerometer, a Y-axis cumulative pulse amount of the accelerometer, a Z-axis cumulative pulse amount of the accelerometer, and a unit time cumulative pulse amount of the gyroscope and the accelerometer.

[0064] The method for obtaining the key index information is as follows: firstly, static testing is performed on the three-self laser inertial measurement unit in different environments, and the X-axis pulse amount, the Y-axis pulse amount, the Z-axis pulse amount of the gyroscope and the X-axis pulse amount, the Y-axis pulse amount, the Z-axis pulse amount of the accelerometer in the three-self laser inertial measurement unit are obtained from the testing data. Then, the X-axis pulse amount, the Y-axis pulse amount, the Z-axis pulse amount of the gyroscope and the X-axis pulse amount, the Y-axis pulse amount, the Z-axis pulse amount of the accelerometer are respectively subjected to differential processing, so as to obtain the differential amount of the X-axis pulse amount, the differential amount of the Y-axis pulse amount, the differential amount of the Z-axis pulse amount of the gyroscope and the differential amount of the X-axis pulse amount, the differential amount of the Y-axis pulse amount, the differential amount of the Z-axis pulse amount of the accelerometer. After that, the feature amount capable of representing the information characteristics is extracted from the differential amount of the X-axis pulse amount, the differential amount of the Y-axis pulse amount, the differential amount of the Z-axis pulse amount of the gyroscope and the differential amount of the X-axis pulse amount, the differential amount of the Y-axis pulse amount, the differential amount of the Z-axis pulse amount of the accelerometer, and the X-axis cumulative pulse amount, the Y-axis cumulative pulse amount, the Z-axis cumulative pulse amount of the gyroscope and the X-axis cumulative pulse amount, the Y-axis cumulative pulse amount, the Z-axis cumulative pulse amount of the accelerometer are selected as the feature amount from the feature amount in combination with the expert knowledge, so as to obtain the key index information. Furthermore, the unit time differential amount of the X-axis of the gyroscope, the unit time differential amount of the Y-axis of the gyroscope, the unit time differential amount of the Z-axis of the gyroscope and the unit time differential amount of the X-axis of the accelerometer, the unit time differential amount of the Y-axis of the accelerometer, the unit time differential amount of the Z-axis of the accelerometer are accumulated, so as to obtain the unit time cumulative pulse amount as the key index information.

[0065] After the key index information is determined, the reference interval is set according to each key index information, and the method comprises the following steps:

[0066] Firstly, the initial reference value of each key index information is set. Specifically, the initial reference value of each key index information can be set by the expert knowledge.

[0067] Then, according to the number of the initial reference values, the initial reference values are equally divided into the reference intervals by the expert system. Specifically, according to the number of the initial reference values, two adjacent initial reference values can be taken as one reference interval, so that all the initial reference values are equally divided into multiple reference intervals; or the interval initial reference values can be taken as one reference interval, so that all the initial reference values are equally divided into multiple reference intervals.

[0068] In one specific embodiment, there are two key indicators 1 and 2, both of which have four initial reference values, and the corresponding initial reference values and reference intervals are set as shown in Table 1.

[0069] Table 1 Reference values and reference intervals of indicators

[0070]

[0071] The embodiment sets the reference intervals according to the key indicator information, which greatly reduces the error and uncertainty compared with a single determined reference value, thereby solving the problem of the explosion of the number of indicator combination rules.

[0072] S2, activate the corresponding rules in the optimized confidence rule base according to the reference interval to which each key indicator information belongs, to obtain a plurality of activated rules. Specifically, it includes the following steps:

[0073] S21, construct an initial confidence rule base based on the key indicator information and the expert knowledge information.

[0074] Specifically, in the initial confidence rule base constructed based on the key indicator information and the expert knowledge information, the kth rule is:

[0075]

[0076] Wherein, x1, x2, …, x Mk represent the key indicator information of the three self-laser inertial measurement units, M k represents the total number of key indicator information, [a1, b1], [a2, b2], …, [a Mk , b Mk ] represent the interval reference values of the key indicator information, D1, D2, …, D N represent N detection results, β 1,k , β 2,k , …, β N,k represent the confidence of each rule detection result, θ k represents the weight of the kth rule, r Mk represents the indicator reliability, L represents the number of rules in the confidence rule base, Δ k represents the availability of the kth rule.

[0077] The availability of rules represents the degree of trust in the rules. In the construction of the three-self laser inertial measurement unit fault detection method, the availability is an important parameter of the rules. Therefore, the embodiment proposes a new calculation method of the availability of rules. The calculation formula of the availability of rules is as follows:

[0078]

[0079] Wherein, Δ k represents the availability of rules, r represents the index reliability, and ψ represents the confidence factor; ψ is equal to the absolute value of the difference between the square of the confidence value, for example, the confidence is {a, b, c}, then ψ = |a 2 -b 2 -c 2 |.

[0080] S22, the confidence of rules, the weight of rules and the index reliability of rules in the initial rule base are optimized by using a projection covariance matrix adaptive strategy (P-CMA-ES) optimization method, so as to obtain the optimized confidence rule base.

[0081] In order to enable the confidence rule base to maintain high accuracy in the actual application process, the initial confidence rule base is optimized by using a model optimization algorithm.

[0082] Please refer to Figure 3 , Figure 3 for a flowchart of a P-CMA-ES optimization method provided by the embodiment of the application. The P-CMA-ES optimization method comprises the following steps: in the first step, the initial parameters are given by expert knowledge, including the confidence of rules β 1,1 ,...,β N,,L , the weight of rules θ1,...,θ L and the index reliability r1,...,r L . In the second step, each generation of solutions is obtained by sampling the population. In the third step, each generation of solutions is projected onto the hyperplane of the feasible region on the constraint feasible region. In the fourth step, the optimal value in the population is selected. In the fifth step, the covariance matrix is updated. In the sixth step, the above steps are repeatedly executed until the overall optimal solution is obtained, at this time, the confidence of rules, the weight of rules and the index reliability satisfy the following constraint conditions:

[0083] min MSE(θ k ,β i,k ,r i )

[0084] st.

[0085] 0≤θ k ≤1,k=1,2,...L

[0086] 0≤β i,k ≤1,i=1,..,N,k=1,2,...L

[0087] 0≤r i ≤1,i=1,...,M k

[0088] wherein, T represents the number of training sample sets, output estimated and output acutal represent the predicted utility value and the real utility value.

[0089] wherein, θ k represents the weight of the kth rule, L represents the number of rules in the confidence rule base, β i,k represents the confidence of each rule detection result, r i represents the index reliability, M k represents the total number of key index information.

[0090] S23, judge the reference interval in which the key index information falls, and activate the rule corresponding to the reference interval in the optimized confidence rule base, to obtain the several activated rules.

[0091] Specifically, according to the reference interval in which the key index information x1, x2, …, x M falls, the corresponding rule in the optimized confidence rule base is activated, for example, x1 falls in the reference interval [a1, b1], then the rule corresponding to [a1, b1] is activated; x2 falls in the reference interval [a2, b2], then the rule corresponding to [a2, b2] is activated; x M falls in the reference interval [a M , b M ], then the rule corresponding to [a M , b M ] is activated.

[0092] Taking the reference values and reference intervals of the indexes in Table 1 as an example, the rule activation mode is: assuming that the input information of index 1 is p1 at this time, and the falling reference interval is , then the rule corresponding to is activated; the input information of index 2 is p2, and the falling reference interval is , then the rule corresponding to is activated, thereby obtaining the two activated rules of corresponding to the rule and corresponding to the rule.

[0093] S3, fusing the several activation rules according to the weight, the confidence and the availability of each of the activation rules, to obtain a rule synthesis result.

[0094] In one specific embodiment, the fusion of the activation rules is performed by using an Evidential Reasoning (ER) rule algorithm according to the weight, the confidence and the availability of each of the activation rules, and specifically includes the following steps:

[0095] S31, calculating a basic probability mass of each activation rule according to the weight and the confidence of each activation rule:

[0096]

[0097] wherein m n,k represents the basic probability mass of the kth activation rule, θ k represents the weight of the kth rule, β n,k represents the confidence of the nth rule detection result, n e {1,...,N}, and Θ represents a recognition framework, and Θ = {D1,...,D N}, D1,D2,...,D N represents N detection results.

[0098] S32, fusing the several activation rules according to the basic probability mass of each of the activation rules and the availability of each of the activation rules, to obtain a rule synthesis result:

[0099]

[0100]

[0101] wherein p n,k represents the confidence of the synthesized rule, represents a joint probability mass, Δ i represents the availability of the ith rule, Δ j represents the availability of the jth rule, m n,j represents the basic probability mass of the jth rule, m n,i represents the basic probability mass of the ith rule, m A,j represents the basic probability mass of the jth rule evaluation result A, m B,i represents the basic probability mass of the ith rule evaluation result B, A represents an evaluation result case in Θ, B represents an evaluation result case in Θ, represents the basic probability mass of the kth rule evaluation result A.

[0102] S4, performing utility conversion on the utility value of the rule synthesis result and the detection result to obtain a utility conversion result.

[0103] S41, setting the utility value of the detection result.

[0104] The detection result of the embodiment sets two utility values: 1 and 2, wherein 1 represents no fault, and 2 represents fault.

[0105] S42, performing utility conversion on the utility value of the rule synthesis result and the detection result to obtain a utility conversion result.

[0106]

[0107] wherein u(D i ) represents the utility value of the detection result D i , and p n,k represents the confidence of the synthesized rule.

[0108] S5, judging whether the three-self laser inertial measurement unit is faulty according to the utility conversion result to obtain a detection result.

[0109] Specifically, according to the comparison between the utility conversion result and the numerical interval, and in combination with the actual physical meaning of the parameters in the fault detection method, the final result is obtained.

[0110] In one specific embodiment, the numerical interval to which the utility conversion result calculated in the judging step S4 belongs, when the calculated utility conversion result belongs to a first numerical interval, the detection result is 1, indicating that the three-self laser inertial measurement unit is not faulty; when the calculated utility conversion result belongs to a second numerical interval, the detection result is 2, indicating that the three-self laser inertial measurement unit is faulty. Specifically, the first numerical interval can be [0.5-1.5], and the second numerical interval can be (1.5-2.5].

[0111] The fault detection method of the embodiment sets a reference interval according to key index information, compared with a single index reference value, the interval reference value greatly reduces the error and uncertainty, thereby solving the problem of the explosion of the number of index combination rules; at the same time, by adopting an optimized confidence rule base, adding the availability parameter of the rule during rule fusion, and calculating the availability of the rule by using the index reliability, the uncertainty and unreliability of the expert knowledge are reduced, the reliability of the index information is improved, and the precision of the method is improved.

[0112] Embodiment Two

[0113] Based on the embodiment one, the embodiment further verifies the effectiveness of the detection method through a simulation experiment of the fault detection of the three-self laser inertial measurement unit. The simulation experiment includes the following steps:

[0114] S1, determine several key index information of the three self-excited laser inertial measurement unit.

[0115] First, the three self-excited laser inertial measurement unit is tested statically in different environments, and the X-axis pulse quantity, Y-axis pulse quantity, Z-axis pulse quantity of the gyroscope and the X-axis pulse quantity, Y-axis pulse quantity, Z-axis pulse quantity of the accelerometer in the three self-excited laser inertial measurement unit are obtained from the test data. Then, the X-axis pulse quantity, Y-axis pulse quantity, Z-axis pulse quantity of the gyroscope and the X-axis pulse quantity, Y-axis pulse quantity, Z-axis pulse quantity of the accelerometer are respectively processed, and the differential quantity of the X-axis pulse quantity, Y-axis pulse quantity, Z-axis pulse quantity of the gyroscope and the differential quantity of the X-axis pulse quantity, Y-axis pulse quantity, Z-axis pulse quantity of the accelerometer are obtained. Then, the characteristic quantity capable of representing the information characteristics is extracted from the differential quantity of the X-axis pulse quantity, Y-axis pulse quantity, Z-axis pulse quantity of the gyroscope and the differential quantity of the X-axis pulse quantity, Y-axis pulse quantity, Z-axis pulse quantity of the accelerometer, and the X-axis cumulative pulse quantity, Y-axis cumulative pulse quantity, Z-axis cumulative pulse quantity of the gyroscope and the X-axis cumulative pulse quantity, Y-axis cumulative pulse quantity, Z-axis cumulative pulse quantity of the accelerometer are selected as the characteristic quantity as the key index information according to the expert knowledge; and the X-axis unit time differential quantity, Y-axis unit time differential quantity, Z-axis unit time differential quantity of the gyroscope and the X-axis unit time differential quantity, Y-axis unit time differential quantity, Z-axis unit time differential quantity of the accelerometer are accumulated to obtain the unit time cumulative pulse quantity as the key index information.

[0116] In summary, the embodiment sets 7 key index information.

[0117] S2, set a reference interval according to each key index information.

[0118] Specifically, for the 7 key index information, 21 initial reference values are set for each key index according to the expert knowledge, and the reference interval of each key index is set to 20 according to the initial reference value. Since one interval corresponds to one rule, 140 rules are set for the 7 indexes. The initial reference value of each key index is shown in Table 2. The reference interval is set according to the initial reference value of the key index, as shown in Table 3.

[0119] Table 2 key index and its initial reference value

[0120] Sequence number Key indicator 1 Sequence number Key indicator 2 ... Sequence number Key indicator 7 1 0.7500 22 0.2500 ... 127 0.2832 2 0.7875 23 0.2625 ... 128 0.2875 3 0.8250 24 0.2750 ... 129 0.2919 4 0.8625 25 0.2875 ... 130 0.2962 5 0.9000 26 0.3000 ... 131 0.3005 ... ... ... ... ... ... ... 20 1.4625 41 0.4875 ... 146 0.3655 21 1.5000 42 0.5000 ... 147 0.3699

[0121] Table 3 reference interval of key index

[0122] Indicator 1 Indicator 2 Indicator 3 ... Indicator 6 Indicator 7 [0.7500，0.7875] [0.2500,0.2625] [0.0000,0.0765] ... [0.6250,0.6716] [0.2832,0.2875] [0.7875,0.8250] [0.2625,0.2750] [0.0765,0.1530] ... [0.6716,0.7183] [0.2875,0.2919] [0.8250,0.8625] [0.2750,0.2875] [0.1530,0.2295] ... [0.7183,0.7649] [0.2919,0.2962] [0.8625,0.9000] [0.2875,0.3000] [0.2295,0.3060] ... [0.7649,0.8115] [0.2962,0.3005] ... ... ... ... ... ... [1.4625,1.5000] [0.4875,0.5000] [1.4535,1.5300] ... [1.5110,1.5576] [0.3005,0.3699]

[0123] S3, constructing an initial confidence rule base based on the key indicator information and the expert knowledge information, and optimizing the confidence of the rules, the weight of the rules and the reliability of the indicators in the initial rule base by using a P-CMA-ES optimization method to obtain the optimized confidence rule base, and testing by using the optimized confidence rule base.

[0124] Specifically, an initial confidence rule base is constructed based on the key indicator information and the expert knowledge information, and the initial parameters are given by the expert knowledge. Since the parameters in the constructed initial confidence rule base are given by the expert, and are affected by the limitation of the cognitive ability of the expert, the initial method is difficult to adapt to the actual working conditions. Therefore, in order to further optimize and learn the method parameters of the initial confidence rule base, 167 groups of data are obtained in this embodiment, 141 groups of data are randomly selected from each of the 167 groups of data as method training data, and the remaining data is test data. The fault detection result of the optimized confidence rule base sets two utility values, 1 represents no fault, and 2 represents fault. The final output result of the trained confidence rule base is as shown in Figure 4 Figure 4 The output result diagram of the three-self laser inertial measurement unit fault detection method based on the interval value confidence rule base provided by the embodiment of the application is shown in Table 4.

[0125] Table 4 Parameters of the trained confidence rule base

[0126] Rule sequence number Indicator reliability Rule weight Rule output 1 0.7867 0.1222 {0.5539 0.4461} 2 0.9784 0.6560 {0.6059 0.3941} 3 0.7894 0.0147 {0.3431 0.6569} 4 0.8442 0.8344 {0.3441 0.6559} 5 0.1265 0.3120 {0.7057 0.2943} 6 0.6340 0.2814 {0.5899 0.4101} 7 0.7561 0.9405 {0.9203 0.0797} 8 0.2669 0.3966 {0.4639 0.5361} 9 0.5491 0.8446 {0.0846 0.9154} 10 0.4305 0.4902 {0.8033 0.1967} 11 0.0498 0.5481 {0.0095 0.9905} 12 0.5310 0.6131 {0.3906 0.6094} 13 0.7992 0.4752 {0.7500 0.2500} 14 0.1652 0.4673 {0.5305 0.4695} 15 0.9950 0.7727 {0.5053 0.4947} 16 0.6744 0.0947 {0.2119 0.7881} 17 0.6471 0.8607 {0.5148 0.4852} 18 0.7697 0.6218 {0.3400 0.6600} 19 0.2216 0.2812 {0.5894 0.4106} 20 0.6538 0.7930 {0.8747 0.1253} ... ... ... ... 137 0.2516 0.4298 {0.5795 0.4205} 138 0.4665 0.4008 {0.5541 0.4459} 139 0.1842 0.1408 {0.7360 0.2640} 140 0.8520 0.8836 {0.0234 0.9766}

[0127] From Figure 4 and Table 4, it can be seen that the trained three-self laser inertial measurement unit fault detection method can accurately detect the faults of the three-self laser inertial measurement unit, and the accuracy rate reaches 92.31%. In order to prove the robustness of the fault detection method, the experiment is repeated for 30 times, and the average fault detection accuracy rate obtained is 91.54%.

[0128] In this embodiment, the information of the gyroscope and the accelerometer of the laser inertial measurement unit is processed as the key indicator information first; then the confidence rule base is constructed to form the three-self laser inertial measurement unit fault detection method, and the key indicator information is taken as the input of the method, and the fault detection utility conversion result is taken as the final output of the method; finally, the practicability of the three-self laser inertial measurement unit fault detection method based on the interval confidence rule base is verified by simulation experiment. The method changes the original single indicator reference value into an indicator interval reference value to solve the problem of the explosion of the number of combined rules, and adds an indicator reliability parameter to improve the reliability of the indicator information, and solves the problems of the explosion of the number of combined rules of the indicators and the unreliability of the indicators in the confidence rule base in the three-self laser inertial measurement unit fault detection.

[0129] ​The above is further detailed description of the present application in combination with specific preferred embodiments, and cannot be deemed as limitation of the specific implementation of the present application to these descriptions. For those skilled in the art to which the present application belongs, without departing from the concept of the present application, a number of simple deductions or substitutions can be made, and all should be deemed as falling within the protection scope of the present application.

Claims

1. A three-axis laser inertial measurement unit fault detection method based on interval-valued confidence rule base, characterized in that, The method comprises the steps of: determining several key index information of the three self-excited laser inertial measurement unit, and setting a reference interval according to each of the key index information; activating corresponding rules in the optimized confidence rule base according to the reference interval to which each of the key index information belongs, to obtain several activated rules, comprising: constructing an initial confidence rule base based on the key index information and expert knowledge information, the kth rule in the initial confidence rule base is: BR k : IF x1∈[a1,b1] ∨x2∈[a2,b2] ∨...∨x MK ∈[a MK ,b MK ] THEN result is {(D1, β 1,k ),(D2, β 2,k ),...,(D N ,β N,k )} WITH rule weightθ k AND indicator reliability r Mk ,rule availabilityΔ k wherein x1, x2, …, x Mk represents the key index information of the three self-excited laser IMU, M k represents the total number of key index information, [a1, b1], [a2, b2], …, [a Mk , b Mk ] represents the interval reference value of the key index information, D1, D2, …, D N represents N detection results, β 1,k , β 2,k , …, β N,k represents the confidence of each rule detection result, θ k represents the weight of the kth rule, r Mk represents the index reliability, L represents the number of rules in the confidence rule base, Δ k represents the availability of the kth rule; wherein ψ represents a confidence factor, ψ is equal to the absolute value of the difference between the square of the confidence value; optimizing the confidence of the rules, the weight of the rules and the index reliability in the initial confidence rule base by using a projection covariance matrix adaptive strategy optimization method, to obtain the optimized confidence rule base; judging the reference interval into which the key index information falls, and activating the rules corresponding to the reference interval in the optimized confidence rule base, to obtain the several activated rules; fusing the several activated rules according to the weight, the confidence and the availability of each of the activated rules, to obtain a rule synthesis result; performing utility conversion by using the rule synthesis result and the utility value of the detection result, to obtain a utility conversion result; judging whether the three self-excited laser inertial measurement unit has a fault according to the utility conversion result, to obtain a detection result.

2. The interval-valued confidence rule-based three-axis laser inertial measurement unit fault detection method according to claim 1, characterized in that, The several key index information comprises: the X-axis cumulative pulse amount of the gyroscope, the Y-axis cumulative pulse amount of the gyroscope, the Z-axis cumulative pulse amount of the gyroscope, the X-axis cumulative pulse amount of the accelerometer, the Y-axis cumulative pulse amount of the accelerometer, the Z-axis cumulative pulse amount of the accelerometer, and the unit time cumulative pulse amount of the gyroscope and the accelerometer.

3. The interval-valued confidence rule-based three-axis laser inertial measurement unit fault detection method according to claim 1, characterized in that, Setting the reference interval according to each of the key index information comprises: setting an initial reference value of each key index information; equally dividing the initial reference values into the reference intervals by an expert system according to the number of the initial reference values.

4. The interval-valued confidence rule-based three-axis laser inertial measurement unit fault detection method according to claim 1, characterized in that, In the process of optimizing the confidence of the rules, the weight of the rules and the index reliability in the initial rule base, the confidence of the rules, the weight of the rules and the index reliability satisfy the following constraint conditions: min MSE(θ k ,β i,k ,r i ) st. 0 < θ k ≤ 1, k = 1, 2,..., L 0 < β i,k ≤ 1, i = 1,.., N, k = 1, 2,..., L 0 < r i ≤ 1, i = 1,..., M k wherein, T represents the number of training sample sets, output estimated and output acutal represents the predicted utility value and the true utility value, θ k represents the weight of the kth rule, L represents the number of rules in the confidence rule base, β i,k represents the confidence of each rule detection result, r i represents the index reliability, M k represents the total number of key index information.

5. The interval-valued confidence rule-based three-axis laser inertial unit fault detection method according to claim 1, characterized in that, Fusing the several activated rules according to the weight, the confidence and the availability of each of the activated rules, to obtain a rule synthesis result, comprising: calculating the basic probability mass of each activated rule according to the weight and the confidence of each activated rule: where m n,k denotes the basic probability mass of the kth activation rule, θ k denotes the weight of the kth rule, β n,k denotes the confidence of the nth rule detection result, n e {1,...,N}, Θ denotes a recognition framework, and Θ = {D1,...,D N}, D1,D2,...,D N denotes N detection results; fusing the several activated rules according to the basic probability mass and the availability of each of the activated rules, to obtain a rule synthesis result: where p n,k denotes the confidence of the synthesized rule, denotes the joint probability mass, Δ i denotes the availability of the ith rule, Δ j denotes the availability of the jth rule, m n,j denotes the basic probability mass of the jth rule, m n,i denotes the basic probability mass of the ith rule, m A,j denotes the basic probability mass of the jth rule evaluating to A, m B,i denotes the basic probability mass of the ith rule evaluating to B, A denotes an evaluation outcome case in Θ, B denotes an evaluation outcome case in Θ, denotes the basic probability mass of the kth rule evaluating to A.

6. The interval-valued confidence rule-based three-axis laser inertial unit fault detection method according to claim 1, characterized in that, Performing utility conversion by using the rule synthesis result and the utility value of the detection result, to obtain a utility conversion result, comprising: setting the utility value of the detection result; Performing utility conversion by using the rule synthesis result and the utility value of the detection result, to obtain a utility conversion result: where u(D n ) represents the utility value of the detection result D n , and p n,k represents the confidence of the synthesized rule.

7. The interval-valued confidence rule-based three-axis laser inertial unit fault detection method according to claim 1, characterized in that, Judging whether the three self-excited laser inertial measurement unit has a fault according to the utility conversion result, to obtain a detection result, comprising: when the utility conversion result is located in a first interval, the detection result is a first numerical value; when the utility conversion result is located in a second interval, the detection result is a second numerical value; According to the first value and the second value, it is judged whether the three self-excited laser inertial measurement unit is faulty, and a detection result is obtained.

Images