Underwater propeller fault feature enhancement method and enhancement system based on stochastic resonance system

Abstract

The application discloses an underwater propeller fault feature enhancement method based on a stochastic resonance system, carries out underwater robot propeller fault test, and collects dynamic signals of the underwater robot; the dynamic signals are input into a bistable stochastic resonance system to obtain a Langevin equation; the fourth-order Runge-Kutta method is used to solve the Langevin equation to obtain a stochastic resonance system output signal; the stochastic resonance system output signal is processed based on a modified Bayesian algorithm to obtain a fault feature sequence, and a maximum value in the fault feature sequence is taken as a fault feature value; structure parameters of the bistable stochastic resonance system are optimized to obtain multiple fault feature values; and a fault feature sequence corresponding to a maximum value in the multiple fault feature values is taken as a fault feature enhancement result. Through optimization of the structure parameters of the stochastic resonance system, multiple fault feature values are obtained, and the maximum value is selected as an enhanced structure, so that the fault feature effect is more obvious.

Description

TECHNICAL FIELD

[0001] The present application relates to underwater robot propeller fault detection, in particular to an underwater propeller fault feature enhancement method and system based on a stochastic resonance system. BACKGROUND

[0002] As the most important power source of underwater robots, underwater propellers are subjected to complex and variable marine environments and heavy loads during operation, and are therefore prone to motor abnormalities, damage, short circuits caused by water entering the propeller, and propeller blade entanglement with water plants and stop rotation. The failure of the underwater propeller will cause the underwater robot to be unable to complete the task, and in severe cases, an accident will occur. Therefore, fault diagnosis of the underwater propeller has great research value and significance. Generally speaking, fault feature extraction is the first link of propeller fault diagnosis.

[0003] In the prior art, a paper entitled "AUV Propeller Fault Feature Extraction and Fusion under Random Disturbance" published in the Journal of Huazhong University of Science and Technology (Natural Science Edition) proposes a method of extracting fault features based on wavelet approximation components. First, the velocity signal is decomposed by wavelet, and the wavelet approximation component is obtained by wavelet reconstruction of the scale coefficient after decomposition. Based on the modified Bayesian algorithm, the fault features are extracted from the wavelet approximation component. For example, the Chinese patent application No. CN202111493572.4 discloses a weak fault feature extraction method for autonomous underwater robot propeller. The method proposes a fault feature enhancement method based on improved MVMD and modified Bayesian classification algorithm (MB). The method judges the Gaussianity of all modalities of multi-source state signals and control signals by negative entropy, optimizes the parameters, completes noise reduction, and extracts and enhances the fault features based on the modified Bayesian algorithm. The above method highlights the fault signal waveform in the signal by reducing the interference noise in the signal, thereby enhancing the modified Bayesian feature of the underwater propeller fault. This method only removes the interference noise in the signal and does not use the energy of the interference noise to enhance the fault signal, so the effect is limited in terms of the difference and ratio of the fault feature value and the noise feature value.

[0004] In the paper "Autonomous Underwater Vehicle Fault Feature Enhancement Method" published in Harbin Engineering University Journal, a method based on adaptive stochastic resonance is proposed to enhance the fault feature. The method takes the original data of AUV longitudinal velocity as the driving force of the bistable stochastic resonance system, changes the potential barrier height to achieve stochastic resonance of AUV longitudinal velocity signal, and transfers the energy of external random interference signal to fault signal to achieve the purpose of enhancing signal fault feature. The method enhances the fault correction Bayesian feature of underwater thruster by transferring the energy of interference noise to the fault signal. However, this method only transfers part of the energy of interference noise to the fault signal, and the remaining interference noise energy continues to affect the fault signal, so that the difference and ratio of fault feature value and noise feature value are limited. SUMMARY

[0005] The purpose of the application is to solve the above problems, and the application provides an underwater thruster fault feature enhancement method based on a stochastic resonance system which significantly enhances the fault feature.

[0006] The application also provides an underwater thruster fault feature enhancement system based on a stochastic resonance system.

[0007] Technical scheme: To solve the above problems, the application adopts an underwater thruster fault feature enhancement method based on a stochastic resonance system, which includes the following steps:

[0008] (1) Perform underwater robot thruster fault test and collect dynamic signals of underwater robots;

[0009] (2) Input the dynamic signal into the bistable stochastic resonance system to obtain the Langevin equation;

[0010] (3) Use the fourth-order Runge-Kutta method to solve the Langevin equation to obtain the output signal of the stochastic resonance system;

[0011] (4) Process the output signal of the stochastic resonance system based on the modified Bayesian algorithm to obtain the fault feature sequence, and take the maximum value in the fault feature sequence as the fault feature value;

[0012] (5) Optimize the structure parameters of the bistable stochastic resonance system, and repeat steps (2) to (4) to obtain multiple fault feature values;

[0013] (6) Take the fault feature sequence corresponding to the maximum value in the multiple fault feature values as the fault feature enhancement result.

[0014] Further, the step (5) sets the structure parameters of the bistable stochastic resonance system and inputs them into the bistable stochastic resonance system in a permutation and combination way to obtain bistable stochastic resonance system with different structural parameters, thereby obtaining a fault characteristic value.

[0015] Further, the step (5) uses an ant colony algorithm to adjust the structural parameters to optimize the bistable stochastic resonance system, and the value range [A min , A max ] of the structural parameter a and the value range [B min , B max ] of the structural parameter b of the bistable stochastic resonance system are divided into grids, the ants are distributed to the grid units according to the random placement principle, each unit corresponds to a set of structural parameter combination [a, b], and the ants on each column search for the next unit position according to the random principle, and the structural parameter combination [a, b] corresponding to the unit is input into the bistable stochastic resonance system.

[0016] Further, the step (6) updates the pheromone concentration of the unit where the ant is located, and whether the unit selected by the ant on each column converges to the same unit is judged. If the unit converges to the same unit, the unit with the highest pheromone concentration in each column is selected, otherwise the ant on each column searches for the next unit position according to the random principle; the value range of the structural parameter value corresponding to the unit with the highest pheromone concentration is reduced, and then optimization is performed until the evaluation function value converges or the maximum iteration number is reached, to obtain the optimal structural parameters a h and b h , at this time the fault characteristic value is maximum, and the fault characteristic sequence corresponding thereto is taken as the fault characteristic enhancement result.

[0017] Further, the step (5) uses a genetic algorithm to adjust the structural parameters to optimize the bistable stochastic resonance system, and the structural parameter combination [a, b] of the bistable stochastic resonance system is taken as an individual, the optimization range of the structural parameter a is [A min , A max ], the optimization range of the structural parameter b is [B min , B max ], N initial individuals are randomly generated, the roulette selection method is used to select the parent individuals, the decoded a and b values of the parent individuals are input into the bistable stochastic resonance system.

[0018] Further, the roulette selection method is used to select the parent individuals, the selected parent individuals are subjected to crossover and mutation operations to obtain offspring individuals, population evolution optimization is performed, the offspring individuals are taken as the parent individuals of the next generation for iteration until the fitness function converges or the maximum iteration number is reached, to obtain the optimal structural parameters a h and b hAt this time, the fault characteristic value is maximum, and the corresponding fault characteristic sequence is taken as a fault characteristic enhancement result.

[0019] Further, the dynamic signal is subjected to multi-resolution wavelet decomposition in the step (2), and a wavelet approximation component is obtained.

[0020] An underwater propeller fault feature enhancement system based on a stochastic resonance system comprises a collecting module for collecting dynamic signals of an underwater robot during a fault test of an underwater robot propeller.

[0021] The signal processing module is configured to input the dynamic signals into the bistable stochastic resonance system to obtain a Langevin equation, solve the Langevin equation by using a fourth-order Runge-Kutta method to obtain a stochastic resonance system output signal, and process the stochastic resonance system output signal based on a modified Bayesian algorithm to obtain a fault characteristic sequence, and take a maximum value in the fault characteristic sequence as a fault characteristic value.

[0022] The optimization module is configured to optimize structure parameters of the bistable stochastic resonance system, and process the dynamic signals through the signal processing module again according to the optimized structure parameters.

[0023] The fault enhancement module is configured to take a fault characteristic sequence corresponding to the maximum value in the multiple fault characteristic values as a fault characteristic enhancement result.

[0024] The present application has the following advantages over the prior art: the structure parameters of the stochastic resonance system are optimized to obtain multiple fault characteristic values, and the maximum value is selected as an enhanced structure, so that the fault characteristic effect is more obvious. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 is an enhancement method flowchart of structure parameter permutation and combination optimization in the present application.

[0026] Figure 2 is an enhancement method flowchart of structure parameter adjustment by an ant colony algorithm in the present application.

[0027] Figure 3 is an enhancement method flowchart of structure parameter adjustment by a genetic algorithm in the present application.

[0028] Figure 4 is an enhancement method flowchart of wavelet decomposition combined with structure parameter permutation and combination optimization in the present application.

[0029] Figure 5 is an enhancement method flowchart of wavelet decomposition combined with structure parameter adjustment by an ant colony algorithm in the present application.

[0030] Figure 6It is the enhanced method flow diagram of wavelet decomposition combined with genetic algorithm to adjust structural parameters in the application.

[0031] Figure 7 It is the underwater robot longitudinal velocity signal waveform diagram collected in the application.

[0032] Figure 8 It is the fault feature enhancement result diagram of embodiment 1 in the application.

[0033] Figure 9 It is the result diagram of random resonance system and the fault feature enhancement result diagram based on ant colony algorithm in embodiment 2 in the application.

[0034] Figure 10 It is the result diagram of random resonance system and the fault feature enhancement result diagram based on genetic algorithm in embodiment 3 in the application.

[0035] Figure 11 It is the comparison diagram of fault feature extraction result of embodiment 4 and embodiment 1 in the application.

[0036] Figure 12 It is the result diagram of random resonance system and the fault feature enhancement result diagram of wavelet decomposition combined with ant colony algorithm in embodiment 5 in the application.

[0037] Figure 13 It is the result diagram of random resonance system and the fault feature enhancement result diagram of wavelet decomposition combined with genetic algorithm in embodiment 6 in the application.

[0038] Figure 14 It is the process diagram of wavelet decomposition combined with genetic algorithm to optimize random resonance system to process longitudinal velocity signal in the application. DETAILED DESCRIPTION

[0039] Embodiment 1:

[0040] As shown in the figure, the underwater propeller fault feature enhancement method based on random resonance system in the embodiment comprises the following steps: Figure 1

[0041] Firstly, the underwater robot propeller fault test is carried out, and the underwater robot dynamic signal u(k) is recorded;

[0042] Secondly, the bistable random resonance system structural parameters a=A min , b=B min are set;

[0043] Thirdly, the underwater robot dynamic signal u(k) is input into the bistable random resonance system, and the Langevin equation is obtained, as shown in formula (1):

[0044] x′(k)=ax(k)-bx 3 ​(k)+u(k) (1)

[0045] In the formula, {x(k)} = [x(1) x(2) ... x(n)] is the output signal of the stochastic resonance system, a and b are the structural parameters of the system, {u(k)} = [u(1) u(2) ... u(n)] is the input signal of the stochastic resonance system (dynamic signal of the underwater robot), and n is the data length of the dynamic signal of the underwater robot.

[0046] The fourth step is to solve the Langevin equation using the fourth-order Runge-Kutta method to obtain the output signal {x(k)} of the stochastic resonance system, where the Runge-Kutta method is shown in equation (2):

[0047]

[0048] In the formula, h is the step size, and f is the sampling frequency. s The reciprocal of f, i.e., h = 1 / f s k is a positive integer, k = 1, 2, 3…n, x(k) is the output value of the random resonance system output signal {x(k)} = [x(1) x(2) … x(k) … x(n)] at each time step, and u(k) is the sampled value of the underwater robot dynamic signal {u(k)} = [u(1) u(2) … u(k) … u(n)] at each time step;

[0049] The fifth step involves processing the output signal {x(k)} = [x(1) x(2) …x(n)] of the stochastic resonance system using the modified Bayesian algorithm to obtain the fault feature sequence {d}. x (k)}=[d x (1) d x (2) … d x (n)], will [d x (1) d x (2) …d x The maximum value d in [n] xmax As a fault characteristic value;

[0050] Step 6: Set the structural parameters of the bistable stochastic resonance system. And use permutations and combinations Input it into the stochastic resonance system in this way, and repeat steps three through five to obtain... Each fault characteristic value

[0051] Step 7: Filter out fault characteristic values ​​{d} xmax The maximum value d in (m)} xb The corresponding fault feature sequence {d xb (k)}=[dxb (1) d xb (2) … d xb (n)] as a fault feature enhancement result.

[0052] Example 2:

[0053] As Figure 2 shown, the underwater thruster fault feature enhancement method based on a stochastic resonance system in this embodiment adopts an ant colony algorithm to optimize the structural parameters of the stochastic resonance system, including the following steps:

[0054] First, conduct an underwater robot thruster fault test and record the underwater robot dynamic signal u(k);

[0055] Second, set the bistable stochastic resonance system structural parameters a = A r , b = B r (A r is a random number in the value range of a, and B r is a random number in the value range of b);

[0056] The third to fifth steps of Example 2 are the same as the third to fifth steps of Example 1; hereinafter, no further elaboration is made;

[0057] Sixth, use the ant colony algorithm to adjust the system structural parameters to optimize the stochastic resonance system. First, set the number of ants in the ant colony to M, the number of grid divisions to N, the total number of units to P, the initial concentration of pheromone in each unit to τ, and the maximum number of iterations to T max , then divide the value range of the system parameter a [A min , A max ] and the value range of b [B min , B max ] into grids, and let get the optimization range of a the optimization range of b According to the principle of random placement, distribute the ants to the grid units, with each unit corresponding to a set of system parameter combination [a, b], and the grid is where Here, the number of ants and the number of grid units are the same;

[0058] Seventh, calculate the pheromone concentration of each unit in the grid j = 1, 2, 3…N, calculate the state transition probability of each unit in each column of the grid according to the formula ​Then, according to this probability, the ants in each column are made to search for the next cell position according to a random principle;

[0059] Step 8: The ant moves to the selected cell position. Input the system parameter combination [a, b] corresponding to this unit into the stochastic resonance system, and repeat steps 3 to 5 to obtain H*D fault characteristic values ​​{d xmax (m)}=[d xmax (1) d xmax (2) … d xmax [H*D], the fault characteristic value d xmax As an evaluation function value, when the parameters of the stochastic resonance system are optimal, the fault characteristic value d xmax maximum;

[0060] Step 9: Calculate the evaluation function values ​​for each unit. According to the formula Update the pheromone concentration of the unit.

[0061] In the formula, τ ij (t) represents the pheromone concentration of the cell in the i-th row and j-th column of the grid sequence at time t, Δτ ij (t) represents the change in unit concentration calculated at time t, ρ is the pheromone volatility coefficient, and Q is the pheromone intensity;

[0062] Step 10: Check whether the cells selected by the ants in each column converge to the same cell [τ]. i1 τ i2 …τ iN If the pheromone concentrations converge to the same cell, select the cell τ with the highest pheromone concentration in each column. max Otherwise, return to step 7 and continue execution;

[0063] Step 11: For the unit τ with the highest pheromone concentration... max The system parameter value corresponding to the location [a] max b max Narrow down the range of its values, then return to step 6 to continue execution until the evaluation function value converges or the maximum number of iterations T is reached. max To obtain the optimal parameter a h With b h At this point, the maximum fault characteristic value is d. xb The corresponding fault feature sequence {d xb (k)}=[d xb (1)d xb (2) … d xb [(n)] is the result of the fault feature enhancement in Example 2.

[0064] Example 3:

[0065] like Figure 3 As shown in this embodiment, a method for enhancing the fault characteristics of an underwater thruster based on a stochastic resonance system is proposed. This method employs a genetic algorithm to adjust structural parameters and optimize the bistable stochastic resonance system, and includes the following steps:

[0066] The first step is to conduct a fault test on the underwater robot's thruster and record the underwater robot's dynamic signal u(k);

[0067] The second step is to set the structural parameters a = A of the bistable stochastic resonance system. r b = B r (A r B is a random number within the range of values ​​for a. r (where b is a random number within the range of values);

[0068] Steps three through five in this embodiment are the same as steps three through five in embodiment 1; therefore, they will not be repeated here.

[0069] The sixth step is to use a genetic algorithm to adjust the system parameters and optimize the stochastic resonance system. First, set the number of individuals N in the population, the encoding length L, and the crossover probability P. c Probability of mutation P m and the maximum number of iterations T max Taking the parameter combination [a, b] of the stochastic resonance system as an individual, the optimization range of parameter a is [A...]. min A max The optimization range for parameter b is [B]. min B max Randomly generate N initial individuals. Each individual consists of two L-bit binary numbers. Let a and b represent the encoding forms respectively. The conversion from encoding to actual value is obtained by the following formula:

[0070]

[0071] In the formula, A 11 B represents the decimal number corresponding to the binary number of individual 'a'. 11 The decimal number corresponding to the binary number of individual b;

[0072] Step 7: Calculate the parent population fitness by decoding the a and b values ​​of the parent individuals. Input a random resonance system, repeat steps 3 to 5, and obtain N fault characteristic values ​​{d}. xmax (m)}=[d xmax (1)d xmax (2)…d xmax(N)], the fault characteristic value d xmax As the fitness function value, when the parameters of the stochastic resonance system are optimal, the fault characteristic value d xmax maximum;

[0073] Step 8: Select offspring individuals using a roulette wheel selection method to choose from parent individuals. The probability of each individual being selected is... This increases the probability of selecting individuals with higher fitness values. Then, the selected individuals are crossoverdone. Operations to obtain offspring individuals

[0074] Step 9: Population evolution optimization. Offspring individuals become parents for the next generation. Steps 7 and 8 are repeated until the fitness function converges or the maximum number of iterations is reached, obtaining the optimal parameter 'a'. h With b h At this point, the maximum fault characteristic value is d. xb The corresponding fault feature sequence {d xb (k)}=[d xb (1) d xb (2) … d xb [(n)] is the result of the fault feature enhancement in Example 3.

[0075] Example 4:

[0076] like Figure 4 As shown in this embodiment, an underwater thruster fault feature enhancement method based on a stochastic resonance system uses wavelet decomposition to process dynamic signals, and includes the following steps:

[0077] The first step is to conduct a fault test on the underwater robot's thruster and record the underwater robot's dynamic signal u(k);

[0078] The second step involves performing multi-resolution wavelet decomposition on the underwater robot's dynamic signal u(k), with the decomposition level being i, to obtain the wavelet approximate components {u A (k)}=[u A (1) u A (2) … u A (n)];

[0079] The third step is to set the structural parameters a = A of the bistable stochastic resonance system. min b = B min ;

[0080] The fourth step is to approximate the wavelet components {u} A When (k)} is input into a bistable stochastic resonance system, the Langevin equation is obtained, as shown below:

[0081] xA (k) = ax A (k) = bx A 3 (k) = u A (k) (5)

[0082] where {x A (k)} = [x A (1) x A (2) … x A (n)] is the output signal of the stochastic resonance system, a and b are the structural parameters of the system, {u A (k)} = [u A (1) u A (2) … u A (n)] is the input signal of the stochastic resonance system;

[0083] In the fifth step, the fourth-order Runge-Kutta method is used to solve the Langevin equation to obtain the output signal {x A (k)} of the stochastic resonance system, and the Runge-Kutta formula is as follows:

[0084]

[0085] where h is the step size, f s is the sampling frequency, i.e. h = 1 / f s , k is a positive integer, k = 1, 2, 3…n, x A (k) is the output value of the stochastic resonance system at each time, x A (k) = [x A (1) x A (2) … x A (k) … x A (n)], u A (k) is the sampling value of the wavelet approximation component at each time, u A (k) = [u A (1) u A (2) … u A (k) … u A (n)];

[0086] In the sixth step, the output signal {x A (k)} = [x A (1) x A (2) … x A (n)] of the stochastic resonance system is processed based on the modified Bayesian algorithm to obtain the fault feature sequence {d xA (k)} = [d xA (1) d xA (2) … dxA (n)], will [d xA (1)d xA (2) … d xA The maximum value d in [n] xAmax As a fault characteristic value;

[0087] Step 7: Set the structural parameters of the bistable stochastic resonance system And use permutations and combinations By inputting it into the stochastic resonance system in this manner, and repeating steps five through seven, we obtain... Each fault characteristic value

[0088] Step 8: Filter out fault characteristic values ​​{d} xAmax The maximum value d in (m)} xAb The corresponding fault feature sequence {d xAb (k)}=[d xAb (1) d xAb (2) … d xAb [(n)] is the result of the fault feature enhancement in Example 4.

[0089] Example 5:

[0090] like Figure 5 As shown in this embodiment, a method for enhancing the fault features of an underwater thruster based on a stochastic resonance system is proposed. This method uses wavelet decomposition to process dynamic signals and combines it with an ant colony algorithm to optimize the structural parameters of the stochastic resonance system. The specific steps are as follows:

[0091] The first and second steps of Example 4 are adopted, and the third step is to set the structural parameter a = A of the bistable stochastic resonance system. r b = B r (A r B is a random number within the range of values ​​for a. r (a random number within the range of values ​​for b); then combine steps four to six of Example 4 with steps six to eleven of Example 2; the specific steps will not be repeated here.

[0092] Example 6:

[0093] like Figure 6 As shown in this embodiment, a method for enhancing the fault characteristics of an underwater thruster based on a stochastic resonance system is proposed. This method uses wavelet decomposition to process dynamic signals and combines it with a genetic algorithm to optimize the structural parameters of the stochastic resonance system. The specific steps are as follows:

[0094] The first and second steps of Example 4 are adopted, and the third step is to set the structural parameter a = A of the bistable stochastic resonance system. r b = B r(A r is a random number in the range of values of a, B r is a random number in the range of values of b); and then combining the fourth to sixth steps of Embodiment 4 with the sixth to ninth steps of Embodiment 3; the specific step contents are not described here again.

[0095] Embodiment 7:

[0096] As shown in Figure 7 , the underwater robot propeller failure test is carried out to obtain the underwater robot longitudinal speed signal. During the test, the underwater robot target speed is set to 0.3 m / s through closed-loop control, and the underwater robot is started from rest. The underwater robot speed gradually increases, and at the 100th beat, the underwater robot reaches 0.3 m / s and starts to run at a uniform speed of 0.3 m / s. At the 200th beat, the propeller output loss of 30% is simulated by using the fault simulation method until the test is completed.

[0097] To study the propeller failure data of the underwater robot when the expected speed is stable, the data of the first to 100th beats are discarded, and the speed data between the 101st to 500th beats are selected as the underwater robot speed signal for research. A underwater propeller failure correction Bayesian feature enhancement method based on a stochastic resonance system is adopted. The bistable stochastic resonance system structure parameters a and b are set to the range [0.1, 20], then a = [0.1 0.2 0.3 … 19.9 20], b = [0.1 0.2 0.3 … 19.9 20], and the maximum value of the fault feature value is 20.36. The corresponding structure parameters a = 6.2 and b = 6.6, and the corresponding fault feature results are shown in Figure 8 .

[0098] When the ant colony algorithm is used to optimize the parameters of the stochastic resonance system, the bistable stochastic resonance system structure parameters a and b are set to the range [0, 20]. The range is divided into grids, the number of segments N = 10, the total number of units P = 100, the number of ants M = 100, the initial concentration of pheromone τ = 1, the volatilization coefficient ρ = 0.2, and the maximum iteration number T max = 100. The results of the ant colony algorithm optimizing the parameters of the stochastic resonance system are shown in Figure 9 (a). When a = 0.08 and b = 0.05, the fault feature value reaches the maximum value of 28.46. The corresponding fault feature results are shown in Figure 9 (b).

[0099] When the genetic algorithm is used to optimize the parameters of the stochastic resonance system, the bistable stochastic resonance system structure parameters a and b are set to the range [0, 20]. The number of individuals in the population N = 50, and the maximum iteration number T max= 100, cross probability P c = 0.9, mutation probability P m = 0.01, genetic algorithm optimization random resonance system parameter results as shown in Figure 10 (a), when a = 6.39 and b = 5.25, the fault characteristic value reaches the maximum value, at this time the fault characteristic value is 28.60, the corresponding fault characteristic results as shown in Figure 10 (b).

[0100] The underwater robot speed signal is subjected to multi-resolution wavelet decomposition, the decomposition layer is 3, the wavelet approximation component {u A (k)} is obtained, the wavelet approximation component {u A (k)} is studied, the wavelet decomposition and random resonance system enhancement method for the fault correction Bayesian feature of the underwater propeller is adopted, the parameter setting is the same as that of the underwater propeller fault correction Bayesian feature enhancement method based on random resonance system, the maximum value of the fault characteristic value is 39.85, the corresponding structural parameters a = 6.0, b = 19.8, the corresponding fault characteristic results as shown in Figure 11 .

[0101] When the dynamic signal wavelet decomposition is combined with the ant colony algorithm optimization random resonance system, the parameter setting is the same as that of the ant colony algorithm optimization random resonance system, the ant colony algorithm optimization random resonance system results as shown in Figure 12 (a), when a = 0.05 and b = 0.02, the fault characteristic value reaches the maximum value, at this time the fault characteristic value is 44.87, the corresponding fault characteristic results as shown in Figure 12 (b).

[0102] When the dynamic signal wavelet decomposition and the genetic algorithm optimization random resonance system are adopted, the parameter setting is the same as that of the genetic algorithm optimization random resonance system, the genetic algorithm optimization random resonance system results as shown in Figure 13 (a), when a = 6.26 and b = 13.65, the fault characteristic value reaches the maximum value, at this time the fault characteristic value is 45.48, the corresponding fault characteristic results as shown in Figure 13 (b).

[0103] As shown in Figure 14 , the longitudinal velocity signal is subjected to wavelet decomposition combined with genetic algorithm optimization random resonance system parameters, the Figure 14 (a) between the 101-500 beats of the speed data is selected as the underwater robot speed signal u(k), the results as shown in Figure 14 (b). Figure 14 (b) data u(k) is subjected to 3-layer wavelet decomposition, the wavelet approximation component {u A (k)} is obtained, as shown in Figure 14(c) shown. The wavelet approximation component {u A (k) is input into the stochastic resonance system with random parameters a=1 and b=1, and the fault feature is obtained, as shown in Figure 14 (d). The stochastic resonance system is optimized based on the genetic algorithm. The parent population fitness, offspring selection fitness, offspring crossover fitness, offspring mutation fitness of the algorithm iteration 1 time, the parent population fitness, offspring selection fitness, offspring crossover fitness, offspring mutation fitness of the algorithm iteration 100 times, and the optimization effect after the algorithm reaches the maximum iteration number are shown in Figure 14 (e). The stochastic resonance system is optimized based on the genetic algorithm, and the optimal parameters a and b are obtained, corresponding to the maximum fault feature value. The corresponding fault feature is shown in Figure 14 (f).

[0104] The fault feature extraction results of the embodiments 1-6 of the present scheme are shown in Table 1.

[0105] Table 1: Fault feature extraction results of embodiments 1-6

[0106]

[0107] The results in Table 1 show that the method of the embodiment 1 of the present scheme is better than the known method in enhancing the fault feature value, the difference value and the ratio of the feature values. The methods of the embodiments 2 and 3 of the present scheme are better than the method of the embodiment 1 of the present scheme in enhancing the fault feature value, the difference value and the ratio of the feature values. The method of the embodiment 4 of the present scheme is better than the method of the embodiment 1 of the present scheme in enhancing the fault feature value, the difference value and the ratio of the feature values. The methods of the embodiments 5 and 6 of the present scheme are better than the methods of the embodiments 2 and 3 of the present scheme in enhancing the fault feature value, the difference value and the ratio of the feature values, respectively. After the noise is processed by the wavelet decomposition and the stochastic resonance for noise reduction and energy transfer, the fault feature of the underwater propeller is extracted. Compared with the wavelet method and the stochastic resonance method, the correction of the Baysian feature is enhanced, and the effect of enhancing the fault feature is more obvious.

Claims

1. A method for enhancing the fault characteristics of underwater thrusters based on stochastic resonance systems, characterized in that, Includes the following steps: (1) Conduct underwater robot thruster failure tests and collect dynamic signals of the underwater robot; (2) Input the dynamic signal into the bistable stochastic resonance system to obtain the Langevin equation; (3) The Langevin equation is solved by the fourth-order Runge-Kutta method to obtain the output signal of the stochastic resonance system; (4) The output signal of the stochastic resonance system is processed based on the modified Bayesian algorithm to obtain the fault feature sequence, and the maximum value in the fault feature sequence is taken as the fault feature value; (5) Optimize the structural parameters of the bistable stochastic resonance system by repeating steps (2) to (4) to obtain multiple fault characteristic values; (6) The fault feature sequence corresponding to the maximum value among multiple fault feature values ​​is used as the fault feature enhancement result; In step (5), the structural parameters of the bistable stochastic resonance system are set. , And using permutations and combinations By inputting it into a bistable stochastic resonance system, we obtain... A set of bistable stochastic resonance systems with different structural parameters were obtained, thus yielding... Each fault characteristic value.

2. The underwater thruster fault feature enhancement method according to claim 1, characterized in that, In step (5), the ant colony algorithm is used to adjust the structural parameters to optimize the bistable stochastic resonance system. The range of values ​​for the structural parameter a of the bistable stochastic resonance system is [A]. min A max The range of values ​​for [B] and [b] min B max The grid is divided, and the ants are assigned to grid cells according to the principle of random placement. Each cell corresponds to a set of structural parameter combinations [a, b]. The ants in each column are allowed to find the next cell position according to the principle of randomness. The structural parameter combination [a, b] corresponding to this cell is input into the bistable stochastic resonance system.

3. The underwater thruster fault feature enhancement method according to claim 2, characterized in that, In step (6), the pheromone concentration of the ant's cell is updated. It is determined whether the cells selected by the ants in each column converge to the same cell. If they converge, the cell with the highest pheromone concentration in each column is selected; otherwise, the ants in each column are allowed to search for the next cell position randomly. The range of the structural parameter value corresponding to the cell with the highest pheromone concentration is narrowed, and then optimization is performed until the evaluation function value converges or the maximum number of iterations is reached, thus obtaining the optimal structural parameter a. h With b h At this point, the fault feature value is at its maximum, and the corresponding fault feature sequence is used as the result of fault feature enhancement.

4. The underwater thruster fault feature enhancement method according to claim 1, characterized in that, In step (5), a genetic algorithm is used to adjust the structural parameters of the bistable stochastic resonance system. The combination of structural parameters [a, b] of the bistable stochastic resonance system is taken as an individual, and the optimization range of structural parameter a is [A...b]. min A max The optimization range for structural parameter b is [B]. min B max N initial individuals are randomly generated. The parent individuals are selected using the roulette wheel selection method. The fitness of the parent population is calculated. The decoded a and b values ​​of the parent individuals are input into the bistable stochastic resonance system.

5. The underwater thruster fault feature enhancement method according to claim 4, characterized in that, The roulette wheel selection method is used to select parent individuals. The selected parent individuals undergo crossover and mutation operations to obtain offspring individuals. Population evolution seeks optimization, with offspring individuals serving as parents for the next generation in an iterative process until the fitness function converges or the maximum number of iterations is reached, thus obtaining the optimal structure parameter 'a'. h With b h At this point, the fault feature value is at its maximum, and the corresponding fault feature sequence is used as the result of fault feature enhancement.

6. The underwater thruster fault feature enhancement method according to any one of claims 1 to 5, characterized in that, In step (2), the dynamic signal is decomposed into multi-resolution wavelet components to obtain wavelet approximate components, which are then input into the bistable stochastic resonance system.

7. The underwater thruster fault feature enhancement method according to claim 1, characterized in that, The Langevin equation in step (2) is: in, Let a and b be the output signal of the bistable stochastic resonance system, and a and b be the structural parameters of the system. is the input signal of the bistable stochastic resonance system, and n is the length of the underwater robot's dynamic signal data.

8. The underwater thruster fault feature enhancement method according to claim 1, characterized in that, In step (3), the Runge-Kutta method uses the following formula: Where h is the step size; k is a positive integer. .

9. An enhancement system for the underwater thruster fault characteristic enhancement method according to any one of claims 1 to 8, characterized in that, It includes a data acquisition module for conducting underwater robot thruster failure tests and acquiring dynamic signals from the underwater robot; Signal processing module: This module is used to input dynamic signals into a bistable stochastic resonance system to obtain the Langevin equation; solve the Langevin equation using the fourth-order Runge-Kutta method to obtain the output signal of the stochastic resonance system; process the output signal of the stochastic resonance system based on the modified Bayesian algorithm to obtain a fault feature sequence, and take the maximum value in the fault feature sequence as the fault feature value; The optimization module is used to optimize the structural parameters of a bistable stochastic resonance system. The optimized structural parameters are then processed multiple times by the signal processing module. The fault enhancement module is used to take the fault feature sequence corresponding to the maximum value among multiple fault feature values ​​as the fault feature enhancement result.

Images