Hydraulic device fault diagnosis method

Abstract

A hydraulic device fault diagnosis method relates to the technical field of hydraulic device state diagnosis and comprises the following steps: step S01, acquiring sensor data of a hydraulic device; step S02, carrying out dimensionality reduction on the collected data of the hydraulic device sensor through a Barnes-Hut t-SNE algorithm; step S03, improving an artificial bee colony algorithm ABC; and step S04, using SCQ-ABC to train a weight w of an input layer, a hidden layer deviation b and the like of the ELM model. According to the fault diagnosis method, dimension reduction processing is performed on the data through Barnes-Hut t-SNE, and the fault diagnosis method of the extreme learning machine ELM is optimized by using the variant artificial bee colony algorithm SCQ-ABC, so that the diagnosis precision is improved.

Description

Technical Field

[0001] The invention relates to the technical field of hydraulic device state diagnosis, and in particular to a hydraulic device fault diagnosis method. Background Art

[0002] Hydraulic transmission systems are widely used in military armored vehicles due to their unique advantages, namely, wide process adaptability, excellent control performance and low cost. Fault diagnosis technology of hydraulic transmission systems mainly includes the following methods: 1) Based on statistical analysis, this type of method identifies faults by analyzing the statistical characteristics of data, such as principal component analysis (PCA) and independent component analysis (ICA), but the data obeys Gaussian distribution and may not be able to describe the low-dimensional data distribution well. 2) Using machine learning methods, such as support vector machine (SVM), there is no better solution because the choice of kernel function has a great influence on the performance of the model, and it is difficult to implement large-scale training samples; extreme learning machine (ELM) because the weights of the input layer and the bias of the hidden layer are randomly generated by orthogonal matrices and no reverse adjustment is performed, there are problems such as overfitting, parameter number sensitivity, and performance fluctuations caused by randomness. With the significant development of hydraulic technology in achieving high pressure, high speed, high power, high efficiency, low noise, durability, high integration and other requirements, higher requirements are put forward for improving proportional control, servo control, digital control, etc. Therefore, using appropriate fault diagnosis methods can effectively detect the operation of hydraulic devices, ensure the safe and reliable operation of equipment, and reduce economic losses. Summary of the invention

[0003] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a hydraulic device fault diagnosis method based on Barnes-Hut t-SNE-SCQ-SSA-ELM, which performs dimensionality reduction processing on data through Barnes-Hut t-SNE, and uses a variant artificial bee colony algorithm SCQ-ABC to optimize the fault diagnosis method of the extreme learning machine ELM to improve the diagnosis accuracy.

[0004] In order to achieve the above object, the main technical solutions adopted by the present invention include: A method for diagnosing a hydraulic device fault comprises the following steps: Step S01, obtaining sensor data of the hydraulic device and dividing it into a test set and a training set; Step S02, reducing the dimension of the collected hydraulic device sensor data by using the Barnes-Hut t-SNE algorithm; Step S03, improving the artificial bee colony algorithm ABC, including using the chaotic sequence Chaotic to initialize the food source position in the ABC algorithm, using the sine-cosine mechanism Sine-Cosine to adjust the position of the individual during the search process, and introducing quantum superposition and revolving door mechanism Quantum in the update phase, to obtain the improved artificial bee colony algorithm, namely SCQ-ABC; Step S04: Use SCQ-ABC to train the weights of the input layer of the extreme learning machine ELM model w and hidden layer bias b , establish a hydraulic device condition assessment model based on SCQ-ABC-ELM; Step S05, classify the samples through k-fold cross validation, calculate the accuracy rate according to the following formula as the fitness value of the artificial bee colony position, retain the best fitness and its corresponding artificial bee colony position, and its worker bee colony moves to the optimal source food in the next iteration; ; Where: P i For the k The calculated accuracy rate; P cv for k The average accuracy of the k-fold cross validation, K represents the number of folds when performing k-fold cross validation; Step S06, calculating the fitness of the updated colony, comparing the fitness value of each worker bee individual after the update with the fitness value of the current food, if the fitness value of the updated artificial bee colony is better than the food, the position of the worker bee with the better fitness value is used as the new food position; Step S07: until the fitness value reaches the termination threshold, terminate and output the global optimal position, that is, the optimal parameter combination corresponding to the ELM model ; Otherwise, return to iterative optimization; Step S08: Combining the optimal parameters Bring in the ELM model to conduct hydraulic device fault assessment tests.

[0005] Furthermore, the step S03 specifically includes the following steps: Step S301: Chaos initialization food source The formula for generating the initial food source position is generated by using Logistic mapping. Assume that the food source position of the optimization problem is The initialization formula for the food source position is: ; in: x i It is i The initial value of the food source location, and They are i The lower and upper bounds of the variables; is the chaotic sequence value generated by Logistic mapping, that is, using Logistic mapping to generate a sequence , each value is between [0,1], map these values ​​to the solution space; Step S302: When searching in the neighborhood, the worker bee generates a new solution by combining the sine and cosine mechanisms. The update formula is as follows: ; in, is the current optimal solution, is a randomly generated angle. The current solution i The value of the parameter, Indicates i The new position of the worker bees after applying the sine-cosine mechanism, S and C are the coefficients of the sine and cosine functions respectively; Step S303: The solution of the swarm is represented as a quantum bit state, and each dimension of the observation swarm is named as a quantum bit. , and has two basic states: and , the state of the quantum bit is regarded as a linear combination of two basic states, namely the superposition state, as shown below: ; in α for β is the probability amplitude, where the probability amplitude is two random complex numbers, one of which is α , another random complex number is β，α and β The relationship satisfies the following formula: ; set up , ,in is the phase, then the state of the quantum bit can be expressed as a probability amplitude or phase as follows: ; one m Weiyou m The state A of the observation bee is expressed as the probability amplitude or phase as follows: ; The probability of observing a bee state in quantum computing is 2 m,During the search process, the observation bee uses the quantum mutation mechanism to perform local perturbations, uses the quantum revolving gate to update the position of the solution, and realizes the position update of the quantum observation bee colony according to the quantum revolving gate of the following formula to jump out of the local optimum; ; In the above formula, For the i The observation bee t In the iteration j The phase of the dimension; Indicates i Observe bees in t +1 iteration j dimensional phase shift, the mapping of the individual state of the bee from the quantum state to the solution space is as follows: ; in, For the i Observation of bees j Dimensional location; is a random number between 0 and 1, It is i The observation bee j The phase of the dimension, and They are j The upper and lower bounds of the dimension.

[0006] The beneficial effects of the present invention are: 1. Since the linear correlation and nonlinear correlation between the parameter variables of the collected hydraulic device information will directly affect the accuracy of the subsequent hydraulic device status assessment, the present invention uses the Barnes-Hut t-SNE algorithm to reduce the dimension of the data for subsequent data set production, and then normalizes the data, which can process large-scale data sets more efficiently; 2. The present invention uses a chaotic sequence to initialize the food source position in the ABC algorithm, and generates solutions through chaotic mapping, thereby avoiding the problem of uneven solution distribution that may be caused by traditional random generation, thereby enhancing the global search capability. During the search process, a sine-cosine mechanism is used to adjust the position of individuals. During the search process, the position of the solution is dynamically adjusted, so that a large-scale search can be performed in the early stage, and gradually focus on better solutions in the later stage, thereby accelerating convergence. In the update stage, quantum superposition and revolving door mechanisms are introduced to generate new solutions with high randomness, and the global search capability is further enhanced through quantum characteristics, which effectively helps the algorithm to jump out of the local optimum. Through the synergistic effect of chaotic initialization, sine-cosine mechanism and quantum computing, the improved ABC algorithm more efficiently balances global search and local development, improves convergence speed and optimization accuracy; 3. The improved SCQ-ABC is used to optimize the ELM related parameters, which makes up for the disadvantage that the weights of the input layer and the bias of the hidden layer are randomly generated by the orthogonal matrix and are not adjusted in reverse during the ELM training process. Taking advantage of the high search accuracy and fast convergence speed of SCQ-ABC, the weights w of the ELM input layer and the bias b of the hidden layer are optimized, thereby avoiding the problems of local optimality, slow convergence speed and low search accuracy in the model during the optimization process, and achieving better classification results; 4. The data processed by Barnes-Hut t-SNE have higher diagnostic accuracy and practicality after being evaluated by the SCQ-ABC optimized ELM state assessment model. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is a schematic diagram of the flow chart of the hydraulic device fault diagnosis method of the present invention; Figure 2 It is a fitness curve diagram comparing the four methods of SCQ-ABC-ELM, ELM, COA-ELM and PSO-ELM in hydraulic fault diagnosis in the specific implementation method. DETAILED DESCRIPTION

[0008] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0009] The present invention provides a method for diagnosing faults of a hydraulic device, such as Figure 1 As shown, the following steps are included: Step S01: obtain the hydraulic device sensor data and divide it into a test set and a training set.

[0010] The sensor data of the hydraulic unit is obtained experimentally on a hydraulic test bench, which consists of a primary working circuit and a secondary cooling and filtering circuit connected via an oil tank. The system cyclically repeats a constant load cycle (lasting 60 seconds) and measures process values ​​such as pressure, volume flow and temperature, while the states of the four hydraulic components (cooler, valve, pump and accumulator) are quantitatively changed.

[0011] The dataset contains raw process sensor data, i.e., no feature extraction. These data are constructed as a matrix (tab-delimited). Table 1 is the data attribute information table, and Table 2 is the fault information table.

[0012] Table 1: Data attribute information table .

[0013] Table 2: Fault information table .

[0014] Step S02: reducing the dimension of the collected hydraulic device sensor data by using the Barnes-Hut t-SNE algorithm.

[0015] In high-dimensional space, suppose the high-dimensional space contains n data samples, represented as Data points in high-dimensional space X i , X j Similar conditional probability It can be expressed as the following formula: ; In low-dimensional space, by finding the [3 u ] neighbors to calculate the sparse approximation and redefine it as the corresponding similarity p ij Is the following formula: ; Among them, N i Represents X i [3 u ] the set of nearest neighbors, The setting makes the perplexity of the conditional distribution equal to u .

[0016] By building an index tree on the data set, the nearest neighbor set can be found in O(uNlogN) time. Specifically: 1. Build a spatial index tree: In the index tree, each node stores a data object and the radius of a sphere centered on the object. All non-leaf nodes have two child nodes: data objects located inside the sphere are stored under the left child of the node, while data objects located outside the sphere are stored under the right child. The tree is constructed by presenting data objects in an arbitrary order, traversing the tree based on whether the current data object is inside or outside the sphere, and creating a new leaf node to store the new object. The radius of the new leaf node is set to the middle distance between its object and all other objects within the sphere represented by its parent node, after which a nearest neighbor search is performed by performing a depth-first search on the tree.

[0017] 2. Used in t-SNE KL Divergence represents the probability distribution in low-dimensional space Q and low-dimensional space probability distribution P The similarity between them is: ; Among them, D KL stands for Kullback-Leibler divergence, which is a measure of the difference between two probability distributions. C express KLDivergence cost function.

[0018] In Barnes-Hut t-SNE, to approximate the t-SNE gradient, the gradient is first decomposed as follows: ; in , the cost function C for the data point The partial derivative of , i.e., the gradient, is used to guide the movement of data points in the low-dimensional space. F atter represents the sum of all gravitational forces, that is, the data point attraction to all other data points; F rep represents the sum of all repulsive forces, that is, the data point The repulsive force between all other data points; The similarity between data points i and j in high-dimensional space; The similarity between data points i and j in low-dimensional space; A function representing the low-dimensional distance between data points i and j, calculating the sum of all gravitational forces F atter , by sparsely distributing P in The t-SNE algorithm performs all repulsive F rep The simple calculation of the sum is O(N 2 ). Note , which is efficiently approximated in O(NlogN) using the Barnes-Hut algorithm F rep With this approximation, only the actual points that are close to the target point and the influence of the centroid that is far away need to be calculated, greatly reducing the amount of calculations that must be performed.

[0019] Step S03, improving the artificial bee colony algorithm ABC, including using the chaotic sequence Chaotic to initialize the food source position in the ABC algorithm, using the sine-cosine mechanism Sine-Cosine to adjust the position of the individual during the search process, and introducing quantum superposition and revolving door mechanism Quantum in the update stage, to obtain the improved artificial bee colony algorithm, namely SCQ-ABC.

[0020] The artificial bee colony algorithm ABC (Artificial Bee Colony) involves food sources, worker bees, observer bees, and scout bees, simulating the roles and behaviors of bees in the foraging process in nature. Worker bees are mainly responsible for preliminary exploration and updating of food sources, and each worker bee has a corresponding food source. Observer bees make selections based on the quality of the food sources found by worker bees, and further optimize high-quality solutions. Scout bees are responsible for randomly searching for new solutions to replace those food sources that are no longer improved, thereby ensuring global exploration. The food source is a candidate solution in the solution space. After continuous optimization by worker bees and observer bees, the algorithm gradually finds the optimal solution. The traditional ABC algorithm also has some shortcomings, such as weak local search capabilities, easy to fall into early convergence, slow convergence speed, etc. The present invention adopts three strategies: chaos theory, sine-cosine mechanism, and quantum computing principle to strengthen the basic ABC. That is, SCQ-ABC.

[0021] The step S03 specifically includes the following steps: Step S301: Chaos initialization food source The formula for generating the initial food source position is generated by using Logistic mapping. Assume that the food source position of the optimization problem is The initialization formula for the food source position is: ; in: x i It is i The initial value of the food source location, and They are i The lower and upper bounds of the variables; It is a chaotic sequence value generated by Logistic mapping. Use Logistic mapping to generate a sequence , each value is between [0,1], and these values ​​are mapped to the solution space. The present invention uses chaotic sequences to initialize the food source positions in the ABC algorithm. The ergodicity of chaotic sequences makes the initial solution distribution more uniform, thereby enhancing the ability of global search.

[0022] Step S302: When searching in the neighborhood, the worker bee generates a new solution by combining the sine and cosine mechanisms. The update formula is as follows: ; in, is the current optimal solution, is a randomly generated angle. is the value of the i-th parameter of the current solution, represents the new position of the ith worker bee after applying the sine-cosine mechanism (SCM), S and Care the coefficients of the sine and cosine functions, respectively. The positions of individuals are adjusted by the sine and cosine functions, which has good global and local search capabilities. The sine-cosine mechanism has a smooth search trajectory in optimization and can flexibly jump in a wide solution space. Through the sine-cosine function, the worker bees can dynamically adjust the search amplitude and direction according to the complexity of the problem. The sine-cosine mechanism can have greater fluctuations in the early stage and is highly exploratory, and gradually converges to a better solution in the later stage, which is highly exploitative.

[0023] Step S303: The solution of the swarm is represented as a quantum bit state, and each dimension of the observation swarm is named as a quantum bit. , and has two basic states: and , the state of the quantum bit is regarded as a linear combination of two basic states, namely the superposition state, as shown below: ; in α and β is the probability amplitude, where the probability amplitude is two random complex numbers, one of which is α , another random complex number is β，α and β The relationship satisfies the following formula: ; set up , ,in is the phase, then the state of the quantum bit can be expressed as a probability amplitude or phase as follows: ; one m Weiyou m The state A of the observation bee is expressed as the probability amplitude or phase as follows: ; The probability of observing a bee state in quantum computing is 2 m ,During the search process, the observation bee uses the quantum mutation mechanism to perform local perturbations, uses the quantum revolving gate to update the position of the solution, and realizes the position update of the quantum observation bee colony according to the quantum revolving gate of the following formula to jump out of the local optimum; ; In the above formula, For the i The observation bee t In the iteration j The phase of the dimension; Indicates i Observe bees in t +1 iteration jdimensional phase shift, the mapping of the individual state of the bee from the quantum state to the solution space is as follows: ; in, For the i Observation of bees j Dimensional location; is a random number between 0 and 1, is the phase of the ith observer bee in the jth dimension, and are the upper and lower bounds of the j-th dimension respectively.

[0024] The invention introduces quantum computing, which can greatly improve the local search accuracy and global search capability of the observer bee. Quantum computing allows multiple solutions to be processed at the same time, thus greatly improving the parallel computing capability.

[0025] The present invention generates an initial population through chaotic mapping to ensure the diversity of solutions and avoid the search from falling into a narrow area. Then, the sine-cosine mechanism is integrated into the search process to dynamically adjust the position of individual solutions to enhance local search capabilities and balance global exploration. Finally, the quantum superposition and revolving door mechanism are introduced in the update phase to further improve the optimization efficiency and global search capabilities by utilizing the high randomness and globality of the quantum state. This combination can effectively improve the performance of the algorithm, avoid local optimality and accelerate convergence.

[0026] Step S04: Use SCQ-ABC to train the weights of the input layer of the extreme learning machine ELM model w and hidden layer bias b , a hydraulic device condition assessment model based on SCQ-ABC-ELM was established.

[0027] Under the SCQ-ABC algorithm framework, the weights of the input layer of the ELM model are adjusted by the foraging behaviors of worker bees, observer bees, and scout bees. w and hidden layer bias b . Worker bees are used to search in the parameter space of ELM, and find better model parameter configurations by randomly perturbing the input layer weights w and hidden layer biases b. During the ELM model training process, observer bees can simulate the selection process and adjust the model weights according to the optimal parameter configuration found by the worker bees. When the ELM model does not improve significantly within a certain number of iterations, scout bees can simulate the restart process and randomly initialize new parameter configurations to jump out of the local optimal solution.

[0028] Step S05, classify the samples through k-fold cross validation, calculate the accuracy rate according to the following formula as the fitness value of the artificial bee colony position, retain the best fitness and its corresponding artificial bee colony position, and its worker bee colony moves to the optimal source food in the next iteration; ; Where: P i For the k The calculated accuracy rate; P cv for k The average accuracy of the k-fold cross validation, K represents the number of folds in the k-fold cross validation. The entire dataset is divided into K disjoint subsets, each of which is used as a test set in one experiment, and the remaining K-1 subsets are used as training sets. This process is repeated K times, and a different subset is selected as the test set each time.

[0029] Step S06, calculating the fitness of the updated colony, comparing the fitness value of each worker bee individual after the update with the fitness value of the current food, if the fitness value of the updated artificial bee colony is better than the food, the position of the worker bee with the better fitness value is used as the new food position; Step S07: until the fitness value reaches the termination threshold, terminate and output the global optimal position, that is, the optimal parameter combination corresponding to the ELM model ; Otherwise, return to iterative optimization; Step S08: Combining the optimal parameters Bring in the ELM model to conduct hydraulic device fault assessment tests.

[0030] The present invention solves the crowding problem in Gaussian distribution in low-dimensional space by using Barnes-Hut t-SNE to reduce the dimension of large-scale data, and reduces the O(N 2 ) is reduced to O(NlogN), making the algorithm more efficient. At the same time, the ELM algorithm with strong generalization ability is selected as the basic algorithm for classification. Due to the disadvantage that the weights of the input layer and the bias of the hidden layer are randomly generated by the orthogonal matrix during the ELM training process and no reverse adjustment is performed, the invention proposes to use the Barnes-Hutt-SNE algorithm to process the original data and then use the variant artificial bee colony (SCQ-ABC) optimization algorithm to optimize the extreme learning machine (ELM), which has higher diagnostic accuracy and practicality.

[0031] Using the fitness function error rate (error rate = 1-accuracy) as the performance indicator, the convergence speed of the four methods SCQ-ABC-ELM, ELM, COA-ELM, and PSO-ELM in hydraulic fault diagnosis is compared. The results are as follows Figure 2As shown in the figure, the ordinate is the fitness value, i.e., the error rate, and the abscissa is the evolutionary generation, i.e., the number of iterations; the downward trend of the curve shows that as the number of iterations increases, the prediction error of the model is decreasing, and the prediction performance of the model is improving. It can be seen that SCQ-ABC-ELM converges the fastest and basically completes convergence in the 20th generation. COA-ELM converges faster and also basically completes convergence in the 20th generation, but the convergence effect is worse than that of SCQ-ABC-ELM. PSO-ELM gradually shows a convergence trend in the 30th generation and completes convergence in the 40th generation. ELM converges more slowly and has a longer number of iterations, and finally completes convergence in the 40th generation. Comparing the above algorithms, it can be concluded that the SCQ-ABC-ELM algorithm has the fastest convergence speed and the best effect.

[0032] Calculate the accuracy, precision, and recall of each model test set respectively: Accuracy: ;

[0033] TP: True positive examples, the number of samples correctly predicted by the model as positive.

[0034] TN: True negative examples, the number of samples correctly predicted by the model as negative.

[0035] FP: False Positives, the number of negative samples that the model incorrectly predicts as positive.

[0036] FN: False negative examples, the number of positive samples that the model incorrectly predicts as negative.

[0037] Recall: ; Precision: ; Comparing the fault diagnosis results of the four methods of SCQ-ABC-ELM, ELM, COA-ELM, and PSO-ELM, after taking the average value of multiple experiments, as shown in Table 3, it can be found that the ELM model has significantly weaker data processing capabilities. The evaluation indicators of the SCQ-ABC-ELM combination model have all increased, and the accuracy rate has reached 98.75%, proving the superiority of the algorithm combination.

[0038] Table 3: Comparison of fault diagnosis results of four methods: SCQ-ABC-ELM, ELM, COA-ELM, and PSO-ELM .

[0039] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. Alterations, modifications, substitutions and variations of the above embodiments by a person skilled in the art are all within the scope of the present invention.

Claims

1. A method for diagnosing a hydraulic device fault, characterized in that: The steps include: Step S01, obtaining sensor data of the hydraulic device and dividing it into a test set and a training set; Step S02, reducing the dimension of the collected hydraulic device sensor data by using the Barnes-Hut t-SNE algorithm; Step S03, improving the artificial bee colony algorithm ABC, including using the chaotic sequence Chaotic to initialize the food source position in the ABC algorithm, using the sine-cosine mechanism Sine-Cosine to adjust the position of the individual during the search process, and introducing quantum superposition and revolving door mechanism Quantum in the update phase, to obtain the improved artificial bee colony algorithm, namely SCQ-ABC; Step S04: Use SCQ-ABC to train the weights of the input layer of the extreme learning machine ELM model w and hidden layer bias b , establish a hydraulic device condition assessment model based on SCQ-ABC-ELM; Step S05, classify the samples through k-fold cross validation, calculate the accuracy rate according to the following formula as the fitness value of the artificial bee colony position, retain the best fitness and its corresponding artificial bee colony position, and its worker bee colony moves to the optimal source food in the next iteration; ； Where: P i For the k The calculated accuracy rate; P cv for k The average accuracy of the k-fold cross validation, K represents the number of folds when performing k-fold cross validation; Step S06, calculating the fitness of the updated colony, comparing the fitness value of each worker bee individual after the update with the fitness value of the current food, if the fitness value of the updated artificial bee colony is better than the food, the position of the worker bee with the better fitness value is used as the new food position; Step S07: until the fitness value reaches the termination threshold, terminate and output the global optimal position, that is, the optimal parameter combination corresponding to the ELM model ; Otherwise, return to iterative optimization; Step S08: Combining the optimal parameters Bring in the ELM model to conduct hydraulic device fault assessment tests.

2. A hydraulic device fault diagnosis method according to claim 1, characterized in that: The step S03 specifically includes the following steps: Step S301: Chaos initialization food source The formula for generating the initial food source position is generated by using Logistic mapping. Assume that the food source position of the optimization problem is The initialization formula for the food source position is: ； in: x i It is i The initial value of the food source location, and They are i The lower and upper bounds of the variables; is the chaotic sequence value generated by Logistic mapping, that is, using Logistic mapping to generate a sequence , each value is between [0,1], map these values ​​to the solution space; Step S302: When searching in the neighborhood, the worker bee generates a new solution by combining the sine and cosine mechanisms. The update formula is as follows: ； in, is the current optimal solution, is a randomly generated angle. The current solution i The value of the parameter, Indicates i The new position of the worker bees after applying the sine-cosine mechanism, S and C are the coefficients of the sine and cosine functions respectively; Step S303: The solution of the swarm is represented as a quantum bit state, and each dimension of the observation swarm is named as a quantum bit. , and has two basic states: and , the state of the quantum bit is regarded as a linear combination of two basic states, namely the superposition state, as shown below: ； in α for β is the probability amplitude, where the probability amplitude is two random complex numbers, one of which is α , and another random complex number is β，α and β The relationship satisfies the following formula: ； set up , ,in is the phase, then the state of the quantum bit can be expressed as a probability amplitude or phase as follows: ； one m Weiyou m The state A of the observation bee is expressed as the probability amplitude or phase as follows: ； The probability of observing a bee state in quantum computing is 2 m During the search process, the observation bees use the quantum mutation mechanism to perform local perturbations, use the quantum revolving gate to update the position of the solution, and use the quantum revolving gate of the following formula to realize the position update of the observation bee colony and jump out of the local optimum; ； In the above formula, For the i The observation bee t In the iteration j The phase of the dimension; Indicates i Observe bees in t +1 iteration j dimensional phase shift, the mapping of the individual state of the bee from the quantum state to the solution space is as follows: ； in, For the i The first bee observation j Dimensional location; is a random number between 0 and 1, It is i The observation bee j The phase of dimension, and They are j The upper and lower bounds of the dimension.

Images