A Blade Sorting Method Based on Threshold-Type Local Iterative Search Algorithm

Abstract

A blade sorting method based on a threshold-based local iterative search algorithm, which relates to a blade sorting method. The present invention aims to solve the problem that the existing local iterative search algorithm has a large number of average iterations to reach the local optimum, resulting in low search efficiency for blade sorting. Step 1 of the present invention: Number the blades; Step 2: Initialize the sequence H in Step 1; Step 3: Conduct local search on the sequence H in Step 2; Step 4: Update the global optimal solution H global ; Step 5: Determine whether the number of elements in the set Σ 1 is greater than the step size step: If so, jump to Step 6, otherwise jump to Step 7; Step 6: Threshold perturbation; Step 7: Use all the elements in the descending trend set Σ perturbation as the initial sequence for Step 3 in turn, and repeat Steps 3 and 4; Step 8: Random perturbation. The present invention introduces the concept of threshold perturbation and sorts the blades within a reasonable calculation time. The present invention is used for blade sorting.

Description

Technical Field

[0001] The present invention relates to a method for sorting steam turbine blades, and more particularly to a method for sorting blades based on a threshold-based local iterative search algorithm, belonging to the technical field of mechanical assembly. Background Art

[0002] The dynamic balance of a steam turbine rotor has a great influence on its performance and service life. However, in the actual manufacturing process of steam turbine blades, it is very difficult to make the mass and the center of gravity position of each blade the same due to machining errors. The product of the mass-radius is the product of the vector radius from the center of gravity of the blade to the rotating shaft and the mass. The difference in the product of the mass-radius will cause a certain initial unbalance of the rotor. A reasonable blade arrangement order can reduce the initial unbalance and minimize the resultant force acting on the shaft.

[0003] The problem of blade sorting is a combinatorial optimization problem. However, the number of blades is usually very large, generally ranging from dozens to hundreds. The solution space of all permutations is of the order of n!, so it is an NP-hard problem and only approximate optimal solutions can be obtained. The blade sorting methods can be divided into traditional methods and heuristic methods. The traditional methods usually use the grouping optimization method, which divides the blades into several groups, optimizes the total mass of each group by exhaustive means, and then optimizes the order between the groups. However, this method has poor robustness and cannot guarantee good grouping every time for different data.

[0004] The heuristic methods select the optimal value by searching the solution space. Commonly used heuristic methods include genetic algorithms, ant colony algorithms, discrete particle swarm algorithms, local search algorithms, etc. Wu Genhong used a genetic algorithm to encode the blade numbers and finally selected the optimal solution after a series of crossover, selection, and mutation operations. Li Dandan et al. used an ant colony algorithm to perform random search based on the principle of the foraging behavior of ants. Multiple ants randomly search for paths and converge to the optimal path according to the magnitude of the resultant force of the sequences represented by each path. Li Yan et al. used a discrete particle swarm method to update multiple particles through local optimal solutions and global optimal solutions, which has the characteristics of simple algorithm, easy implementation, and fast convergence speed. However, the global search algorithm searches the entire space, and when the state space is large, the efficiency is relatively low.

[0005] The local search algorithm defines the concept of neighborhood in the solution space. Then, starting from an initial solution, it selects a solution in a certain neighborhood according to the iterative strategy for iteration. The local search algorithm will fall into the local optimum. Many algorithms have specified different strategies for jumping out of the local optimum. Brito, J, etc. proposed an iterative local search algorithm. When falling into the local optimum each time, it randomly selects one from the current neighborhood for transfer. By performing this step several times, the purpose of perturbation is achieved. Wang, Y, etc. proposed a greedy random adaptive search algorithm. When falling into the local optimum each time, it reselects an initial solution in a randomly constructed manner and uses the restart mechanism to jump out of the local optimum. Wayne Pullan, etc. proposed a dynamic local search algorithm. This method affects the value of the evaluation function by adding a penalty function to make it jump out of the local optimum.

[0006] Among them, the local iterative search algorithm requires more iteration times when iterating to the local minimum each time. The reason for this phenomenon is that when the local iterative search algorithm reaches the local optimum each time, it jumps out of the local optimum through random neighborhood perturbation. Due to its randomness, the size of the perturbation cannot be controlled. Therefore, the average number of steps to iterate to the local optimum is large, and there is a problem of low search efficiency for blade sorting. Summary of the Invention

[0007] The purpose of the present invention is to solve the problem that the existing local iterative search algorithm has a large average number of steps to iterate to the local optimum and has a low search efficiency for blade sorting. Furthermore, a blade sorting method based on a threshold-based local iterative search algorithm is provided.

[0008] The technical solution of the present invention is: a blade sorting method based on a threshold-based local iterative search algorithm, which includes the following steps:

[0009] Step 1: Number the blades:

[0010] Step 1.1: For n blades evenly distributed on the steam turbine shaft, starting from any one blade, in the counterclockwise direction, number the blades as 1, 2... n respectively;

[0011] Step 1.2: Define any combination of the n numbers as a sequence H of the product of the mass radius of the blades;

[0012] Step 1.3: Form a solution space set with all the sequences H, denoted as Σ;

[0013] Step 2: Initialize the sequence H in Step 1.2:

[0014] Randomly initialize a sequence H. Define the neighborhood transformation of the local search algorithm as any two elements in the sequence H, denoted as τ(H). Define the neighborhood of the sequence H as the set of all its neighborhood transformations, denoted as Σ H, set the threshold of the threshold-based local iterative search algorithm to C th , the step size is step, and the bias probability is bia;

[0015] Step 3: Perform local search on the sequence H in Step 2:

[0016] Using the sequence H as the initial sequence, perform local search iterative descent and return the locally optimal solution sequence H obtained by local search local , and all objective function values less than the threshold C during the descent process th of the set Σ of sequences 1 ;

[0017] Step 4: Update the global optimal solution H global ;

[0018] Step 5: Judge whether the number of elements in the set Σ 1 is greater than the step size step: if so, jump to Step 6, otherwise jump to Step 7;

[0019] Step 6: Threshold perturbation:

[0020] Take the first element value of the set Σ 1 and record it as Take out the element value All sequences with a downward trend in all neighborhood transformations are recorded as the downward trend set Σ perturbation ;

[0021] Step 7: Use all elements in the downward trend set Σ perturbation as the initial sequence H in Step 3 and repeat Step 3 and Step 4;

[0022] Step 8: Random perturbation:

[0023] Perform a random neighborhood transfer on the optimal solution sequence H with a bias probability of 1 - bia local as the initial sequence H, randomly initialize an initial sequence H with a bias probability of bia, if the maximum execution time is reached, the algorithm ends, otherwise continue to execute Step 2. Thus, the search for the optimal sorting of the blades is completed.

[0024] Furthermore, the definition process of the sequence H of the product of the mass radius of the blades in Steps 1 and 2 is as follows:

[0025] S1: Number the sequence H:

[0026] Given an initial sequence M of the product of the mass radius of the blades = (m1r1, m2r2,......, m n r n ), the sequence H is the sequence composed of the corresponding subscripts after reordering the sequence M of the product of the mass radius.

[0027] S2: Calculate the objective function:

[0028] For a sequence H, first convert it into the corresponding product of mass-radius sequence M. When n blades are evenly distributed in a circle, the included angle between adjacent blades is:

[0029] Define the vector corresponding to the i-th element in the product of mass-radius sequence M as:

[0030] Define the objective function value as:

[0031]

[0032] Furthermore, in Steps 1 and 2, the product of mass-radius of the blade is the product of the radius vector from the center of gravity of the blade to the axis of rotation and the mass of the blade; for example: if the product of mass-radius sequence corresponding to sequence H is M, and they are evenly arranged on the axis, the calculation method of the total product of mass-radius is S2.

[0033] Further, the specific implementation method in Step 3 is as follows:

[0034] Step 3-1: Initialize a sequence H, and obtain the neighborhood Σ of sequence H through the neighborhood transformation τ(H) H ;

[0035] Step 3-2: Randomly select a sequence from the neighborhood Σ of sequence H H and denote it as H new , satisfying H new ∈Σ H , calculate f(H new ), and remove it from the neighborhood Σ of sequence H H ;

[0036] Step 3-3: If f(H new ) < f(H), then H = H new ; otherwise execute Step 3-5;

[0037] Step 3-4: If f(H new ) < C th , add H new to the set Σ 1 , and execute Step 3-2;

[0038] Step 3-5: If the neighborhood Σ of sequence H H is not an empty set, execute Step 3-2, otherwise H local = H, return the local optimal solution H local and the set Σ 1 and then end.

[0039] Further, the process of updating the global optimal solution in Step 4 is as follows;

[0040] If f(H global ) < f(H local ), update H global = H local .

[0041] Furthermore, the specific implementation method of step six is as follows:

[0042] Step six - one: Take the first element in the set Σ 1 and calculate the neighborhood of the element through the neighborhood transformation . of the element

[0043] Step six - two: Randomly select a sequence from the neighborhood of the element and denote it as H new . Calculate f(H new ) and remove it from the neighborhood of the element .

[0044] Step six - three: If , add H new to the set Σ perturbation . This set is the candidate set of the threshold perturbation sequence for providing the threshold perturbation sequence;

[0045] Step six - four: If the neighborhood of the element is not an empty set, execute step six - two; otherwise, return the set Σ perturbation and end.

[0046] Furthermore, the specific implementation method of step eight is as follows:

[0047] Step eight - one: Execute step eight - two and step eight - three with probabilities of bias probability bia and 1 - bia respectively;

[0048] Step eight - two: Calculate the neighborhood local of H local through the neighborhood transformation τ(H and randomly sample a sequence H in and then return;

[0049] Step eight - three: Randomly sample a sequence H from the solution space set Σ and then return.

[0050] The present invention has the following effects compared with the prior art:

[0051] 1. The present invention adds a threshold limit perturbation, which effectively reduces the number of iterations required to reach the local minimum by controlling the perturbation amount. Thus, the search efficiency is effectively improved.

[0052] 2. Compared with the ILS algorithm (i.e., the local iterative search algorithm), the present invention adds a threshold limit perturbation in step six. By restricting the perturbation within the threshold, the number of steps required to iteratively reach the local optimal solution on average is greatly reduced. As shown in Table 2, for the combinatorial optimization problems in Table 1, two groups of times, 84s and 1470s respectively, are selected in Table 2, and each group of experiments is repeated 20 times to calculate the average value. Among them, the average number of local optimal solutions represents the number of local optimal solutions found within this time, the average minimum local optimal solution represents the minimum value among all local optimal solutions, and the average number of iterative steps represents the number of steps required to iteratively reach the optimal solution from the initial solution on average. It can be seen from Table 2 that the average number of iterative steps of the TILS algorithm (i.e., the threshold-based local iterative search algorithm) is greatly reduced. Therefore, more local minima are searched within the same time. The search efficiency is improved by more than 20% compared with the ILS algorithm. Specific Embodiment

[0053] Specific Embodiment 1: A blade sorting method based on the threshold-based local iterative search algorithm in this embodiment includes the following steps:

[0054] Step 1: Number the blades:

[0055] Step 1-1: For the n blades evenly distributed on the steam turbine shaft, starting from any one blade, in the counterclockwise direction, number the blades as 1, 2... n respectively;

[0056] Step 1-2: Define any combination of the n numbers as a sequence H of the product of the mass radius of the blades;

[0057] Step 1-3: Form the solution space set of all sequences H, denoted as Σ;

[0058] Step 2: Initialize the sequence H in Step 1-2:

[0059] Randomly initialize a sequence H, define the neighborhood transformation of the local search algorithm as any two elements in the sequence H, denoted as τ(H), and define the neighborhood of the sequence H as the set of all its neighborhood transformations, denoted as Σ H Set the threshold of the threshold-based local iterative search algorithm as C th Set the step size as step and the bias probability as bia;

[0060] Step 3: Conduct local search on the sequence H in Step 2:

[0061] Using sequence H as the initial sequence, perform local search iterative descent and return the locally optimal solution sequence H obtained from the local search local , and all the objective function values less than the threshold C during the descent process th of the set Σ of sequences 1 ;

[0062] Step Four: Update the global optimal solution H global ;

[0063] Step Five: Determine whether the number of elements in the set Σ 1 is greater than the step size step: If so, jump to Step Six; otherwise, jump to Step Seven;

[0064] Step Six: Threshold perturbation:

[0065] Take the first element value of the set Σ 1 and denote it as Take out the element value All sequences with a downward trend in all neighborhood transformations, denoted as the downward trend set Σ perturbation ;

[0066] Step Seven: Use all the elements in the downward trend set Σ perturbation as the initial sequence H in Step Three, and repeat Step Three and Step Four;

[0067] Step Eight: Random perturbation:

[0068] Perform a random neighborhood transfer on the optimal solution sequence H local with a bias probability of 1 - bia as the initial sequence H, randomly initialize an initial sequence H with a bias probability of bia. If the maximum execution time is reached, the algorithm ends; otherwise, continue to execute Step Two. Thus, the search for the optimal sorting of the blades is completed.

[0069] To further illustrate the technical advantages of the present invention compared with the prior art, the process of the local search algorithm is as follows:

[0070]

[0071] The process of the ILS algorithm with random perturbation is as follows:

[0072]

[0073]

[0074] And the process of the TILS algorithm with threshold perturbation adopted by the present invention is as follows:

[0075]

[0076]

[0077] Among them, Algorithm2 is the ILS algorithm, and Algorithm2-1 is a function in Algorithm2.

[0078] Algorithm3 is the TILS algorithm, and Algorithm3-1 is a function in Algorithm3.

[0079] The TILS algorithm of this embodiment, on the basis of the ILS algorithm, in view of the characteristics of the blade sorting problem, when falling into a local optimum, adopts a method combining threshold-limited perturbation and random perturbation to jump out of the local optimum. This method greatly reduces the average number of steps to reach the local optimum through local iteration after perturbation compared with the iterative local search algorithm, and improves the efficiency of local search.

[0080] In addition, the present invention belongs to a combinatorial optimization method. This method improves the ILS algorithm for blade sorting problems. In view of using random perturbation in ILS to propose local optimum values, the TILS algorithm introduces the concept of threshold perturbation. The difference from the ILS algorithm is that step six threshold perturbation is added. The purpose of the present invention is to reduce the initial mass-radius product imbalance by sorting steam turbine blades within a reasonable calculation time, and the algorithm is applicable to various numbers of blades and different data distributions, and has universality.

[0081] Among them, the selection of the threshold needs to be artificially adjusted according to the size of the data. It is a hyperparameter, and the thresholds for different data are different. The perturbation is to jump out of the local minimum.

[0082] Combinatorial optimization means that the blade sorting problem is a combinatorial optimization problem mathematically, and ILS and TILS are methods to solve such combinatorial optimization problems. The goal of this combinatorial optimization problem is to find a sequence, and the mass-radius product after the force synthesis corresponding to this sequence is the smallest.

[0083] Specific Embodiment 2: The process of defining the sequence H of the mass-radius product of the blades in Steps 1 and 2 of this embodiment is as follows:

[0084] S1: Number the sequence H:

[0085] Given an initial sequence M of the mass-radius product of blades = (m1r1, m2r2,......, m n r n ), the sequence H is the sequence composed of the corresponding subscripts after reordering the mass-radius product sequence M;

[0086] S2: Calculate the objective function:

[0087] For a sequence H, first convert it into the corresponding product of mass and radius sequence M. When n blades are evenly distributed in a circle, the included angle between adjacent blades is:

[0088] Define the vector corresponding to the i-th element in the product of mass and radius sequence M as:

[0089] Define the objective function value as:

[0090]

[0091] Other compositions and connection relationships are the same as those in the first specific implementation manner.

[0092] Specific implementation manner three: In steps one and two of this implementation manner, the product of mass and radius of the blade is the product of the radius vector from the center of gravity of the blade to the axis of rotation and the mass of the blade; for example: The product of mass and radius sequence corresponding to sequence H is M, which is evenly arranged on the axis, and the calculation method of the total product of mass and radius is S2. Other compositions and connection relationships are the same as those in the first or second specific implementation manner.

[0093] Specific implementation manner four: The specific implementation method in step three of this implementation manner is:

[0094] Step three one: Initialize a sequence H, and obtain the neighborhood Σ of sequence H through the neighborhood transformation τ(H) H ;

[0095] Step three two: Randomly select a sequence from the neighborhood Σ of sequence H H and denote it as H new , satisfying H new ∈Σ H , calculate f(H new ), and remove it from the neighborhood Σ of sequence H H ;

[0096] Step three three: If f(H new ) < f(H), then H = H new ; otherwise, execute step three five;

[0097] Step three four: If f(H new ) < C th , add H new to the set Σ 1 , and execute step three two;

[0098] Step three five: If the neighborhood Σ of sequence H H is not an empty set, execute step three two, otherwise H local = H, and return the local optimal solution H local and the set Σ 1It ends later. Other compositions and connection relationships are the same as those in the first, second, or third specific implementation manners.

[0099] Specific implementation manner five: The process of updating the global optimal solution in step four of this implementation manner is as follows;

[0100] If f(H global ) < f(H local ), update H global = H local .

[0101] With such settings, the local search algorithm has a strong purpose and can find the local optimal solution under this definition in a relatively short time. Other compositions and connection relationships are the same as those in the first, second, third, or fourth specific implementation manners.

[0102] Specific implementation manner six: The specific implementation method of step six of this implementation manner is as follows:

[0103] Step six one: Take the first element 1 in the set Σ and calculate the neighborhood of the element through neighborhood transformation

[0104] Step six two: Randomly select a sequence from the neighborhood of the element and denote it as H new , calculate f(H new ), and remove it from the neighborhood of the element ;

[0105] Step six three: If add H new to the set Σ perturbation , and this set is the candidate set of the threshold perturbation sequence for providing the threshold perturbation sequence;

[0106] Step six four: If the neighborhood of the element is not an empty set, execute step six two, otherwise return the set Σ perturbation and end.

[0107] With such settings, the previous search process can be effectively utilized. In addition, imposing a threshold on the perturbation to limit the size of the perturbation is beneficial to reducing the number of times of finding the local optimal solution. Other compositions and connection relationships are the same as those in the first, second, third, fourth, or fifth specific implementation manners.

[0108] Specific implementation manner seven: The specific implementation method of step eight of this implementation manner is as follows:

[0109] Step VIII-1: Execute Step VIII-2 and Step VIII-3 with probabilities bia and 1 - bia respectively;

[0110] Step VIII-2: Calculate the neighborhood of H local ) through the neighborhood transformation τ(H local and randomly sample a sequence H in and then return it;

[0111] Step VIII-3: Randomly sample a sequence H in the solution space set Σ and then return it.

[0112] With such settings, the algorithm is restarted by randomly perturbing and re - selecting the initial sequence. By adopting the method of combining random neighborhood transfer and randomly selecting the initial sequence, each transfer method has a certain probability. Random neighborhood transfer can achieve small perturbations, which is beneficial for fewer iteration times, but it is easier to search for the same local optimal solution. While randomly selecting the initial sequence can reduce the number of times of searching for the same local optimal solution. The two are combined with a certain probability to achieve better search for specific data. Other compositions and connection relationships are the same as those in the specific embodiments one, two, three, four, five, or six.

[0113] Embodiment 1:

[0114] To better understand the foregoing invention content, according to the method of the present invention, taking the sorting of 88 - blade steam turbine blades as an example, as shown in Table 1, two sets of solution times are set, 84 ± 0.5 s and 1470 ± 1 s, and the ILS and TILS algorithms are used for calculation respectively. Set the parameters step = 3, C th = 1000, bia = 0.1, and the results are shown in Table 2.

[0115] Among them, the average number of iteration steps is the number of local iterations required to reach the local optimal solution on average each time. It can be seen that the number of iteration steps of TILS is significantly reduced. Therefore, the number of average local optimal solutions it searches for increases, and thus there is a greater probability of searching for a smaller local optimal solution.

[0116] Table 1 Mass - radius product data of 88 - blade steam turbine blades

[0117]

[0118]

[0119] Table 2 Comparison of ILS and TILS results

[0120]

[0121] Embodiment 2:

[0122] ​In order to further illustrate the advantages of the present invention in the leaf sorting algorithm, the present invention is now compared with the following in the prior art: 1. "A method for sorting turbine moving blades [J]. Gas Turbine Technology, 2018, 31(03): 39-42", which is used in steam turbines and adopts the grouping sorting method. 2. "Application of genetic algorithm in optimizing the sorting of blade mass moment [J]. Journal of Aerospace Power, 2011, 26(01): 204-209", which is used in aero engines and adopts the genetic algorithm. 3. "A blade sorting method for minimizing the imbalance of concentric rotors in aero engines", which is used in aero engines and adopts the CAGA algorithm. The above three algorithms are respectively experimented with the present invention at the same time, and the following data are obtained, as shown in Table 3:

[0123] Table 3 Comparison effects of TILS and other algorithms

[0124]

[0125]

[0126] It can be seen from the above experimental comparison that within an observable time, compared with the grouping sorting, genetic algorithm, and CAGA algorithm, the final solution results of the present invention are respectively reduced to 0.3% - 31% of their minimum values.

[0127] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A blade sorting method based on a threshold-based local iterative search algorithm, characterized in that: It includes the following steps: Step 1: Number the blades: Step 1.1: For the n blades evenly distributed on the steam turbine shaft, starting from any one blade, number the blades as 1, 2... n in the counterclockwise direction; Step 1.2: Define any combination of the n numbers as a sequence H of the product of mass and radius of the blades; Step 1.3: Form the solution space set with all the sequences H, denoted as Σ; Step 2: Initialize the sequence H in Step 1.2; Randomly initialize a sequence H. Define the neighborhood transformation of the local search algorithm as any two elements in the sequence H, denoted as τ(H). Define the neighborhood of the sequence H as the set of all its neighborhood transformations, denoted as Σ H Set the threshold of the threshold-based local iterative search algorithm to C th The step size is step, and the bias probability is bia; Step 3: Conduct local search on the sequence H in Step 2; Using sequence H as the initial sequence, perform local search iterative descent and return the locally optimal solution sequence H obtained by local search local , and all the sequences whose objective function values are less than the threshold C during the descent process th in the set Σ 1 ; Step 4: Update the global optimal solution H global ; Step Five: Determine the set Σ 1 to check if the number of elements is greater than the step size step: if yes, jump to Step Six; otherwise, jump to Step Seven; Step 6: Threshold perturbation; Take the set Σ 1 The value of the first element is denoted as Take out the element value Among all neighborhood transformations, the sequences with a downward trend are denoted as the downward trend set Σ perturbation ; Step Seven: All elements in the downward trend set Σ perturbation are successively used for the initial sequence H in Step Three, and Steps Three and Four are repeated; Step 8: Random perturbation; Perform a random neighborhood transfer on the optimal solution sequence H with a bias probability of 1 - bia local and use it as the initial sequence H. Randomly initialize an initial sequence H with a bias probability of bia. If the maximum execution time is reached, the algorithm ends; otherwise, continue to execute step two. Thus, the search for the optimal sorting of the blades is completed.

2. The leaf sorting method based on the threshold-based local iterative search algorithm according to claim 1, wherein: The definition process of the sequence H of the product of mass and radius of the blades in Step 1.2 is as follows: S1: Number the sequence H; Given an initial sequence of the product of mass and radius of a blade M = (m1r1, m2r2,......, m n r n ), the sequence H is the sequence composed of the corresponding subscripts after reordering the sequence of the product of mass and radius M; S2: Calculate the objective function; For a sequence H, first convert it into the corresponding product of mass and radius sequence M. When n blades are evenly distributed in a circumferential manner, the included angle between adjacent blades is: Define the vector corresponding to the i-th element in the mass-radius product sequence M as: Define the objective function value as:

3. A blade sorting method based on a threshold-based local iterative search algorithm according to claim 2, characterized in that: The product of mass and radius of the blades in Step 1.2 is the product of the vector radius from the center of gravity of the blade to the axis of rotation and the mass of the blade; Suppose: the mass-radius product sequence corresponding to the sequence H is M, evenly arranged on the axis, and the calculation method of the total product of mass and radius is S2.

4. A method for sorting blades based on a threshold-based local iterative search algorithm according to claim 3, characterized in that: The specific implementation method in Step 3 is: Step 3-1: Initialize a swap sequence H, and obtain the neighborhood Σ of sequence H through the neighborhood transformation τ(H). H ; Step 32: Randomly select a sequence from the neighborhood Σ of sequence H H and denote it as H new , such that H new ∈Σ H . Calculate f(H new ) and remove it from the neighborhood Σ of sequence H H ; Step 33: If f(H new ) < f(H), then H = H new ; otherwise, execute Step 35; Step 3-4: If f(H new ) < C th , add H new to the set Σ 1 , and execute Step 3-2; Step 35: If the neighborhood Σ of sequence H H is not an empty set, execute Step 32; otherwise, H local = H, return the local optimal solution H local and set Σ 1 and then end.

5. A blade sorting method based on a threshold-based local iterative search algorithm according to claim 4, characterized in that: The process of updating the global optimal solution in Step 4 is as follows; If f(H global ) < f(H local ), update H global = H local .

6. A blade sorting method based on a threshold-based local iterative search algorithm according to claim 1 or 5, characterized in that: The specific implementation method of Step 6 is as follows: Step 6-1: Take the first element in the set Σ 1 and calculate the neighborhood of the element through neighborhood transformation ​​ Step 6-2: Randomly extract a sequence from the neighborhood of element and denote it as H . Calculate f(H new ) and remove it from the neighborhood of element new . The neighborhood of is removed; Step Six Three: If Add H new to the set Σ perturbation , which is a candidate set for the threshold perturbation sequence and is used to provide the threshold perturbation sequence; Step Six Four: If the neighborhood of the element is not an empty set, execute Step Six Two; otherwise, return the set Σ and end. perturbation ​ 7. A method for sorting blades based on a threshold-based local iterative search algorithm according to claim 6, characterized in that: The specific implementation method of Step 8 is as follows: Step 8.1: Execute Step 8.2 and Step 8.3 with probabilities of bias probability bia and 1 - bia respectively; Step VIII-2: Calculate the neighborhood of H local by means of the neighborhood transformation τ(H local ), and randomly sample a sequence H from it and then return it; ​ Step 8.3: Randomly sample a sequence H from the solution space set Σ and then return.

Images