Ship pipeline path optimization method based on multi-strategy improved mucus algorithm

Abstract

A ship pipeline path optimization method based on a multi-strategy improved colistia algorithm comprises the following steps: determining an objective function, initializing population positions, calculating individual fitness and sorting, performing T distribution variation on colistia individual positions, recalculating colistia individual fitness and sorting, and performing random reverse learning to generate a new solution. And carrying out next sorting preferentially, and finally outputting an optimal solution after judgment. According to the method, cat mapping is firstly used for replacing a random population generation method in an original myxomycete algorithm, then a T distribution variation strategy and a random reverse learning strategy are adopted to increase population diversity and improve the ability of the algorithm to jump out of a local space, a high-quality ship distribution scheme can be provided for engineers in a short time under various constraints and target limitations, and the ship distribution efficiency is improved. And the method is applied to ship pipeline path planning.

Description

Technical Field

[0001] The present invention belongs to the field of ship pipeline design and construction, and particularly relates to a ship pipeline path optimization method based on a multi-strategy improved slime mold algorithm. Background Technique

[0002] The essence of ship pipeline layout design is to find the connected path between equipment interfaces in three-dimensional space and meet the path legality, effectiveness and optimization. Current mainstream CAD software, such as CATIA, FORAN, TRIBON, SolidWorks, AVEVA MARINE, although provides an automatic piping module, can only generate simple pipe paths in a limited mode and does not consider obstacle avoidance, and relies on manual adjustment by designers. In addition, ship production has the characteristics of customization and pipeline design often needs to be reworked with the adjustment of equipment layout, resulting in a large amount of manual design work and low efficiency. Therefore, exploring an automatic pipeline layout design algorithm is of great significance for improving ship design efficiency, and can also provide reference for pipeline or cable layout in the fields of aeroengines, building spaces, mechatronic equipment, and robot path planning.

[0003] Research on pipeline layout has received extensive attention from scholars at home and abroad. Representative research includes: Park et al. first proposed to use the unit generation method to automatically layout the pipelines in the ship engine room; Asmara et al. compared the differences of various ship pipeline design algorithms and proposed a solution strategy combining a heuristic algorithm (PSO) with a deterministic algorithm (A*); Fan et al. first applied the ant colony algorithm to ship pipeline layout design, and Jiang et al. and Wang et al. respectively improved the ant colony algorithm for ship piping by using co-evolution and human-machine combination strategies, but the algorithm efficiency and layout diversity still need to be improved; Liu et al. proposed an ant colony path-finding algorithm based on an adaptive dynamic adjustment strategy and applied it to two-dimensional plane pipeline design; Niu et al. proposed to layout ship pipelines based on the multi-objective optimization algorithm NSGA-II, reflecting the value of providing multiple piping solutions at the same time. The above research methods have developed from solving small-scale simple problems to solving complex practical problems, but there are still deficiencies. The main reason is that ship pipeline layout is a multi-objective optimization problem under complex constraints, with great difficulty in solving, and single methods have limitations.

[0004] The slime mold algorithm (SMA) is a stochastic optimization algorithm proposed by Li et al. It is an optimization algorithm based on the oscillation mode of slime molds in nature, simulating the behavior and morphological changes of slime molds during the foraging process. The algorithm uses an adaptive weight to simulate the process of positive and negative feedback generated by the slime mold propagation wave based on a biological oscillator to form the best path connecting food. Due to problems such as the reduction of population diversity in the later stage of iteration and the tendency to fall into local space in the original slime mold algorithm, in view of the above deficiencies, the present invention proposes a ship pipeline path optimization method based on a multi-strategy improved slime mold algorithm. Summary of the Invention

[0005] To solve the above problems, the present invention provides a method for optimizing the path of ship pipelines based on a multi-strategy improved slime mold algorithm. The technical solution adopted is as follows:

[0006] A method for optimizing the path of ship pipelines based on a multi-strategy improved slime mold algorithm, and the specific operation steps are as follows:

[0007] Step 1: Determine the objective function and initialize the population position.

[0008] The essence of ship pipeline layout is a multi-objective optimization problem under certain constraints. First, it is stipulated that the objectives and constraints can be described in the following form:

[0009]

[0010] s.t.g(path) = 0, h(path) = 0 (1)

[0011] Among them, minf length (path) calculates the path length, minf bend (path) calculates the number of path bends, minf power (path) calculates the energy value of the positions passed by the path, minf pocket (path) calculates the number of "concave pocket" structures formed by the path in the vertical direction, minf times (path) calculates the number of times the distance between path bends is less than the limit length. g(path) = 0 means that the path does not collide with the layout space, and h(path) = 0 means that the path is orthogonally arranged with the floor or wall (the bend is a right angle).

[0012] After determining the objective function, perform function transformation on it according to the actual optimization objective to obtain the fitness function, and determine the relative fitness value through the value of the objective function to facilitate the subsequent calculation of the algorithm.

[0013] Since the cat mapping has a simple structure and is not easily trapped in small cycles or fixed points, generating the initial population through cat mapping has better traversal uniformity. The cat mapping expression is:

[0014]

[0015] In the formula: d and e are arbitrary real numbers; mod1 is the fractional part of d.

[0016] Step 2: Calculate the individual fitness and sort.

[0017] Through v α,1 and v β,1The synergistic effect simulates the selective behavior of slime molds, enabling them to continue exploring higher-quality food sources while maintaining existing food sources. v α,1 The value oscillates randomly between [-a, a] and gradually approaches 0 as the number of iterations increases. v β,1 The value oscillates between [0, 1] and eventually tends to 0.

[0018] Slime molds can approach food based on the smell in the air. The approximation behavior expression of the model is:

[0019]

[0020] In the formula: v α,1 The value range of is [-a, a]; v β,1 is a parameter that decreases from 1 to 0; l is the number of iterations; X b,1 is the position of the individual with the current best fitness value; X1 is the position of the slime mold; X A,1 and X B,1 are the positions of randomly selected individuals. The update expressions for the control parameter p, the parameter v α , 1, and the weight coefficient W1 are:

[0021] p = tanh|S(δ) - R| (4)

[0022] In the formula: δ is a positive integer; S(δ) is the fitness value of X1; R is the current best fitness value.

[0023] v α,1 = [-a, a] (5)

[0024]

[0025] In the formula: lmax is the maximum number of iterations; r is a random number between 0 and 1; under certain conditions, M is the current best fitness; N is the current worst fitness; S is the sorted fitness value sequence.

[0026] Step 3: Perform T-distribution mutation on the positions of slime mold individuals.

[0027] Perform T-distribution mutation on the individual positions according to formula (9). In view of the problems such as the weak oscillation effect in the later iteration of the slime mold algorithm, being easily trapped in the local optimum, and the slow convergence speed due to the weak contraction mechanism, a mutation strategy is generally adopted to interfere with the individuals to increase the population diversity and jump out of the local optimum situation. The T-distribution is also called the Student's distribution, and the probability density function is related to the parameter degree of freedom n. The larger n is, the higher the middle part of its curve is.

[0028] To further improve the optimization ability of the SMA algorithm, a T-distribution mutation strategy is adopted for the positions of slime mold individuals, and the expression is:

[0029]

[0030] Wherein: is the position of the mutated slime mold individual; xj is the position of the j-th slime mold individual; T(ε) is the number of iterations of the algorithm, and ε is the T-distribution of the parameter degrees of freedom. The T-distribution mutation uses the number of iterations ε as the degrees of freedom parameter of the T-distribution to perturb the position of the solution, making full use of the information interference of the current population, enabling the algorithm to obtain better global search ability in the early stage of iteration and good local development ability in the later stage of iteration, and improving the convergence speed of the slime mold algorithm.

[0031] Step 4: Recalculate the fitness of the slime mold individuals and sort them.

[0032] After random perturbation by the T-distribution mutation, the fitness and the corresponding slime mold positions need to be updated. The expression for updating the slime mold position is:

[0033]

[0034] Wherein: L and U are the upper and lower bounds of the search space; τ is a random value between 0 and 1; z is the probability of slime mold mutation, generally 0.03. Then, the remaining parameters are updated through formulas (4) to (8), and the new position of the slime mold is updated through formula (10).

[0035] Step 5: Generate new solutions through random opposition-based learning and select the better ones for the next sorting

[0036] Generate new solutions through random opposition-based learning of the global optimal solution. After comparing the quality of the new solutions according to the objective function, the better solution is used as the new global optimal solution. The opposition-based learning strategy is an opposition-based learning method proposed by TIZHOOSH. It generates an opposite solution based on the current solution, compares the objective function values of the current solution and the opposite solution, and selects the better one to enter the next iteration. The expression for the opposite solution is:

[0037] k = L + U - K (11)

[0038] Wherein: K is the current solution (K ∈ [L, U]); k is the opposite solution.

[0039] Since the opposite solutions generated by the opposition-based learning strategy lack randomness and cannot effectively enhance the diversity of the population in the search space. Therefore, the random opposition-based learning strategy (ROBL) not only enhances the population diversity but also improves the ability of the population to jump out of the local optimum. The calculation expression is:

[0040] k τ = L + U - rK (12)

[0041] In the formula, kτ is the random reverse solution. Fitness can be understood as the degree of satisfaction of the solution with respect to the objective function. After optimization by random reverse learning, by comparing the fitness values of the current individual and the optimized individual, the individual with the better fitness value is selected as the global optimal solution.

[0042] Step 6: Determine whether the preset goal is reached. If it is reached, the loop ends and the optimal solution is output. If not, return to the T-distribution mutation.

[0043] For the above method for optimizing the path of ship pipelines based on a multi-strategy improved slime mold algorithm, further, the maximum number of iterations is set to 800.

[0044] For the above method for optimizing the path of ship pipelines based on a multi-strategy improved slime mold algorithm, further, in step 4, z is generally 0.03.

[0045] For the above method for optimizing the path of ship pipelines based on a multi-strategy improved slime mold algorithm, further, the set value of the parameter degree of freedom in formula (9) is 0.1.

[0046] The present invention first uses the cat map to replace the method of randomly generating the population in the original slime mold algorithm, and then adopts the T-distribution mutation strategy and the random reverse learning strategy to increase the population diversity and improve the ability of the algorithm to jump out of the local space, so as to provide high-quality ship piping solutions for engineers in a short time under various constraints and objective limitations, and apply them to the ship pipeline path planning. Description of the Drawings

[0047] Figure 1 It is a flowchart of operations. Detailed Embodiments

[0048] The main engine fuel pipeline system of a 200,000-ton bulk carrier needs to meet the following requirements: the total path ≤ 85 meters, the number of right-angle elbows ≤ 8; the number of vertical concave pocket structures ≤ 2 (to prevent fuel gas blockage); the elbow spacing ≥ 1.2 meters (to avoid stress concentration).

[0049] Step 1: Determine the objective function and the number of iterations, initialize the population position, and generate the predator matrix. The decomposition steps are as follows:

[0050] a) Determine the objective function through formula (1) after setting the following constraints and their weights.

[0051] Parameter Name Target Weight <![CDATA[minf length (path)]]> Path length ≤ 210 m 20％ <![CDATA[minf bend (path)]]> Number of bends ≤ 30 40％ <![CDATA[minf power (path)]]> Energy value at the position where the path passes The higher the better 20％ <![CDATA[minf pocket (path)]]> Number of pockets ≤ 3 20％ <![CDATA[minf times (path)]]> Elbow spacing ≥ 1.2 m Meet Must meet g(path) Collision At 0 Must meet h(path) Orthogonal arrangement Meet Must meet l Maximum number of iterations 800

[0052] b) Generate the initial population position through the cat map by formula (2).

[0053]

[0054]

[0055] Initialization parameters of the cat mapping, mapping formula:

[0056]

[0057] Initialize the initial population matrix P initial

[0058]

[0059] Step 2: Calculate the individual fitness and sort them.

[0060] The following parameters need to be set in formulas (3)(4)(5)(6)(7)(8).

[0061]

[0062]

[0063] After Step 2 is executed, the fitness ranking matrix is as follows:

[0064] Individual number X coordinate Y coordinate Z coordinate Fitness Rank 1 55.0 45.0 28.75 0.82 1 2 32.5 22.1 11.8 0.78 2 3 38.9 18.4 13.5 0.75 3 ... ... ... ... ... ... 28 215.0 85.0 35.0 0.45 28 50 5.0 25.0 28.75 0.75 32

[0065] Step 3: Perform T-distribution mutation on the positions of the slime mold individuals.

[0066] The degree of freedom parameter needs to be set in formula (9).

[0067] Parameter Name Set value ε Parameter degree of freedom 0.1

[0068] The population matrix P after mutation mutated (Partial example):

[0069]

[0070] Boundary correction: If it exceeds the search space after mutation (such as X>215), reset it to the boundary value.

[0071] Step 4: Recalculate the fitness of the slime mold individuals and sort them.

[0072] The degree of freedom parameter needs to be set in formula (10).

[0073]

[0074] Adjustment example (individual 1): Conditional judgment: τ = 0.02 < z, triggering random reset.

[0075] New position:

[0076] The population matrix P after adjustment adjusted (Partial):

[0077]

[0078] Step 5: Randomly perform reverse learning to generate a new solution, and select the better one for the next sorting.

[0079] Parameter Name Set value λ Flight coefficient 0.7 α Step size scaling factor 0.5

[0080] In this example, the global optimal solution K = [55.0, 45.0, 28.75], and the following reverse solutions are generated:

[0081] r = 0.6k τ = [5 + 215 - 0.6×55.0, 5 + 85 - 0.6×45.0,

[0082] 10 + 35 - 0.6×28.75] = [181.0, 57.0, 30.75]

[0083] 2. Perform random reverse learning. The fitness of the current solution K is f(K) = 0.82, and the fitness of its random reverse solution kτ is f(kτ) = 0.75. Since f(kτ) < f(K) (the lower the fitness value, the better the path), kτ is retained as the better solution.

[0084] Step 6: When the preset goal is reached, the loop ends and the optimal solution is output.

[0085] The fuel pipeline optimization scheme based on the improved slime mold algorithm has achieved remarkable results in the main engine fuel system of a 208,000-ton bulk carrier: after optimization, the path length has been shortened from 58.2 meters to 49.7 meters (a reduction of 14.6%), the number of elbows has been reduced from 7 to 4 (a reduction of 42.9%), effectively reducing the fuel delivery resistance and the risk of local pressure drop; through the calculation and verification of the energy value, the pipeline stress index has been optimized from 1.32 to 0.87 (a decrease of 34.1%), significantly improving the anti-vibration and anti-fatigue performance; in addition, the 2 "pocket" structures and 3 elbow spacing violations in the original design have been completely eliminated, meeting the mandatory code requirements for ship pipeline layout. Combining with engineering verification, the optimization scheme reduces the material cost of the fuel system by 12% and the fuel pump power demand by 8%. At the same time, through finite element analysis, it is confirmed that the maximum stress of the pipeline is lower than the RCC-M code limit, achieving the multi-objective collaborative optimization of economy, safety and compliance.

Claims

1. A ship pipeline path optimization method based on a multi-strategy improved slime mold algorithm, characterized in that: The specific operations are as follows: Step 1: Determine the objective function and initialize the population position; First, it is stipulated that the objectives and constraints can be described as follows: Among them, min f lengt h (path) Calculate path length, min f bend (path) Calculate the number of path bends, minf power (path) Calculate the energy value of the path passing through the position, min f pocket (path) Calculate the number of "concave pockets" formed by the path in the vertical direction, min f times (path) Calculate the number of times the path bending spacing is less than the limit length. g(path) = 0 means that the path does not collide with the layout space. h(path) = 0 means that the path is arranged orthogonally to the floor or wall, and the bend is a right angle. After determining the objective function, the fitness function is obtained by performing function transformation according to the actual optimization goal. The relative fitness value is determined by the numerical value of the objective function, and the initial population position is generated by cat mapping through formula (2); Where: d and e are arbitrary real numbers; mod1 is the decimal part of d; Step 2: Calculate individual fitness and sort them; By v a,1 and v β,1 The synergistic effect of the slime mold simulates the selective behavior of the slime mold, allowing it to continue to explore higher quality food sources while maintaining the existing food sources. α,1 The value oscillates randomly between [-a, a] and gradually approaches 0 as the number of iterations increases. β,1 The value oscillates between [0,1] and eventually tends to 0; Slime mold can approach food based on the smell in the air. The model’s approach behavior is expressed as: Where: v α,1 The value range of v is [-a, a]; β,1 is a parameter that decreases from 1 to 0; l is the number of iterations; X b,1 is the position of the individual with the best current fitness value; X1 is the position of the slime mold; X A,1 and X B,1 To randomly select the position of an individual, control parameters p and v α,1 The update expression of the weight coefficient W1 is: p = tanh|S(δ)-R| (4); Where: δ is a positive integer; S(δ) is the fitness value of X1; R is the current optimal fitness value; in α,1 =[-a,a] (5); Where: l max is the maximum number of iterations; r is a random number between 0 and 1; M is the current best fitness; N is the current worst fitness; S is the sorted fitness value sequence; Step 3: Perform T distribution variation on the individual positions of slime molds; According to formula (9), the T distribution variation of individual positions is performed; Where: is the position of the individual slime mold after mutation; x j is the position of the jth slime mold individual; T(ε) is the number of iterations of the algorithm, and ε is the parameter freedom of the T distribution; T distribution mutation uses the number of iterations ε as the freedom parameter of the T distribution to perturb the position of the solution; Step 4: Recalculate the fitness of slime mold individuals and sort them; After random perturbation by T distribution mutation, the fitness and the corresponding slime mold position are updated by formula (10); Where: L and U are the upper and lower bounds of the search space, τ is a random value between 0 and 1, and z is the probability of the slime mold mutating. Then, the remaining parameters are updated using formulas (4) to (8), and the new position of the slime mold is updated using formula (10); Step 5: Random reverse learning generates new solutions and selects the best ones for the next sorting; The global optimal solution is subjected to random reverse learning through formula (12) to generate a new solution. The quality of the new solution is compared according to the objective function and the better solution is used as the new global optimal solution. k τ =L+U-rK (12); In the formula, k τ is a random reverse solution, rK is the current solution (rK∈[L,U ] ); Step 6: Determine whether the preset target is achieved. If so, the loop ends and the optimal solution is output. If not, return to step 3 and perform T distribution variation again.

2. According to claim 1, a ship pipeline path optimization method based on a multi-strategy improved slime mold algorithm is characterized in that: Set the maximum number of iterations to 800.

3. According to claim 1, a ship pipeline path optimization method based on a multi-strategy improved slime mold algorithm is characterized in that: In step 4, z is generally 0.

03.

4. According to claim 1, a ship pipeline path optimization method based on a multi-strategy improved slime mold algorithm is characterized in that: The parameter freedom value in formula (9) is set to 0.1.

Images