Shipyard steel plate storage yard dynamic optimization method based on space-time coupling risk propagation

Abstract

The invention discloses a shipyard steel plate storage yard dynamic optimization method based on space-time coupling risk propagation. The method aims at solving the problem that the plate turnover operation efficiency is low due to the fact that a traditional steel plate storage yard management method of a shipyard cannot dynamically adapt to production plan changes. The core innovation of the invention lies in that a dynamic'plate turnover risk field 'model is provided, and a global risk field of a steel plate storage yard is constructed to guide optimization by quantifying the potential plate turnover risk of each stacked steel plate on a vacant site below the steel plate in the future based on the warehouse-out time of the steel plate and considering the exponential attenuation effect of the risk along with the layer distance. On the basis, the method combines a self-adaptive greedy randomization construction strategy and a path reconnection enhanced local search strategy, the self-adaptive greedy randomization construction strategy generates a high-quality initial solution through dynamic balance greedy and randomness, the path reconnection enhanced local search strategy evolves to a historical elite solution through guiding search to achieve deep optimization, and local optimum is effectively avoided. According to the method, the plate turning frequency can be remarkably reduced by 30%-50%, the strong real-time dynamic response capability is achieved, the optimal stacking scheme can be rapidly regenerated when the warehouse-in and warehouse-out plan is changed, and the operation efficiency, the intelligent level and the production flexibility of the shipyard steel plate storage yard are greatly improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of industrial intelligent manufacturing and logistics optimization, especially to the technical field of intelligent scheduling and management of steel plate yards in shipbuilding, heavy machinery and other industries, and particularly to a shipyard steel plate yard dynamic optimization method based on space-time coupling risk propagation. BACKGROUND

[0002] The steel plate yard is a key logistics hub of a shipbuilding enterprise. The core of its management is how to reasonably stack a large number of steel plates of different specifications and different in-out warehouse times in a limited yard area, so that the target steel plate can be taken out with the least plate turning operation when out of the warehouse. The plate turning operation not only significantly increases the operation time and cost, but also brings safety risks and equipment wear and tear. The traditional technical path, whether it is an experience rule or a traditional algorithm without improvement, lacks forward-looking assessment of the future chain reaction caused by the current stacking decision. They usually only consider the local optimum of the current state, that is, whether the current in-out warehouse is convenient, and cannot quantify the cumulative plate turning cost that a seemingly reasonable stacking position may cause in the future. When the plan changes, the original scheme may even become a new obstacle. In addition, the traditional optimization method is difficult to balance between solving efficiency and solution quality, and cannot meet the optimization calculation demand of large-scale yard dynamic rolling every day. SUMMARY

[0003] The purpose of the present application is to provide a shipyard steel plate yard dynamic optimization method based on space-time coupling risk propagation. The innovation of the method is to quantify the "number of plate turning that may be caused by future out-of-warehouse" as a calculable and propagatable "plate turning risk", and to guide each in-out warehouse decision by simulating the diffusion and superposition of the risk in the space and time dimensions of the yard, so as to avoid high-risk stacking strategies from the source, substantially reduce the number of plate turning, and improve the efficiency of yard operation.

[0004] The technical solution to achieve the purpose of the present application is: a shipyard steel plate yard dynamic optimization method based on space-time coupling risk propagation, the method comprising the following steps:

[0005] Step 1, yard state modeling: establishing a yard state model based on a three-dimensional matrix, the dimensions of the three-dimensional matrix including a stack position number , a stack layer height and a time , for representing the occupancy state of each layer of each stack position at any planning time point ; the occupancy state includes being occupied by a steel plate, being empty;

[0006] Step 2, plate turning risk quantification and propagation: constructing a dynamic plate turning risk field model, for each empty position in the yard Calculate its flip-over risk value ;

[0007] Step 3, Generate initial stacking scheme: Based on the risk field, an adaptive greedy randomization construction strategy is used to sequentially assign stacking positions to each steel plate to be put into storage, thereby obtaining an initial stacking scheme;

[0008] Step 4, Enhance the local search optimization strategy: Perform enhanced local search optimization based on path reconnection technology on the initial stacking scheme obtained in Step 3;

[0009] Step 5, Scheme generation and output: Repeat the iterative process from Step 1 to Step 4 until the termination condition is met, and finally output the optimal stacking scheme that minimizes the total number of flips during the planning period.

[0010] Furthermore, the stockpile state model based on a three-dimensional matrix in step 1 is represented as follows: ,in, It is the total number of heap positions. It is the maximum stack height of each stack position, i.e., the maximum number of layers. It is the total duration of the planning period; matrix elements in Indicates in Time, Number The first heap position The layer is occupied by a single steel plate; This means that in Time, Number The first heap position The layer is empty.

[0011] Further, in step 2, the risk value of the vacant space being flipped over is calculated. The specific process includes:

[0012] Identify empty spaces For each steel plate already stacked above, the risk contribution value of the steel plate above to the empty space is calculated based on the difference between its outbound time and the planned outbound time of the steel plate to be placed in the empty space, as well as the spatial layer distance between the two, using a preset risk attenuation factor.

[0013] The total risk value of the vacant space is obtained by summing the risk contribution values ​​of all the upper steel plates. .

[0014] Further, in step 2, the risk value of the vacant space being flipped over is calculated. The calculation formula is:

[0015]

[0016] in, It is the first a risk value of the stack position, is located in the th stack position, and the height of the layer is greater than ; and is the planned delivery time of the upper steel plate ; is the delivery time of the steel plate to be placed; represents the positive value of the time difference between the two; is a risk attenuation factor between 0 and 1, used to control the attenuation speed of the risk as the layer distance increases; and are the heights of the upper steel plate and the empty position, respectively.

[0017] Further, in step 3, the adaptive greedy randomization construction strategy includes:

[0018] defining a greedy evaluation function that takes into account the current stack position risk value and the delivery time-layer height consistency;

[0019] constructing a restricted candidate list RCL according to the current adaptive randomization parameter ;

[0020] randomly selecting a stack position for stacking from the RCL, and dynamically adjusting the parameter according to the iterative search feedback to balance the greediness and randomness of the search.

[0021] Further, the greedy evaluation function is represented as :

[0022]

[0023] wherein, is the risk value of the th stack position; is a preset space-time coordination weight coefficient; is the maximum stack height of the stack position, is the total duration of the planning period; is the delivery time of the steel plate c; is used to coordinate the tendency of steel plates with early delivery times to be placed on higher layers, is the height of the th stack position.

[0024] Further, the restricted candidate list RCL is constructed according to the current adaptive randomization parameter , specifically including:

[0025] for the steel plate , calculate the value;

[0026] select the maximum value of the values and the minimum value ;

[0027] set a threshold value wherein is a randomization parameter, and ;

[0028] include all empty slots in the restricted candidate list RCL.

[0029] Further, in step 4, the initial stacking scheme obtained from step 3 is further optimized by a reinforcement local search based on path reconnection technique, specifically: maintaining an elite solution set; constructing a path from the current solution to a selected elite solution; evaluating all intermediate solutions on the path; and using a threshold acceptance criterion to decide whether to accept the new solution, to effectively avoid local optimum.

[0030] Further, step 4 specifically includes:

[0031] (1) Design neighborhood actions, including:

[0032] a. Exchange: randomly select two steel plates with different delivery times, and exchange their stacking positions;

[0033] b. Promotion: randomly select a steel plate that is pressed, and promote it to the top of its stack position;

[0034] (2) Path reconnection, the specific steps are:

[0035] a. Maintain an elite solution set to save a number of best solutions found so far;

[0036] b. Randomly select an elite solution from as a guide solution;

[0037] c. Starting from the current solution , gradually change it to the guide solution through a series of domain actions, and each intermediate solution on the "path" of the change will be evaluated; on this path, the solution with the best evaluation is selected as the new current solution;

[0038] (3) Acceptance criterion: threshold acceptance criterion is adopted, that is, not only better solutions are accepted, but also solutions that are worse than the current solution, but the degree of difference is within a dynamically decreasing threshold .

[0039] Further, the threshold Linear or exponential decay with increasing iteration number.

[0040] Compared with the prior art, the present application has the following significant advantages:

[0041] (1) By means of the precisely quantified "turning risk" model, the algorithm can prospectively evaluate the long-term impact of stacking decisions, thereby avoiding the stacking mode with high turning frequency from the source. In actual data simulation, the optimization effect is significantly improved compared with the traditional genetic algorithm and greedy algorithm, and the total turning frequency can be reduced by 30%-50%.

[0042] (2) The algorithm proposed in the present application adopts a rolling optimization framework with strong dynamic adaptability. When the production plan changes, the algorithm can start from the current yard state, combine the new in-out warehouse plan, and quickly regenerate a high-quality stacking scheme in a short time, thereby greatly enhancing the robustness of the production system.

[0043] (3) The "adaptive greedy randomized construction" is innovatively combined with the "path reconnection local search", which not only ensures the exploration ability of the algorithm in the vast solution space, but also accelerates the convergence speed through guided search, ensuring the balance between solving efficiency and quality, and effectively processing large data problems.

[0044] (4) The "spatiotemporal coupling risk propagation model" proposed in the present application provides a new modeling perspective for the yard optimization problem. The method of converting time series constraints into spatial risk fields, as well as the application of path reconnection and other advanced optimization strategies, has significant innovation and originality in the technical field.

[0045] (5) The method proposed in the present application has clear logic, high modularization degree, and is easy to program and integrate into existing manufacturing execution systems or warehouse management systems. Its parameters have clear physical meaning and are easy to adjust, and have high promotion value in practical applications.

[0046] The present application will be described in further detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The figure is a general flowchart of the shipyard steel plate yard dynamic optimization method based on spatiotemporal coupling risk propagation of the present application.

[0048] Figure 2 The figure is a turning risk quantification and propagation diagram.

[0049] Figure 3 The figure is a flowchart of the adaptive greedy randomized construction strategy.

[0050] Figure 4 The figure is a flowchart of the local search strategy based on path reconnection. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0052] It should be noted that if the present application embodiments involve "first", "second" and the like, the "first", "second" and the like are only for description purposes and cannot be understood as indicating or implying the relative importance of the indicated technical features or implying the number of the indicated technical features. Therefore, the features with "first", "second" can be explicitly or implicitly included at least one of the features. In addition, the technical solutions of each embodiment can be combined with each other, but it must be based on the realization of the ordinary skilled in the art, when the combination of technical solutions appears contradictory or unachievable, it should be considered that the combination of technical solutions does not exist and is not within the scope of protection required by the present application.

[0053] In one embodiment, in combination Figure 1 , a shipyard steel plate yard dynamic optimization method based on space-time coupling risk propagation is provided, the method comprising the following steps:

[0054] Step 1, yard state modeling: establishing a yard state model based on a three-dimensional matrix, the dimensions of the three-dimensional matrix including berth number , berth height and time , for representing the occupancy state of each layer of each berth at any planning time point ; the occupancy state includes being occupied by a steel plate, empty;

[0055] Step 2, risk quantification and propagation of plate turning: constructing a dynamic plate turning risk field model, calculating the plate turning risk value of each empty berth in the yard;

[0056] Step 3, generating an initial stacking scheme: based on the risk field, using an adaptive greedy randomization construction strategy to sequentially assign a stacking position for each steel plate to be stored, obtaining an initial stacking scheme;

[0057] Step 4, reinforcement local search optimization strategy: performing reinforcement local search optimization based on path reconnection technology on the initial stacking scheme obtained from step 3;

[0058] Step 5, scheme generation and output: repeating the iteration process of steps 1 to 4 until the termination condition is met, and finally outputting an optimal stacking scheme that can minimize the total number of plate turning in the planning period.

[0059] Further, in one of the embodiments, the yard state model based on three-dimensional matrix in step 1 is represented as wherein, is the total number of the stacks, is the maximum stack height, i.e. the maximum number of layers, of each stack; is the total duration of the planning period; the matrix has an element representing that the layer of the stack is occupied by a steel plate at the time; represents that the layer of the stack is empty at the time.

[0060] Further, in one of the embodiments, the flipping risk value of the empty stack in step 2 is calculated as and the specific process includes:

[0061] identifying all the stacked steel plates above the empty stack, for each of the upper steel plates, calculating the risk contribution value of the upper steel plate to the empty stack by using a preset risk decay factor according to the difference between the delivery time of the upper steel plate and the planned delivery time of the steel plate to be placed in the empty stack, and the space layer distance between the two;

[0062] superimposing the risk contribution values of all the upper steel plates to obtain the total risk value of the empty stack.

[0063] Further, in one of the embodiments, the flipping risk value of the empty stack in step 2 is calculated as and the calculation formula is:

[0064]

[0065] wherein, is the flipping risk value of the stack, is the set of all the stacked steel plates in the stack with a layer height greater than ; is the planned delivery time of the upper steel plate ; is the delivery time of the steel plate to be placed in; represents the positive value of the time difference between the two; is a risk decay factor between 0 and 1, used to control the decay speed of the risk with the increase of the layer distance; and steel plate above and the height of the empty space.

[0066] Here, the derivation process of the turnover risk value is as follows:

[0067] 1. Derivation of single steel plate risk contribution: consider a steel plate already stacked at a position , and its delivery time is . Now evaluate its risk contribution to the empty space below it (where ).

[0068] a. Principle: if a new steel plate is placed at , and its delivery time is . If , when the delivery is made, the steel plate above it must be removed first, resulting in a turnover. The later , the higher the probability that needs to be removed, and the greater the risk. At the same time, the risk will decay with the increase of the spatial distance.

[0069] b. Mathematical model: this invention uses an exponential decay function to simulate the decay of risk with layer distance, and uses time difference to amplify risk. Define the risk contribution as:

[0070]

[0071] where: is the planned delivery time of the steel plate ; is the delivery time of the steel plate to be placed ; is a risk decay factor between 0 and 1, used to control the decay speed of risk with the increase of layer distance; and are the layer heights of the steel plate and the empty space, respectively, and the difference between the two represents the layer distance. .

[0072] c. Construction of overall risk field: the total risk value of the empty space in a certain stacking position in the steel plate yard is the superposition of the risk contributions of all the steel plates stacked above it, and the formula is as follows:

[0073]

[0074] where is the steel plate located in the stacking position The middle layer height is greater than All the set of stacked steel plates. The higher the matrix value, the greater the possibility of triggering the plate in the future.

[0075] Further, in one embodiment, in step 3, an adaptive greedy randomized construction strategy is constructed, which is used to gradually construct an initial stacking scheme. Instead of simply selecting the position with the lowest risk, randomness is introduced to avoid local optimization, and the randomness is adaptively adjusted. Specifically, it includes:

[0076] Define a greedy evaluation function that takes into account the current stacking position risk value and the consistency of the delivery time and stacking layer height;

[0077] According to the current adaptive randomization parameter , build a restricted candidate list RCL;

[0078] Randomly select a stacking position from RCL for stacking, and dynamically adjust the parameter according to the feedback of iterative search to balance the greediness and randomness of search.

[0079] Preferably, in some embodiments, for the steel plate to be warehoused (delivery time ), the "cost" of placing it in the empty space is determined by the risk value of the position. The greedy evaluation function is represented as :

[0080]

[0081] Wherein, is the risk value of the th stacking position; is the preset spatiotemporal coordination weight coefficient; is the maximum stacking height of the stacking position, is the total duration of the planning period; is the delivery time of steel plate c; is used to coordinate the tendency of steel plates with early delivery time to be placed in higher layers, is the layer height of the th stacking position.

[0082] Preferably, in some embodiments, the restricted candidate list RCL is constructed according to the current adaptive randomization parameter , specifically including:

[0083] For steel plate , calculate the value of all feasible empty spaces;

[0084] Select maximum value of the values and minimum value ;

[0085] setting a threshold value wherein is a randomization parameter, and ;

[0086] including all empty slots in the restricted candidate list RCL.

[0087] Further, in one of the embodiments, the initial stacking scheme obtained from step 3 is subjected to a reinforcement local search optimization based on path reconnection technique in step 4, specifically: maintaining an elite solution set; constructing a path from the current solution to a selected elite solution; evaluating all intermediate solutions on the path; and using a threshold acceptance criterion to decide whether to accept the new solution to effectively avoid local optimum; specifically: randomly selecting an elite solution from the elite solution set as a guide solution, generating a path from the current solution to the guide solution by gradually replacing the stacking position of the steel plate in the current solution with the corresponding steel plate position in the guide solution, and evaluating each intermediate solution on the path for feasibility and objective function value to find a better solution.

[0088] Preferably, in some embodiments, step 4 specifically includes:

[0089] (1) Design neighborhood actions, including:

[0090] a. Swap: randomly select two steel plates with different delivery times and exchange their stacking positions (if the constraints are met);

[0091] b. Promote: randomly select a steel plate that is pressed and promote it to the top of its stack;

[0092] (2) Path reconnection, the specific steps are:

[0093] a. Maintain an elite solution set to save a number of best solutions found so far;

[0094] b. Randomly select an elite solution from as a guide solution;

[0095] c. Starting from the current solution , gradually change it to the guide solution through a series of neighborhood actions, and evaluate each intermediate solution on the "path" of the transition; on this path, select the best evaluated solution as the new current solution;

[0096] Here, the path reconnection effectively combines the characteristics of the current solution and the historical excellent solution, can make a guided and efficient jump search in the solution space, and avoid falling into local optimum.

[0097] (3) Acceptance criterion: threshold acceptance criterion is adopted, that is, not only better solutions are accepted, but also solutions worse than the current solution but within a dynamically decreasing threshold .

[0098] Preferably, in some embodiments, the threshold decreases linearly or exponentially with the increase of the number of iterations.

[0099] In one embodiment, a shipyard steel plate yard dynamic optimization system based on space-time coupled risk propagation is provided, which comprises sequentially executed:

[0100] A first module for implementing yard state modeling: a yard state model based on a three-dimensional matrix is established, the dimensions of the three-dimensional matrix including berth number , berth height and time , for representing the occupancy state of each layer of each berth at any planning time point ; the occupancy state includes being occupied by a steel plate, empty; A second module for implementing plate flipping risk quantification and propagation: a dynamic plate flipping risk field model is constructed, and the plate flipping risk value of each empty berth in the yard is calculated;

[0101] A third module for implementing the generation of an initial stacking scheme: based on the risk field, an adaptive greedy randomization construction strategy is used to sequentially assign a stacking position for each steel plate to be stored, and an initial stacking scheme is obtained;

[0102] A fourth module for implementing a reinforced local search optimization strategy: the initial stacking scheme obtained in step 3 is subjected to a reinforced local search optimization based on the path reconnection technique;

[0103] A fifth module for implementing scheme generation and output: repeat the iteration process of the first module to the fourth module until the termination condition is met, and finally output the optimal stacking scheme that can minimize the total number of plate flipping during the planning period.

[0104] A fifth module for implementing scheme generation and output: repeat the iteration process of the first module to the fourth module until the termination condition is met, and finally output the optimal stacking scheme that can minimize the total number of plate flipping during the planning period.

[0105] ​​The specific definitions of the shipyard steel plate yard dynamic optimization system based on spatiotemporal coupling risk propagation can refer to the definitions of the shipyard steel plate yard dynamic optimization method based on spatiotemporal coupling risk propagation in the foregoing, and will not be repeated here. Each module in the shipyard steel plate yard dynamic optimization system based on spatiotemporal coupling risk propagation can be realized by software, hardware, or a combination thereof, in whole or in part. Each module can be embedded in or independent of a processor in a computer device in hardware form, or can be stored in a memory in a computer device in software form, so as to be called and executed by a processor to perform the operations corresponding to each module.

[0106] In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the following when executing the computer program:

[0107] Step 1, yard state modeling: a yard state model based on a three-dimensional matrix is established, the dimensions of the three-dimensional matrix including a yard location number , a yard layer height , and a time , for representing an occupancy state of each layer of each yard location at any planning time point ; the occupancy state includes being occupied by a steel plate, being empty;

[0108] Step 2, risk quantification and propagation of plate flipping: a dynamic plate flipping risk field model is constructed, and a plate flipping risk value is calculated for each empty location in the yard;

[0109] Step 3, generating an initial stacking scheme: based on the risk field, an adaptive greedy randomization construction strategy is used to sequentially assign a stacking position for each steel plate to be stored, to obtain an initial stacking scheme;

[0110] Step 4, reinforcement local search optimization strategy: the initial stacking scheme obtained in Step 3 is subjected to reinforcement local search optimization based on path reconnection technology;

[0111] Step 5, scheme generation and output: the iteration process of Steps 1 to 4 is repeated until a termination condition is met, and finally an optimal stacking scheme that minimizes the total number of plate flipping during the planning period is output.

[0112] The specific definitions of each step can refer to the definitions of the shipyard steel plate yard dynamic optimization method based on spatiotemporal coupling risk propagation in the foregoing, and will not be repeated here.

[0113] In one embodiment, a computer readable storage medium is provided, having a computer program stored thereon, wherein the computer program is executed by a processor to implement:

[0114] Step 1, Stockyard Condition Modeling: Establish a stockyard condition model based on a three-dimensional matrix, wherein the dimensions of the three-dimensional matrix include the stockpile number. Stack height and time Used to characterize at any planned time point Each heap position Each layer The occupancy status includes being occupied by a steel plate and being empty;

[0115] Step 2, Risk Quantification and Propagation of Flip-over: Construct a dynamic flip-over risk field model for each empty space in the yard. Calculate its flip-over risk value ;

[0116] Step 3, Generate initial stacking scheme: Based on the risk field, an adaptive greedy randomization construction strategy is used to sequentially assign stacking positions to each steel plate to be put into storage, thereby obtaining an initial stacking scheme;

[0117] Step 4, Enhance the local search optimization strategy: Perform enhanced local search optimization based on path reconnection technology on the initial stacking scheme obtained in Step 3;

[0118] Step 5, Scheme generation and output: Repeat the iterative process from Step 1 to Step 4 until the termination condition is met, and finally output the optimal stacking scheme that minimizes the total number of flips during the planning period.

[0119] For specific limitations on each step, please refer to the limitations of the dynamic optimization method for shipyard steel plate yards based on spatiotemporal coupling risk propagation mentioned above, which will not be repeated here.

[0120] As a specific example, the invention will be further verified and illustrated in one embodiment.

[0121] Using five months of actual operational data from a steel plate storage yard at a domestic shipyard as an example, this invention will be described in detail. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0122] Implementation Environment: A steel plate storage yard at a domestic shipyard has Each heap space has a limited heap size. Zhang steel plate, with The daily rolling plan is an optimization cycle.

[0123] Step 1: Data Preparation and Parameter Setting

[0124] (1) Input data: Daily steel plate inbound list and outbound plan for the next 5 months (including steel plate ID, specifications, and planned inbound / outbound dates).

[0125] (2) Algorithm parameter settings: Through preliminary experimental debugging, a set of robust parameters was set: risk decay factor Spatiotemporal coordination weights Initial randomization parameters initial threshold value Elite group size Maximum number of iterations .

[0126] Step 2: Monthly baseline plan generation

[0127] Combination Figure 1 The overall process of the method of the present invention is as follows: (1) Stockyard state modeling: establish a stockyard state model based on a three-dimensional matrix, wherein the dimensions of the three-dimensional matrix include the stockyard number. Stack height and time Used to accurately characterize at any planned time point Each heap position Each layer The occupancy status (occupied by steel plates or empty). (2) Risk quantification and propagation of flipping: Construct a dynamic flipping risk field model for each empty space in the yard. Calculate its flip-over risk value The calculation process includes: identifying all steel plates already stacked above the empty space; for each steel plate, calculating the risk contribution value of the steel plate above the empty space based on the difference between its outbound time and the planned outbound time of the steel plate to be placed in the empty space, as well as the spatial layer distance between them, using a preset risk attenuation factor; and summing the risk contribution values ​​of all the steel plates above to obtain the total risk value of the empty space. (3) Adaptive Greedy Randomization Construction Strategy: Based on the risk field, an adaptive greedy randomization strategy is adopted to sequentially assign stacking positions to each steel plate to be put into storage. This strategy includes: defining a greedy evaluation function that takes into account both the risk value of the current position and the consistency between the outbound time and the stacking layer height; and according to the current adaptive randomization parameters... Construct a restricted candidate list (RCL); randomly select positions from the RCL for stacking; and dynamically adjust parameters based on iterative search feedback. To balance the greediness and randomness of the search. (4) Strengthen the local search optimization strategy: for the initial stacking scheme obtained by step (3), a strengthened local search optimization based on path reconnection technology is carried out. This step includes: maintaining an elite solution set; from the current solution, a path leading to the selected elite solution is constructed; all intermediate solutions on the path are evaluated; and a threshold acceptance criterion is used to decide the acceptance of new solutions to effectively avoid local optimum. (5) Scheme generation and output: repeat the iteration process of steps (1) to (4) until the termination condition is met, and finally output the optimal stacking scheme that can minimize the total turnover times in the planning period.

[0128] In combination Figure 2 , the turnover risk quantification and propagation process is: the maximum stacking height of the steel plate is 70 layers, assuming that the turnover risk value of the top layer (5th layer) steel plate (the number of layers of the steel plate is not more than 70 layers) is 0, the turnover risk values of the 3rd layer and the 1st layer are calculated. From top to bottom, the risk values are sequentially superimposed and propagated, at the 3rd layer, the risk values of the 5th and 4th layers are sequentially received, the risk contribution of the 4th layer is propagated downward, and the risk of the 5th layer is not received. The 1st layer sequentially receives the risk values of the 5th, 4th and 2nd layers, the risk contribution of the 2nd layer is propagated downward, and the risks of the 5th and 4th layers are not received. In this way, the risk values of each position of the stacking position can be calculated.

[0129] In combination Figure 3 , the adaptive greedy randomization construction strategy process is: according to the current yard state and the outbound scheduled plan, the yard risk value is calculated, the greedy evaluation function of the current steel plate is calculated from the yard risk value, the top items of the restricted candidate list (RCL) are constructed by selecting the highest value of the greedy evaluation function, and a stacking position is randomly selected from the RCL to assign the current steel plate. Repeat the above steps until all steel plates are assigned to obtain a starting solution. According to the quality of the starting solution, the value is adaptively adjusted until a better solution is found.

[0130] In combination Figure 4 , the local search strategy based on path reconnection is as follows:

[0131] A starting solution, i.e. a stack position allocation scheme, is generated according to an adaptive greedy randomization construction strategy, and a number of optimal solutions are iterated in history. A guide solution is randomly selected from the starting solution and the optimal solutions. A difference set is calculated by comparing the starting solution and the guide solution to find all the different steel plates, which form a difference set. A "move" operation is defined by moving a steel plate from its current stack position in the starting solution to its target stack position in the guide solution. A feasible move with the largest improvement of the objective function is selected, and the selected move is executed to move the steel plate to the new stack position, while removing the steel plate from the difference set. The difference set is updated to obtain an intermediate solution. The intermediate solution is compared with the guide solution. If the intermediate solution is better, the intermediate solution is directly output as the global optimal solution. Otherwise, it is determined whether the difference set is empty. If the difference set is empty, the guide solution is the global optimal solution. Otherwise, it is further determined whether the iteration number exceeds an iteration threshold. If the iteration number exceeds the iteration threshold, it is indicated that a better solution cannot be found, and the path reconnection is terminated to save calculation time. Otherwise, the path reconnection iteration is continued until the difference set is empty.

[0132] At the beginning of each month, the algorithm is called. The current storage state of the yard and the detailed in-out storage plan for the next month are input. The algorithm is run, and an optimized stacking scheme is output, including the stack position and layer height of each new in-storage steel plate in the month.

[0133] Step 3: Dynamic response and rolling optimization

[0134] (1) At the beginning of each day, it is checked whether there is a change in the production plan. If there is no big change, the original scheme is executed.

[0135] (2) If there is a big change in the plan (such as an emergency order or delayed out-of-storage), the algorithm is triggered to recalculate immediately. The input is changed to the actual state of the yard at the current time and the latest in-out storage plan after the change until the end of the plan period.

[0136] (3) The algorithm outputs a new optimized scheme to guide the in-storage operation of the day.

[0137] Step 4: Turnover frequency calculation and verification

[0138] (1) During the entire simulation period, the turnover frequency is recorded according to the rules: when the target steel plate is taken, one turnover is recorded for each upper steel plate removed.

[0139] (2) The results show that the total turnover frequency of the method of the present application is significantly lower than that of the method based on the experience rule and the traditional optimization algorithm.

[0140] In summary, the present application can significantly reduce the turnover frequency by 30%-50%, and has strong real-time dynamic response capability. When the in-out storage plan changes, the optimal stacking scheme can be quickly regenerated, greatly improving the operation efficiency, intelligent level and production flexibility of the shipyard steel plate yard.

[0141] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only to illustrate the principles of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A shipyard steel plate yard dynamic optimization method based on spatiotemporal coupling risk propagation, characterized in that, The method comprises the following steps: Step 1, yard state modeling: a yard state model based on a three-dimensional matrix is established, the dimensions of the three-dimensional matrix including a stack position number , a stack position height , and a time , for representing an occupancy state of each layer of each stack position at any planning time point ; the occupancy state including occupied by a steel plate, empty; Step 2, Risk Quantification and Propagation of Flip-over: Construct a dynamic flip-over risk field model for each empty space in the yard. Calculate its flip-over risk value ; Step 3, generating an initial stacking scheme: based on a dynamic turn-over risk field model, an adaptive greedy randomized construction strategy is used to sequentially assign a stacking position for each steel plate to be stored in the warehouse, and an initial stacking scheme is obtained; Step 4, strengthening local search optimization strategy: the initial stacking scheme obtained in step 3 is subjected to a strengthening local search optimization based on path reconnection technology; Step 5, scheme generation and output: repeat the iteration process of steps 1 to 4 until the termination condition is met, and finally output the optimal stacking scheme that can minimize the total number of turn-overs within the planning period.

2. The shipyard steel plate stockyard dynamic optimization method based on spatiotemporal coupling risk propagation according to claim 1, characterized in that, The yard state model based on three-dimensional matrix in step 1 is represented as wherein, is the total number of stacks, is the maximum stack height, i.e. the maximum number of layers, of each stack; is the total duration of the planning period; the matrix has an element which indicates that the layer of the stack is occupied by a steel plate at the moment; which indicates that the layer of the stack is empty at the moment.

3. The shipyard steel plate stockyard dynamic optimization method based on space-time coupling risk propagation according to claim 1, characterized in that, The plate risk value for the empty position is calculated in step 2 The specific process includes: Identifying an empty slot All the steel plates above the empty slot are identified, and for each steel plate above the empty slot, a risk contribution value of the steel plate above the empty slot to the empty slot is calculated according to a difference between a delivery time of the steel plate above the empty slot and a planned delivery time of a steel plate to be placed in the empty slot and a space layer distance between the two, using a preset risk decay factor. The risk contribution values of all the upper steel plates are superimposed to obtain the total risk value of the gap .

4. The shipyard steel plate stockyard dynamic optimization method based on spatiotemporal coupling risk propagation according to claim 3, characterized in that, The risk value of the turntable in the empty position in step 2 is calculated The calculation formula is: ； wherein, is the risk value of the turn plate for the th stack position, is the set of all stacked steel plates in the th stack position with a height greater than ; is the planned delivery time of the upper steel plate ; is the delivery time of the steel plate to be placed ; represents the positive value of the time difference between the two; is a risk decay factor between 0 and 1, used to control the decay speed of the risk as the layer distance increases; and are the heights of the upper steel plate and the empty space, respectively.

5. The shipyard steel plate stockyard dynamic optimization method based on spatiotemporal coupling risk propagation according to claim 1, characterized in that, In step 3, the adaptive greedy randomized construction strategy comprises: Defining a greedy evaluation function that takes into account the current stacking risk value and the consistency of the delivery time and stacking layer height; According to the current adaptive randomization parameter , a restricted candidate list RCL is constructed; Randomly select a stack from the RCL for stacking and dynamically adjust parameters based on iterative search feedback to balance the greediness and randomness of the search.

6. The shipyard steel plate stockyard dynamic optimization method based on spatiotemporal coupling risk propagation according to claim 5, characterized in that, The greedy evaluation function is represented as : ； wherein, is the risk value of the th stack position; is a preset spatiotemporal coordination weight coefficient; is the maximum stack height of the stack position, is the total duration of the planning period; is the delivery time of the steel plate c; steel plates with early delivery times tend to be placed in higher layers, is the layer height of the th stack position.

7. The shipyard steel plate stockyard dynamic optimization method based on spatiotemporal coupling risk propagation according to claim 6, characterized in that, The current adaptive randomization parameter constructing a restricted candidate list RCL, specifically comprising: For a steel sheet , the values of all possible vacancies are calculated; selecting maximum value of the values and minimum value ; Setting a threshold wherein is a randomization parameter, and ; All vacancies are included in the restricted candidate list RCL.

8. The shipyard steel plate yard dynamic optimization method based on space-time coupling risk propagation according to claim 1, characterized in that, In step 4, the initial stacking scheme obtained in step 3 is subjected to a strengthening local search optimization based on path reconnection technology, specifically: maintaining an elite solution set; from the current solution, a path is constructed to the selected elite solution; all intermediate solutions on the path are evaluated; and a threshold acceptance criterion is used to decide whether to accept the new solution, in order to effectively avoid local optimum.

9. The shipyard steel plate yard dynamic optimization method based on space-time coupling risk propagation according to claim 1, characterized in that, Step 4 specifically comprises: (1) Designing neighborhood actions, including: a. Exchange: randomly select two steel plates with different delivery times and exchange their stacking positions; b. Promotion: randomly select a pressed steel plate and promote it to the top of its stacking position; (2) Path reconnection, the specific steps are: a. maintain an elite set of solutions , save a number of best solutions found so far; b. From a random elite solution is selected from as a guiding solution; c. From the current solution a series of domain actions, gradually transforming it towards a guiding solution Each intermediate solution on the "path" of this transformation is evaluated; on this path, the best evaluated solution is chosen as the new current solution; (3) acceptance criterion: a threshold acceptance criterion is adopted, i.e. not only the better solution is accepted, but also the solution which is worse than the current solution but the degree of the difference is within a dynamic decreasing threshold solutions.

10. The shipyard steel plate stockyard dynamic optimization method based on spatiotemporal coupling risk propagation according to claim 9, characterized in that, Threshold value Linear or exponential decay with increasing number of iterations.

Images