A multi-objective intelligent compilation and optimization method for comprehensive multi-factor construction organization design

Abstract

The present invention discloses a multi-objective intelligent compilation and optimization method for comprehensive multi-factor construction organization design, aiming to study and solve the multi-objective optimization method that simultaneously considers the lowest cost and the smallest resource volatility in the multi-mode resource-constrained project scheduling problem, and provide a set of Pareto sets that simultaneously meet the conflicting optimization objectives and construction requirements for decision-makers and construction units. The proposed method can solve the compilation and optimization problem of large-scale construction progress plans with hundreds of activities. The present invention will provide support for the compilation of construction organization designs with more comprehensive and refined management for automated construction fields such as railway engineering, and efficiently promote the implementation of projects. The beneficial effects of the present invention are as follows: it can quickly find solutions that meet the construction period in the early stage of the optimization of the multi-objective intelligent optimization algorithm. On this basis, in the middle and late stages of the optimization of the multi-objective intelligent optimization algorithm, a compilation plan with lower construction costs can be searched, ensuring the diversity of the finally obtained compilation plan and the balance of conflicting objectives.

Description

Technical Field

[0001] The present invention relates to the technical field of construction organization design, and particularly to a multi-objective intelligent compilation and optimization method for comprehensive multi-factor construction organization design. Background Technique

[0002] Construction organization design plays a crucial role in the construction management of railway projects, and can overall arrange multiple main project elements such as the construction sequence, construction period, resource allocation, and process plan of the project construction. At present, the traditional critical path method is usually used to compile the construction organization plan in railway projects. This method only focuses on the construction period target and ignores the consideration of various factors such as the site environment, production factors, process, cost, and resource balance. Therefore, it is easy to cause cost waste or plan failure and is difficult to achieve refined management of the construction process. However, during the implementation of railway projects, fluctuations in the quantity of resource allocation will bring a series of impacts. The most prominent one is the problem of the entry and exit of large-scale mechanical equipment caused by unstable allocation of large-scale mechanical resources. Frequent entry and exit will not only lead to increased wear of large-scale mechanical equipment, but also be unfavorable to the stability of the project progress, thereby increasing the uncertainty of the construction period. In addition, the entry and exit of large-scale mechanical equipment will also result in high costs. Therefore, reducing the fluctuations in resource use is crucial for the progress of the project. To sum up, there is an urgent need in the railway construction field to conduct in-depth research on the compilation principle of the construction organization plan. By establishing an effective engineering model that comprehensively considers constraint conditions such as the site environment, production factors, and process within a certain construction period and designing a multi-objective algorithm that considers two objectives of cost and resource volatility, the automatic compilation of the construction organization plan can be realized, providing accurate and quantitative decision-making support for the construction unit and the construction company, and also laying a foundation for the intelligent construction technology of less manpower / no manpower.

[0003] The goal of the resource allocation problem is to reasonably allocate limited resources in a project to achieve the lowest cost for the entire project. The resource-constrained project scheduling problem is a challenging resource allocation problem and is also a typical problem in the field of optimizing the compilation of railway engineering construction schedules. It mainly considers the duration of activities and the associated resource allocation, and resources exist in a discrete and divisible manner. The compilation of the construction organization design can be modeled through the multi-mode resource-constrained project scheduling problem. The multi-mode resource-constrained project scheduling problem takes into account that there may be multiple execution modes in actual activities. Even for the same activity, different modes will result in different operation times and resource usage amounts. However, most of the current research on solving the multi-mode resource-constrained project scheduling problem is single-objective optimization, such as minimizing cost, minimizing duration, minimizing resource volatility, etc., and the methods used are mostly exact algorithms such as the branch and bound method, which cannot be applied to the optimization of large-scale construction schedule compilation. But in practical problems, it is often necessary to comprehensively consider multiple construction objectives, such as minimizing duration, minimizing resource volatility, etc. The above optimization objectives conflict with each other and cannot be modeled in the same optimization objective, and the scale of the construction processes is generally large. Summary of the Invention

[0004] The purpose of the present invention is to provide a multi-objective intelligent compilation and optimization method for comprehensive multi-factor construction organization design, aiming to study and solve the multi-objective optimization method that simultaneously considers the lowest cost and the smallest resource fluctuation in the multi-mode resource-constrained project scheduling problem, provide a set of Pareto sets that simultaneously meet multiple optimization objectives and construction requirements for decision-makers and construction units, and can solve the optimization problem of large-scale construction schedule compilation with hundreds of activities, so as to solve the problems raised in the above background technology.

[0005] To achieve the above purpose, the present invention provides the following technical solutions:

[0006] A multi-objective intelligent compilation and optimization method for comprehensive multi-factor construction organization design specifically includes the following steps:

[0007] S1. Problem modeling: According to resource fluctuation, project cost, and project scheduling constraints, construct a multi-objective function that is optimal for cost-resource volatility, and design a multi-objective intelligent optimization algorithm;

[0008] S2. Algorithm design: Adopt a hybrid repair strategy and an adaptation value calculation method based on two-layer penalty;

[0009] S3. Output the construction schedule: Decode all individuals on the Pareto front and output the corresponding construction schedule;

[0010] The project costs in S1 include working costs and idle costs. The project scheduling constraints in S1 specifically include precedence constraints, resource constraints, duration constraints, shift-out constraints, and environmental constraints. The multi-objective function for optimal cost-resource volatility in S1 is as follows:

[0011]

[0012]

[0013] Where is for calculating costs, is for calculating the fluctuation of resource quantity, is the number of process types, is the th number of activities included in the i th process, c is the single-shift working cost of the th process, the th number of shifts arranged for the th activity of the th process, is the working time required for the th activity of the th process, is the start time of the th activity of the th process, is the end time of the th activity of the th process, is the number of resource types, is the quantity of the th type of resource used for the th activity of the th process,

[0014] The specific steps for calculating the fitness value in S2 are as follows:

[0015] Set two penalty factors and is much larger than ;

[0016] For any individual, calculate the shortest duration according to the shift arrangement and determine whether the shortest duration meets the constraints;

[0017] If the shortest duration does not meet the constraints, do not proceed to the next step of plan preparation. Set its th gene locus to 1 and impose Multiply by the penalty value for the number of days exceeding the construction period range and stop further execution; if the shortest construction period meets the construction period constraint, use the overall compilation strategy to compile the plan and check whether it meets the resource constraint, and continue to execute the following steps;

[0018] If it does not meet the resource constraint, set its th gene position to 1 and stop further execution; if it meets the resource constraint, continue to detect whether the newly compiled strategy meets the construction period constraint and continue to execute the following steps;

[0019] If it does not meet the construction period constraint, set its th gene position to 2 and impose multiplied by the penalty value for the number of days exceeding the construction period range on its fitness value; if it meets the construction period constraint, set its th gene position to 3;

[0020] The hybrid repair strategy in S2 is specifically as follows:

[0021] (1) Minimization of cost strategy: For individuals that do not meet the shift constraint after crossover, check forward from the crossover point. If the number of shifts at the crossover point is less than the number of shifts at the checkpoint, gradually reduce the number of shifts at the checkpoint until the number of shifts at the checkpoint is not greater than the number of shifts at the crossover point, or when the checkpoint and the crossover point do not belong to the same process, stop the check; for individuals that do not meet the shift constraint after mutation, perform the same check method on the mutation point as on the crossover point;

[0022] (2) Minimization of time strategy: For individuals that do not meet the shift constraint after crossover, check backward from the crossover point. If the number of shifts at the checkpoint is less than the number of shifts at the point before the crossover point, increase the number of shifts at the checkpoint and keep checking backward until the number of shifts at the checkpoint is greater than or equal to the number of shifts at the point before the crossover point or the checkpoint and the crossover point do not belong to the same process, then stop the check; for individuals that do not meet the shift constraint after mutation, perform the same check method on the mutation point as on the crossover point;

[0023] With probability to select and use the minimization of cost strategy, and with probability to select and use the minimization of construction period strategy, initially set to 0.2, every 100 generations , when , do not increase anymore.

[0024] As a preferred embodiment of the present invention: The specific steps of S2 are as follows:

[0025] (1), First, determine the basic information of the project;

[0026] (2), Randomly initialize the population,

[0027] (3) Calculate the fitness value for each individual in the initialized population;

[0028] (4) Using the current population as the parental population, apply the binary tournament selection operator to select individuals to be added to the mating pool from the current population;

[0029] (5) Subsequently, randomly pair the individuals in the mating pool, and perform multi-point crossover on the successfully paired individual pairs to generate offspring individuals;

[0030] (6) Then randomly mutate the generated offspring individuals;

[0031] (7) Use a hybrid repair strategy to repair the newly generated individuals that do not meet the shift constraints;

[0032] (8) Use a fitness value calculation method based on two-layer penalty to evaluate the fitness value of the newly generated offspring population, and classify the individuals into feasible solutions, potential feasible solutions, and completely infeasible solutions;

[0033] (9) Use the Pareto front approximation strategy to improve the potential feasible solutions;

[0034] (10) Combine the parental and offspring populations and perform non-dominated sorting, crowding degree calculation, and elite retention in sequence;

[0035] (11) Repeat the steps in (4)-(10) above until the maximum number of iterations is reached.

[0036] As a preferred solution of the present invention: The random initialization operation of the population in S2 specifically includes the following steps:

[0037] First, number different activities in the order of the process logical relationship;

[0038] Use a one-dimensional array with a length of +1 to represent the chromosome, where the first gene positions in the chromosome represent the number of shifts used by the activity, and the th gene position represents whether the chromosome individual is feasible. 1 indicates that the chromosome individual is a completely invalid solution individual, 2 indicates that the chromosome individual is a potential valid solution individual, and 3 indicates that the chromosome individual is a feasible individual;

[0039] Assign random integers that satisfy the shift constraints to the first (4) Using the current population as the parental population, apply the binary tournament selection operator to select individuals to be added to the mating pool from the current population;

[0029] (5) Subsequently, randomly pair the individuals in the mating pool, and perform multi-point crossover on the successfully paired individual pairs to generate offspring individuals;

[0030] (6) Then randomly mutate the generated offspring individuals;

[0031] (7) Use a hybrid repair strategy to repair the newly generated individuals that do not meet the shift constraints;

[0032] (8) Use a fitness value calculation method based on two-layer penalty to evaluate the fitness value of the newly generated offspring population, and classify the individuals into feasible solutions, potential feasible solutions, and completely infeasible solutions;

[0033] (9) Use the Pareto front approximation strategy to improve the potential feasible solutions;

[0034] (10) Combine the parental and offspring populations and perform non-dominated sorting, crowding degree calculation, and elite retention in sequence;

[0035] (11) Repeat the steps in (4)-(10) above until the maximum number of iterations is reached.

[0036] As a preferred solution of the present invention: The random initialization operation of the population in S2 specifically includes the following steps:

[0037] First, number different activities in the order of the process logical relationship;

[0038] Use a one-dimensional array with a length of +1 to represent the chromosome, where the first gene positions in the chromosome represent the number of shifts used by the activity, and the th gene position represents whether the chromosome individual is feasible. 1 indicates that the chromosome individual is a completely invalid solution individual, 2 indicates that the chromosome individual is a potential valid solution individual, and 3 indicates that the chromosome individual is a feasible individual;

[0039] Assign random integers that satisfy the shift constraints to the first gene positions of each individual, and initialize the th bit to 1.

[0040] As a preferred solution of the present invention: The Pareto front approximation strategy in S2 means that for any bit to 1.

[0040] As a preferred solution of the present invention: The Pareto front approximation strategy in S2 means that for any It should be noted that there seems to be some repetition and potential errors in the original text you provided. I have translated it as accurately as possible based on the given rules. If you have any further questions or need clarification, please feel free to ask. For a potentially feasible individual with the second gene locus, randomly select a feasible individual with the third gene locus, compare the sizes of the two individuals in terms of the resource volatility objective. If the potentially feasible individual performs better in the resource volatility objective, perform a local search on this individual and recalculate its fitness.

[0041] As a preferred embodiment of the present invention: The multi-point crossover and random mutation operations in S2 are specifically as follows:

[0042] Multi-point crossover: After the selection operation, randomly pair up the individuals in the population. Each pair of individuals crosses over with a probability . If the crossover operation is performed, for each pair of parent and , randomly select multiple crossover points, and according to the parity of the crossover points, copy gene segments from different parents to the offspring and ; if the crossover is not performed, directly copy the two parents to the next generation;

[0043] Random mutation: Each individual in the population independently mutates with a probability of . At each gene locus of the individual to be mutated, also with a probability of , select whether to mutate. If a mutation occurs, randomly select a new shift number within the range of optional shift numbers at this gene locus.

[0044] As a preferred embodiment of the present invention: The specific steps of the non-dominated sorting strategy in S2 are as follows:

[0045] (1), Calculate two parameters for each individual p in the population: the number of dominated individuals and the set of solutions dominated by this individual ;

[0046] (2), Put the individuals with the parameter equal to 0 into the set ;

[0047] (3), For each individual p in F1: For the individual p, traverse each individual l in the set S: Subtract 1 from . If is equal to 0, add the individual l to the set ;

[0048] (4), Continue to execute step 3 for the individuals in the set , and so on, until all Pareto levels are completely divided.

[0049] As a preferred embodiment of the present invention: The specific steps for calculating the congestion degree in S2 are as follows:

[0050] Initialize the congestion degree = 0;

[0051] For each objective function Perform the following operations: Sort the individuals of the same level for this objective function, and record as the maximum value of the individual objective function value , as the minimum value of the individual objective function value , and set the congestion degrees of the two boundaries after sorting to ∞; Set to ∞;

[0052] Calculate , where is the objective function value of the individual Q's next position after sorting, is the objective function value of the individual Q's previous position after sorting.

[0053] As a preferred embodiment of the present invention: The specific steps of the elitist retention strategy in S2 are as follows:

[0054] Combine the parent population and the offspring population to form a combined population ;

[0055] Generate a new parent population R i from the population according to the following rules: ① Arrange the individuals of the entire layer in ascending order of Pareto rank and put them into the parent population , until the individuals of a certain layer cannot all be put into the parent population ; ② Arrange the individuals of this layer in descending order of congestion degree and put them into the parent population in turn, until the size of the parent population reaches the specified upper limit.

[0056] Compared with the prior art, the beneficial effects of the present invention are:

[0057] 1. Hybrid repair strategy: The present invention proposes a hybrid repair strategy, which adaptively repairs the individuals that do not meet the shift constraints after crossover or mutation operations. It can help the population find solutions that meet the construction period faster in the early stage of evolution, help the population find solutions with lower costs in the middle and late stages of evolution, and always ensure that the population does not evolve excessively in a certain direction during the entire evolution process.

[0058] 2. Two - layer Feasible Solution Screening Mechanism: The present invention proposes a two - layer feasible solution screening mechanism. By imposing different levels of penalties on the fitness values of individuals that do not satisfy the constraints to different degrees, feasible solutions and candidate feasible solutions are screened. The first - layer screening mechanism can enable the population to quickly approach the feasible region in the early stage of the search, and significantly reduce the time used for formulating strategies and checking resources for invalid solutions. The second - layer screening mechanism can enable the population to search on the boundary of the feasible region, increasing the possibility of the algorithm finding better solutions.

[0059] 3. Pareto Front Approximation Strategy: The present invention proposes a Pareto front approximation strategy. Since there are multiple optimization objectives in the algorithm, there is a certain probability that the solutions with optimal values in the resource allocation optimization objective are infeasible solutions. Using the Pareto front approximation strategy to repair rather than directly delete these infeasible solutions that are likely to be retained helps the population move from the infeasible direction towards the Pareto front. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 is the flowchart of the multi - objective intelligent optimization algorithm of the present invention;

[0061] Figure 2 is the Gantt chart of the Pareto - optimal solution 1 of the present invention;

[0062] Figure 3 is the Gantt chart of the Pareto - optimal solution 2 of the present invention;

[0063] Figure 4 is the Gantt chart of the Pareto - optimal solution 3 of the present invention;

[0064] Figure 5 is the comparison chart of the daily funds of the three Pareto - optimal solutions of the present invention;

[0065] Figure 6 is the comparison chart of the maximum resource usage of the three Pareto - optimal solutions of the present invention;

[0066] Figure 7 is the scatter plot of the Pareto - optimal solutions and feasible solutions of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0067] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0068] Please refer to Figures 1-7 , the present invention provides an implementation method, taking a certain railway bridge project as an example:

[0069] The construction of bridge engineering mainly includes the construction of the upper part of the bridge and the construction of the lower part of the bridge. The construction of the upper part of the bridge is divided into the construction of ordinary beams and the construction of special beams. The construction of ordinary beams includes two processes: precast beam and beam erection. The construction of each special beam is regarded as an independent process. The construction of the lower part of the bridge includes three processes: pier construction, cap construction, and pile foundation construction. Each process contains multiple schedulable activities. In the same process, all activities use the same single shift, and the number of shifts can be increased in sets within the allowable range. Different numbers of shifts correspond to different execution modes. Once an activity starts, the number of shifts cannot be increased or decreased, that is, each activity can only choose one execution mode. Since the entry and exit of large equipment will have a huge impact on the project construction, therefore, in any process, once the shift enters the site, it cannot leave until all activities in this process are completed.

[0070] Consider the following two parts of costs to construct the total project cost: (1) Working cost: Workers and machinery will generate working costs during work. The single shift daily working cost and working rate of each activity are fixed, but limited by site restrictions and other complex factors, increasing the number of shifts will not directly double the working rate, but will result in a certain working rate loss. Therefore, using different numbers of shifts to complete an activity will generate different working costs. (2) Idle cost. Since the activities in the same process are not continuous, and the used shifts cannot leave the site until the process is completed, idle costs are generated. The idle cost is times the working cost, where .

[0071] When there are resource fluctuations (such as large machinery) in the arrangement of large-scale projects, it will cause very high costs. Therefore, decision-makers hope that the resource usage during the project process is as stable as possible. In this problem, the same type of process uses the same shifts for work. Therefore, in the process of solving this problem, the goal of the present invention is to ensure that the resource usage among various activities within the same process is as stable as possible. The present invention uses the sum of the absolute values of the differences in the number of shift resources used between two adjacent activities within the same process to measure the resource volatility.

[0072] There is a certain negative correlation between reducing resource fluctuations and reducing costs, and the analysis is as follows: When decision-makers need to shorten the construction period to meet the construction period constraints, in order to minimize costs, decision-makers need to increase shifts at the later positions in the same process, while reducing resource fluctuations requires decision-makers to make all activities in this process increase shifts as much as possible. To balance resource fluctuations and costs as much as possible, the present invention constructs a multi-objective function that is optimal for cost-resource volatility, and designs a multi-objective intelligent optimization algorithm to solve this problem.

[0073] The constructed multi-objective function that is optimal for cost-resource volatility is as follows:

[0074]

[0075]

[0076] The calculation cost is divided into two aspects: working cost and idle cost. The working cost is the sum of the product of the completion time required for all activities and the cost of each work shift arranged for each activity; the idle cost involves the time interval between adjacent activities of the same process, that is, the difference between the start time of the latter activity and the end time of the former activity. During this period, the work shifts arranged for the former activity are idle, and the idle cost is calculated according to the normal working cost of the work shifts that have entered the site, and the total idle cost is the sum of all idle costs.

[0077] Calculate the fluctuation of resource quantity, specifically, it is the sum of the absolute values of the differences in the usage of each resource between adjacent activities of the same process.

[0078] In a multi-objective problem, for any two feasible solutions x and y, if solution x is not worse than solution y in all objective functions and is better than solution y in at least one objective function, then solution x is said to dominate solution y, denoted as , if solutions x and y have advantages and disadvantages in different objectives, then solutions x and y are said to be in a non-dominated relationship. Further, if there does not exist any solution that can dominate solution x, then solution x is called a Pareto non-dominated solution, and all Pareto non-dominated solutions constitute the Pareto front. In a multi-objective problem, the solution algorithm needs to find a set of Pareto non-dominated solutions evenly distributed on the Pareto front as much as possible. In the solution set obtained by the algorithm, the Pareto rank of the non-dominated solution is defined as 1. Delete all non-dominated solutions from the current solution set, and a new set of non-dominated solutions will be generated in the solution set. The Pareto rank of this new set of non-dominated solutions is 2, and so on, the Pareto ranks of all solutions in the solution set can be obtained.

[0079] Engineering scheduling needs to meet the following constraints:

[0080] (1) Predecessor constraint: There is a predecessor constraint relationship between activities, and any activity can start only after all its predecessor activities are completed.

[0081] (2) Resource constraint: At the same moment, ensure that the types of labor and mechanical resources required for different activities within each process do not exceed their maximum supply limits.

[0082] (3) Project duration constraint: The project duration must be completed between , where is the earliest completion time, and is the latest completion time.

[0083] (4) Work shift departure constraint: Once the same type of process starts, the work shifts can only increase and cannot decrease.

[0084] (5)Environmental constraints: Limited by the construction environment constraints, each type of process has a maximum number of available shifts

[0085] Specifically, when in use:

[0086] 1. Randomly initialize the population: Encode the individuals: First, number different activities in the order of the process logical relationship, and then number different scheduling strategies. Therefore, the chromosome is represented by a one-dimensional array with a length of , where the first gene positions in the chromosome represent the number of shifts used for this work, and the th gene position represents whether this individual is feasible. 1 indicates that this individual is a completely invalid solution individual, 2 indicates that this individual is a potentially valid solution individual, and 3 indicates that this individual is a feasible individual. When decoding, schedule the project according to the overall scheduling strategy and the shift arrangement included in the first gene positions. The overall scheduling strategy method is as follows: Treat all activities within each process as a whole, arrange all activities within each process to be executed continuously, so that the idle cost is 0, and arrange the earliest start time that satisfies the precedence constraint for all activities within the process as the start time for the first activity in the process.

[0087] Randomly initialize the population, randomly allocate the number of shifts for each individual, ensure that the values of the first gene positions are legal integers that satisfy the shift constraint, and at the same time initialize the th gene position to 1.

[0088] 2. Fitness calculation: Only calculate the fitness of individuals with the th gene position being 1. For the resource volatility index, all individuals are directly calculated according to the resource volatility objective function. For the cost index, a two-layer feasible solution screening mechanism is proposed based on the present invention for calculation, specifically as follows: Set two penalty factors , where is much larger than ; For any individual, first calculate the shortest construction period according to the shift arrangement. If the shortest construction period does not meet the constraint, do not proceed with the next step of the project scheduling, set its th gene position to 1, and impose a penalty value of multiplied by the number of days exceeding the construction period range on its fitness; If the shortest construction period meets the construction period constraint, use the overall scheduling strategy for project scheduling, check whether it meets the resource constraint. If it does not meet, set its th gene position to 1. If it meets, continue to check whether the new scheduling strategy meets the construction period constraint. If it does not meet, set its Set a gene position to 2 and impose on its fitness value a penalty value multiplied by the number of days exceeding the construction period. If satisfied, set its th gene position to 3.

[0089] 3. Pareto front approximation strategy: For any potential feasible individual with the th gene position being 2, randomly select a feasible individual with the th gene position being 3, compare the magnitudes of the two individuals in terms of the resource volatility objective. If the potential feasible individual performs better in the resource volatility objective, perform a local search on this individual and recalculate its fitness. The local search can adopt a critical path-based method: First, calculate the critical path, and then search for the activity with the minimum crashing cost on the critical path and add a set of shifts for this activity.

[0090] 4. Multi-point crossover and random mutation: Multi-point crossover: After the selection operation, randomly pair up the individuals in the population. Each pair of individuals crosses over with a probability . If the crossover operation is performed, for each pair of parent and , randomly select multiple crossover points. According to the parity of the crossover points, copy gene segments from different parents to the offspring and . Specifically, the odd gene positions of the offspring are obtained from the parent , and the even gene positions are obtained from the parent ; the even gene positions of the offspring are obtained from the parent , and the odd gene positions are obtained from the parent , and set the th gene position of the offspring individual to 1. If the crossover is not performed, directly copy the two parents to the next generation.

[0091] Random mutation: Each individual in the population mutates independently with a probability . At each gene position of the individual to be mutated, also with a probability select whether to mutate. If mutation occurs, randomly select a new number of shifts within the range of available shifts for this gene position.

[0092] 5. Hybrid repair strategy: For individuals that do not satisfy the shift constraints after the crossover operation or mutation operation, perform the hybrid repair strategy. The specific steps are as follows:

[0093] (1) Minimization cost strategy: For individuals that do not meet the shift constraints after crossover, check forward from the crossover point. If the number of shifts at the crossover point is less than that at the checkpoint, gradually reduce the number of shifts at the checkpoint until the number of shifts at the checkpoint is not greater than that at the crossover point, or when the checkpoint and the crossover point do not belong to the same process, stop the check; for individuals that do not meet the shift constraints after mutation, perform the same checking method on the mutation point as on the crossover point.

[0094] (2) Minimization time strategy: For individuals that do not meet the shift constraints after crossover, check backward from the crossover point. If the number of shifts at the checkpoint is less than that at the point before the crossover point, increase the number of shifts at the checkpoint and keep checking backward until the number of shifts at the checkpoint is greater than or equal to that at the point before the crossover point or the checkpoint and the crossover point do not belong to the same process, then stop the check; for individuals that do not meet the shift constraints after mutation, perform the same checking method on the mutation point as on the crossover point.

[0095] With probability, select to use the minimization cost strategy, and with probability, select to use the minimization construction period strategy. The initial value is set to 0.2, and every 100 generations , when , it will not increase anymore.

[0096] 6. Non-dominated solution sorting strategy: Merge the parent and offspring populations, perform fast non-dominated sorting on the merged population, and quickly screen out solutions with a higher Pareto rank to make the entire population move rapidly towards the Pareto front. The specific steps are as follows:

[0097] (1) Calculate two parameters for each individual p in the population: the number of dominated individuals and the set of solutions dominated by this individual .

[0098] (2) Put the individuals with the parameter equal to 0 into the set .

[0099] (3) For each individual p in F: For individual p, traverse each individual l in the set S: Decrease by 1. If is equal to 0, then add individual l to the set .

[0100] (4) The above steps obtain the set of individuals with Pareto rank 2. Continue to execute step (3) for the individuals in the set , and so on, until all Pareto ranks are completely divided.

[0101] 7. Crowdedness calculation: To make the obtained solutions more uniform in the objective space, crowdedness is introduced here , and the specific steps are as follows:

[0102] (1) Initialization = 0;

[0103] (2) For each objective function perform the following operations: Sort the individuals of this level for this objective function, and denote as the maximum value of the individual objective function value , as the minimum value of the individual objective function value , and set the crowdedness of the two boundaries after sorting to ∞.

[0104] (3) Calculate , where is the objective function value of the individual after sorting and the next one, is the objective function value of the individual before sorting and the previous one.

[0105] 8. Elite retention strategy: By selecting excellent solutions from the offspring and the parent generation, the solutions at the forefront are always preserved in multiple population iterations to accelerate the convergence process of the population. The specific steps are as follows:

[0106] (1) Combine the parent population and the offspring population into a population ;

[0107] (2) Generate a new parent population from the population according to the following rules: ① Put the entire layer of the population into the parent population in ascending order of Pareto rank until the individuals of a certain layer cannot all be put into the parent population ; ② Arrange the individuals of this layer in descending order of crowdedness and put them into the parent population in turn until the size of the parent population

[0108] reaches the specified upper limit. 9. Iteration: Repeat steps 3 - 8 until the maximum number of iterations is reached, decode all individuals on the Pareto front, and output the corresponding construction progress plan.

[0109] In the above case, the minimum construction period is set to 650 days, the maximum construction period is 750 days, the population size is 20, and the maximum number of iterations is 500. The following table shows the comparison of the effects of three Pareto-dominated solutions obtained by the multi-objective intelligent optimization algorithm (the cost objective and the resource volatility objective are essentially discrete, so the obtained solutions are not a uniform front, but discrete non-dominated solutions) on four indicators: project cost, work cost, vacancy cost, and resource volatility, and gives the running time of the algorithm. From the data in the table, it can be observed that the obtained Pareto optimal solutions are in a non-dominated relationship with each other. The first Pareto optimal solution has the lowest project cost, the second Pareto optimal solution has the smallest resource fluctuation, and the third Pareto optimal solution is an equilibrium between the two objectives. All three solutions meet the construction period requirements.

[0110]

[0111] Figure 2 and 3 and 4 respectively show the Gantt charts of the three Pareto optimal solutions obtained. From Figure 2 and 4 , it can be seen that the time span of the sixth process of the first Pareto optimal solution and the third Pareto optimal solution 3 is large, which proves that the number of shifts used by the first Pareto optimal solution and the third Pareto optimal solution is not much. This is also the reason why the first Pareto optimal solution and the third Pareto optimal solution have lower costs. In addition, it can be clearly observed that there are obvious differences in the completion times between different activities of the third process of the first Pareto optimal solution and the second Pareto optimal solution, which indicates that the first Pareto optimal solution and the second Pareto optimal solution have different resource fluctuations. The increase in the degree of resource fluctuation here makes the first Pareto optimal solution have a lower cost, which also shows that there is a conflict between the lowest cost and the smallest resource fluctuation.

[0112] From Figure 5 , it can also be seen that for the first Pareto optimal solution and the third Pareto optimal solution, the daily capital use is less than that of the second Pareto optimal solution after 140 days of the construction period as the project progresses. This is also the reason why the first Pareto optimal solution and the third Pareto optimal solution have lower work costs. The daily capital use of the second Pareto optimal solution is very stable throughout the construction period without fluctuations, so the resource volatility is 0. And from Figure 6 , it can be seen that the second Pareto optimal solution has lower resource pressure compared to the first Pareto optimal solution and the third Pareto optimal solution. This is because the second Pareto optimal solution uses enough shifts to work from the start of the process and does not need to rush work in the later stage of the project. This is also the reason for the higher cost and smaller resource fluctuation of the second Pareto optimal solution.

[0113] Figure 7 The Pareto front solution set is shown, where the solid dots represent the Pareto optimal solutions, and the hollow dots represent the Pareto feasible solutions. It can be observed from the figure that the Pareto feasible solutions are evenly distributed in the objective function set, indicating that the proposed intelligent optimization algorithm has a broader exploration ability in the objective function solution set.

[0114] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-objective intelligent compilation and optimization method for comprehensive multi-factor construction organization design, characterized in that, Specifically, it includes the following steps: S1. Problem Modeling: According to resource fluctuations, project costs, and project scheduling constraints, construct a multi-objective function that is optimal for cost-resource volatility, and design a multi-objective intelligent optimization algorithm; S2. Algorithm Design: Adopt a hybrid repair strategy and an adaptation value calculation method based on two-layer penalty; S3. Output the construction schedule: Decode all individuals on the Pareto front and output the corresponding construction schedule; The project costs in S1 include work costs and vacancy costs. The project scheduling constraints in S1 specifically include precedence constraints, resource constraints, duration constraints, shift unit dispatch constraints, and environmental constraints. The multi-objective function that is optimal for cost-resource volatility in S1 is as follows: in To calculate the cost, To calculate the fluctuation of the number of resources, is the number of process types, For the The number of activities included in each process, c i is the single-shift work cost of the i-th process, For the The first step of the process Number of shifts arranged for each activity, To complete the The first step of the process Working hours required for each activity, For the The first step of the process The start time of the activity, For the The first step of the process The end time of the activity, is the number of resource types, For the The first step of the process The activity used The number of resources, The vacancy cost is a multiple of the labor cost and satisfies ; The specific steps for calculating the adaptation value in S2 are as follows: Set two penalty factors and much greater than ; For any individual, calculate the shortest duration according to the shift unit arrangement and determine whether the shortest duration meets the constraints; If the shortest construction period does not meet the constraints, the next step of plan compilation will not be carried out. Set its th gene position to 1, and impose a penalty value multiplied by the number of days exceeding the construction period range, and do not continue to execute downward; if the shortest construction period meets the construction period constraints, use the overall compilation strategy to compile the plan, check whether it meets the resource constraints, and continue to execute the following steps; If the resource constraint is not met, set its th gene bit to 1 and stop further execution; if the resource constraint is met, continue to check whether the newly compiled strategy meets the construction period constraint and continue to execute the following steps; If the construction period constraint is not met, the th gene position is set to 2, and a penalty value of multiplied by the number of days exceeding the construction period range is imposed on its fitness value; if the construction period constraint is met, the th gene position is set to 3; The specific hybrid repair strategy in S2 is as follows: (1) Cost minimization strategy: For individuals that do not meet the shift unit constraints after crossover, check forward from the crossover point. If the number of shift units at the crossover point is less than that at the checkpoint, gradually reduce the number of shift units at the checkpoint until the number of shift units at the checkpoint is not greater than that at the crossover point, or when the checkpoint and the crossover point do not belong to the same process, stop the check; For individuals that do not meet the shift unit constraints after mutation, perform the same check method as the crossover point on the mutation point; (2) Time minimization strategy: For individuals that do not meet the shift unit constraints after crossover, check backward from the crossover point. If the number of shift units at the checkpoint is less than the point before the crossover point, increase the number of shift units at the checkpoint and keep checking backward until the number of shift units at the checkpoint is greater than or equal to the point before the crossover point or the checkpoint and the crossover point do not belong to the same process, then stop the check; For individuals that do not meet the shift unit constraints after mutation, perform the same check method as the crossover point on the mutation point; With a probability of , the minimum cost strategy is selected, and with a probability of , the minimum construction period strategy is selected. The initial value is set to 0.2, and every 100 generations . When , it will not increase anymore.

2. The multi-objective intelligent compilation optimization method for comprehensive multi-factor construction organization design according to claim 1, characterized in that: The specific steps of S2 are as follows: (1) First, determine the basic information of the project; (2) Randomly initialize the population, (3) Calculate the adaptation value for each individual in the initialized population; (4) Use the current population as the parent population and apply the binary tournament selection operator to select individuals to be added to the mating pool from the current population; (5) Then randomly pair the individuals in the mating pool and perform multi-point crossover on the successfully paired individual pairs to generate offspring individuals; (6) Then randomly mutate the generated offspring individuals; (7) Use the hybrid repair strategy to repair the newly generated individuals that do not meet the shift unit constraints; (8) Use the adaptation value calculation method based on two-layer penalty to evaluate the adaptation value of the newly generated offspring population and classify the individuals into feasible solutions, potential feasible solutions, and completely infeasible solutions; (9) Use the Pareto front approximation strategy to improve the potential feasible solutions; (10) Merge the parent and offspring populations and perform non-dominated solution sorting, crowding degree calculation, and elite retention in sequence; (11) Repeat the steps (4)-(10) above until the maximum number of iterations is reached.

3. The multi-objective intelligent compilation optimization method for comprehensive multi-factor construction organization design according to claim 2, characterized in that: The operation of randomly initializing the population in S2 specifically includes the following steps: First, number the different activities in the order of the process logical relationship; Use a set of length +1 for one-dimensional arrays Represents a chromosome, where the front The first gene bit represents the number of shifts used in the activity, Each gene position indicates whether the chromosome individual is feasible, 1 means that the chromosome individual is a completely invalid solution individual, 2 means that the individual is a potential valid solution individual, and 3 means that the chromosome individual is a feasible individual; Allocate random integers that satisfy the shift constraints to the first gene positions of each individual, and initialize the th position to 1.

4. The multi-objective intelligent compilation and optimization method for comprehensive multi-factor construction organization design according to claim 2, characterized in that: The Pareto front approximation strategy in S2 means that for any potential feasible individual with the gene position of 2, randomly select a feasible individual with the gene position of 3, compare the magnitudes of the two individuals in terms of the resource volatility objective. If the potential feasible individual performs better in the resource volatility objective, perform a local search on this individual and recalculate its fitness.

5. The multi-objective intelligent compilation optimization method for comprehensive multi-factor construction organization design according to claim 2, characterized in that: The multi-point crossover and random mutation operations in S2 are specifically as follows: Multi-point crossover: After the selection operation is completed, randomly pair up individuals in the population. Each pair of individuals undergoes crossover with probability . If the crossover operation is performed, for each pair of parent and , randomly select multiple crossover points. According to the parity of the crossover points, copy gene segments from different parents to the offspring and ; If the crossover is not performed, directly copy the two parents to the next generation; Random mutation: Each individual in the population independently mutates with a probability of . For each gene locus of the individual to be mutated, it also mutates with a probability of . If a mutation occurs, a new shift number is randomly selected within the range of available shift numbers for that gene locus.

6. A multi-objective intelligent compilation and optimization method for comprehensive multi-factor construction organization design according to claim 2, characterized in that: The specific steps of the non-dominated sorting strategy in S2 are as follows: (1) Calculate two parameters for each individual p in the population: the number of individuals it is dominated by and the set of solutions dominated by this individual ; (2) Put the individuals with parameter being 0 into the set . (3) For each individual p in F1: For the individual p, traverse each individual l in the set S: Subtract by 1. If is equal to 0, then add the individual l to the set ; (4)Continue to perform step (3) on the individuals in the set and so on until all Pareto levels are completely divided.

7. A multi-objective intelligent compilation and optimization method for comprehensive multi-factor construction organization design according to claim 6, characterized in that: The specific steps of the crowding degree calculation in S2 are as follows: Initialize congestion degree = 0; For each objective function perform the following operations: perform individual sorting of the same level for this objective function, and record as the maximum value of the individual objective function value , as the minimum value of the individual objective function value , and set the crowding degree of the two boundaries after sorting to ∞; Calculation , where is the objective function value of the last individual Q after sorting, is the objective function value of the previous individual Q after sorting.

8. A multi-objective intelligent compilation and optimization method for comprehensive multi-factor construction organization design according to claim 7, characterized in that: The specific steps of the elite retention strategy of S2 are as follows: Combine the parental population A h and the offspring population B h to form a synthetic population R h ; Generate a new parental population from the population according to the following rules R h : ① Place the entire population layer by layer into the parental population in ascending order of Pareto rank until all individuals in a certain layer cannot be fully placed into the parental population ; ② Arrange the individuals in this layer in descending order of crowding degree and place them into the parental population in turn until the size of the parental population reaches the specified upper limit .

Images