Optimization of prefabricated building construction based on dragonfly algorithm and BIM technology

Abstract

With the continuous development of construction industrialization, prefabricated buildings have been widely used due to their advantages of resource conservation, shortened construction period, environmental protection, and low pollution. However, the structural optimization design and construction scheduling optimization of prefabricated building concrete components urgently need to be solved to meet the sustainable development. Therefore, the research adopts the dragonfly algorithm optimized by multiple strategies to optimize the overall structure of prefabricated concrete components in buildings. BIM technology is applied to build a construction scheduling optimization model that satisfies resource constraints and structural constraints to optimize the scheduling plan of prefabricated components during assembly construction and find the shortest construction period and optimal scheduling plan. The experimental results show that in the comparison of intelligent algorithm iteration curves, the improved algorithm was superior to the standard DA, Nelder-Mead, NS-FPA, and CPSO algorithms. In the multi-modal function test, the optimal result of the research model was 17.54, while the standard DA, Nelder-Mead, NS-FPA, and LMA algorithms were 72.5, 52.5, and 20.3, respectively. The convergence and optimization performance of the research model were both the best. In the specific construction optimization, the improved algorithm effectively reduced the cost of concrete components and increased the ultimate bearing capacity of welded joints by 6.24%. The construction scheduling optimization model based on BIM technology saved 0.2631 working days, or 6.31 h, improving the work efficiency and reducing cost. This indicates that the research results have application value in the field of prefabricated building optimization.

Keywords

prefabricated building dragonfly algorithm BIM structural optimization scheduling optimization

Introduction

With the continuous development of construction industry, the high energy consumption of traditional buildings in labor, resources, and environment has an impact on the green development.¹ To promote industry innovation, the above issues need to be addressed. Prefabricated buildings can use the standardized and industrialized production forms to ensure project quality, accelerate project progress, reduce resource consumption, and minimize environmental pollution and labor demand, thereby achieving social and economic benefits.² How to optimize the structure and construction scheduling of prefabricated buildings is increasingly receiving attention from scholars both domestically and internationally. Yuan Z et al. developed a methodological framework to address the multiple non-lean issues faced by prefabricated wall hoisting in prefabricated building construction, like inaccurate hoisting time estimation, and unreasonable process design and resource allocation. The designed scheme reduced the Total Lift Time (TLT) by 6.3% and reduced the uncertainty of TLT by 20.6%.³ To address the significant deviations caused by various unknown factors during various stages in prefabricated buildings, Zhenmin Y et al. built an System Dynamics (SD) model based on SD theory from the supply chain. The sensitivity analysis method based on dynamic simulation was applied to analyze the data output by the SD. The installation phase had the greatest impact on construction accuracy, and the accuracy of cast-in-place structures was the most sensitive factor.⁴ Li Z et al. discovered defects such as cracks and holes on the surface of prefabricated concrete components in prefabricated buildings. A surface defect image recognition method was proposed. This method first denoised and enhanced the obtained component defect image to improve its clarity. The image was segmented and defect features of the component surface defect image were extracted based on the segmentation results. The recognition effect and accuracy of component defect image recognition were good.⁵ Quality issues caused by construction delays and errors are frequent in prefabricated buildings. Therefore, Zhao X et al. designed a monitoring method based on point cloud model feature extraction. Single column beam component data were processed to measure the quality of prefabricated column beam components. The completion model was compared with the design model to visualize the construction progress. The experimental results showed that the beam column size measurement proposed in this paper had high accuracy, with an accuracy of 95.7% in a certain beam column test. In addition, real-time visualization of progress monitoring was achieved.⁶

To weaken the construction risks of prefabricated building projects, Wei L et al. established a risk assessment index system based on factors such as personnel, machinery, materials, management, and environment. A risk assessment model was constructed based on the combination of weighted and disaster progression methods. The results indicated that the AHP-CRITIC weighting method reflected the risk issues faced by the research subjects, optimized the information data, and improved the overall evaluation performance. The research conclusion had meaningful practical significance for improving the risk management of prefabricated building construction.⁷ Qiang C et al. established a risk impact index system to address issues such as the large number of participants and the difficulty in improving risk control levels. The method provided a theoretical basis for risk control, as well as a new idea for its risk assessment.⁸ To evaluate the damage to prefabricated building structures under earthquake action, Yange L et al. explored the vulnerability of prefabricated buildings and the theory of building depreciation. A method for quantitatively evaluating the structural damage risk under seismic action was proposed. The method had advantages and assists in risk management in prefabricated building and disaster prevention and reduction.⁹ To response the information visualization, Xue O et al. developed an information visualization model based on the P-ISOMAP and Building Information Modeling (BIM) technology. In the questionnaire survey, 87.61% of respondents hoped to visualize building related information, 82.30% of structural information, and 91.15% of model display, all of which were high. Intelligent building information visualization met the needs of the public.¹⁰ Pereira J L J et al. conducted research on multi-objective optimization problems in the field of mechanical engineering. The method is to systematically review the most cited articles in this field and introduce multi-objective optimization algorithms and methods. The results indicated that the status of classical optimization methods declined due to the new algorithm, while metaheuristic algorithms were a modern trend with fierce competition among new algorithms.¹¹ Khodadadi N proposed a metaheuristic algorithm based on Newton’s cooling law to solve multi-objective optimization problems. The method was to test performance through mathematical and engineering problems and structural design examples. The algorithm could provide high-quality Pareto frontiers.¹² Lin S explored the carbon emission reduction effect from the perspective of optimizing transportation structure and transforming transportation modes. The method was to construct a multi-objective optimization model and adjust the traffic structure by combining algorithms and techniques. The results indicated that optimizing transportation structure could reduce carbon emissions, and the effects varied in different regions and transformation modes.¹³ Xue Q proposed a multi-objective optimization method based on simulation to reduce building costs and carbon emissions from the perspective of sustainable development. This method involved establishing an energy simulation model, selecting design variables, running parameter simulations, and using algorithms to search for the optimal solution. The result was a reduction in costs and carbon emissions, which could guide designers to achieve economic and environmental goals.¹⁴

BIM technology and machine learning have important applications in construction and concrete structure optimization. Especially, BIM technology provides visual building information models, while machine learning provides scheduling optimization for various construction processes, significantly improving the comprehensive construction effect of the building construction process. However, These studies have some shortcomings, such as not fully considering the synergistic effect between the overall structural optimization of prefabricated concrete components and construction progress. This study introduces the Dragonfly Algorithm (DA) and BIM technology. The key parameters output by DA, such as the optimal size of components and reinforcement, are integrated into the BIM component library through specific mapping rules. Specifically, the parameters optimized by DA are converted into a BIM recognizable format, accurately matching the corresponding BIM component attributes. For example, the optimized beam section size parameters are directly mapped to the size attributes of beam components in BIM. BIM updates the component library based on these parameters to generate building models that match the optimization results, allowing DA’s optimization results to be visually presented and applied in the BIM environment. The two work together to promote the optimization and upgrading of prefabricated buildings, thereby improving the collaborative effect between construction and schedule. There are two innovations. Firstly, the DA can effectively optimize the structure of concrete components through various strategies, improve the ultimate bearing capacity of welded joints, and reduce costs. The second is to establish a construction scheduling optimization model using BIM technology, which can find the shortest construction period and optimal scheduling plan, and improve work efficiency. Research technology also provides new solutions for optimizing prefabricated buildings, significantly improving the overall effectiveness of current construction and ensuring construction quality.

Optimization of prefabricated building construction

Optimization of prefabricated concrete structures based on improved dragonfly algorithm

The DA is a population-based intelligent biomimetic optimization algorithm, which has a simple operating mechanism and fewer parameterization problems.¹⁵ The advantages of DA mainly lie in its powerful optimization ability, efficient processing speed, and ability to adapt to large-scale operations. This algorithm simulates the static and dynamic clustering behavior of dragonflies in nature, achieving efficient exploration and development operations and demonstrating powerful optimization performance. Meanwhile, the processing speed of the DA is only related to the input bit width, making it significantly advantageous in handling complex operations such as large-scale product sums. The DA can expand radius search to enhance global search and local development capabilities, significantly improving object optimization performance. Therefore, it is used to enhance the concrete structure of prefabricated buildings and find the optimal size and reinforcement. The specific process of the algorithm is shown in Figure 1.

Figure 1.

Basic dragonfly algorithm process diagram.

From Figure 1, the principle of this algorithm is to simulate the behavior of dragonflies searching for prey.¹⁵ It first initializes the parameters, including the maximum iterations, population size, and location information. Then, based on the initial algorithm parameters, the fitness of the individual is calculated and the optimal and worst solutions in the current population are determined. Based on the results obtained from the above steps, each weight factor is updated and the relevant weight values are calculated. Weight updates include separating weights, aligning weights, and aggregating weights. The function of weight separation is to achieve individual collision avoidance through the sum of negative vectors, preventing excessive group aggregation and falling into local optima. The alignment weight is based on the average speed of neighbors to adjust the direction of motion and guide the population to migrate towards the dominant area in a coordinated manner. The aggregation weight calculates the centroid position of neighbors, prompting individuals to approach the center of the group and enhancing the algorithm development capability. By adjusting the parameters of three weights, the attractiveness of the target, and the resistance of natural enemies, the population can be accelerated towards the optimal region, as shown in equation (1).

{\begin{cases} \begin{array}{l} S_{i} = - \sum_{j = 1}^{N} (X - X_{j}) \\ A_{i} = \frac{\sum_{j = 1}^{N} V_{j}}{N} \\ C_{i} = \frac{\sum_{j = 1}^{N} X_{j}}{N} \\ F_{i} = X^{+} - X \end{array} \\ E_{i} = X^{-} - X \end{cases}

(1)

In equation (1),

S_{i}

signifies the separation amount of the

i

-th individual.

X

signifies the current location of the individual.

X_{j}

signifies the position of the

j

-th adjacent individual.

N

represents the neighboring individuals.

A_{i}

signifies the position vector of the

i

-th individual.

V_{j}

signifies the velocity of the

j

-th adjacent individual.

C_{i}

signifies the aggregation amount of the

i

-th individual.

F_{i}

represents the attractiveness to the target.

X^{+}

represents the target location.

E_{i}

represents the resistance to natural enemies.

X^{-}

represents the location of the natural predator.¹⁶ After obtaining the weight values representing different behaviors of the population, they can be used to calculate the search radius

r

, as shown in equation (2).

r = \frac{(u b - l b)}{4} + 2 (u b - l b) \frac{i t e r}{Max_i t e r a t i o n}

(2)

In equation (2),

Max_i t e r a t i o n

signifies the maximum number of iterations.

l b

signifies the minimum search space.

i t e r

signifies the model iteration.

u b

signifies the maximum search space. Next, whether there are any other individuals in the search neighborhood is determined based on the obtained search radius. If the distance between individuals is less than

r

, they are adjacent. If there are, the position update is performed, as shown in equation (3).

{\begin{cases} Δ X_{t + 1} = (s S_{i} + a A_{i} + c C_{i} + f F_{i} + e E_{i}) + ω Δ X_{t} \\ X_{t + 1} = X_{t} + Δ X_{t + 1} \end{cases}

(3)

In equation (3),

Δ X_{t + 1}

signifies the step vector.

s

signifies the separation weight.

a

stands for the alignment weight.

c

refers to the cohesive weight.

f

signifies the target weight.

e

signifies the weight of natural enemies.

ω

refers to the inertia weight.

t

signifies the current iteration count.

X_{t}

is the current population location. When there are no other individuals within the search radius, to further enhance the performance, the Levy flight strategy is introduced to adjust and update the position, thereby optimizing the search ability and random performance, as shown in equation (4).

X_{t + 1} = X_{t} + L e v y (d) \times X_{t}

(4)

In equation (4),

L e v y

signifies the Levy flight function.

d

represents the dimension. After the position update is completed, the position information of the population under the existing position conditions is updated and the fitness values of individuals is recalculated.¹⁷ The DA is prone to getting stuck in local optima, with slow convergence speed, making it difficult to preserve the historical optimal solution positions during the iteration process. Therefore, to optimize the optimization accuracy and its ability to escape from local optima, various strategies are employed to optimize the DA, as shown in Figure 2.

Figure 2.

Improved dragonfly algorithm process.

From Figure 2, the population generated by the DA during the initialization stage has randomness and uncertainty in its position.¹⁸ Therefore, the diversity of the initial positions is improved. The elite reverse learning strategy is introduced to adjust the initial population, so that the population is more likely to form the optimal population during training. The specific process is shown in equation (5).

{\begin{cases} X_{i, j}^{E^{'}} = θ (l b_{j} + u b_{j}) - X_{i, j}^{E} \\ X_{i, j}^{E^{'}} = r a n d (l b_{j} + u b_{j}) \end{cases}

(5)

In equation (5),

X_{i, j}^{E}

is the extremum point of the individual.

X_{i, j}^{E^{'}}

is the inverse solution.

θ \in [0, 1]

l b_{j}

signifies the upper boundary of the population space.

u b_{j}

signifies the lower boundary of the population space. Next, to optimize the local development performance of the population, the golden sine strategy is adopted, which effectively guides the position and movement direction of the population, as shown in equation (6).

X_{t + 1} = X_{t} \times | \sin R_{1} | + R_{2} \times \sin R_{1} \times | x_{1} \times X^{*} - x_{2} \times X_{t} |

(6)

In equation (6),

X_{t}

represents the population position.

R_{2}

represents updating the new individual position direction.

x_{1}

and

x_{2}

are both coefficients of the golden ratio.

R_{1}

is the individual action distance.

X^{*}

represents the optimal current position. To improve the global development capability and search performance, the original inertia weight is adjusted to an adaptive weight, as shown in equation (7).

ω = ω_{\max} \times \cos \frac{π \times η_{i}}{2 η_{\max}} + ω_{\min} \times (1 - \cos \frac{π \times η_{i}}{2 η_{\max}})

(7)

In equation (7),

η_{i}

signifies the current iterations.

ω_{\min}

signifies the minimum inertia factor value.

η_{\max}

is the maximum number of iterations.

ω_{\max}

signifies the maximum inertia factor value. Introducing adaptive weights can optimize the global search performance of early populations and enhance the convergence performance of the model. Next, the perturbation operations are performed on the population position to improve population diversity and avoid local convergence during training. Therefore, a random differential mutation strategy is adopted, as shown in equation (8).

X_{t + 1} = r_{1} \times (X^{*} - X_{t}) + r_{2} \times (X_{t}^{'} - X_{t})

(8)

In equation (8),

X_{t}^{'}

is the selected random population individual. In project construction management, the overall structure, beam column structure, and cost of the project are comprehensively considered to further reduce the overall project cost. The Improved Dragonfly Algorithm (IDA) is applied to design the overall structure and beam column construction of the project. The entire technical process is shown in Figure 3.

Figure 3.

Process flowchart of prefabricated component optimization.

From Figure 3, in practical optimization design, the mathematical software system is used to encode the beams and columns, and construct corresponding geometric mathematical models. The height of the beam section is $h$ . The thickness of the beam section is $b$ . In addition, the cross-sectional area of the longitudinal reinforcement of the beam is defined as $A_{1}$ . The hoop reinforcement is $A_{3}$ . The diameter of the beam hoop reinforcement is defined as $d$ . The waist reinforcement is defined as $A_{2}$ . The relevant cost calculations are shown in equation (9).

{\begin{cases} C_{1} = b h L_{1} C_{p} + A_{1} L_{1} ρ_{g} C_{z} + A_{2} L_{1} ρ_{g} C_{y} + 2 (A_{3} (b + h - 4 c_{0} + 8 d)) ρ_{g} C_{g} \\ C_{2} = h L_{2} C_{p} + A_{1} L_{2} ρ_{g} C_{z} + 2 (A_{3} (b + h - 4 c_{0} + 8 d)) ρ_{g} C_{g} + S_{a} ζ + n_{k} c_{a} \end{cases}

(9)

In equation (9),

C_{1}

represents the total cost of the beam.

C_{2}

represents the total cost of the column.

L_{1}

represents the beam length.

L_{2}

represents the column length.

c_{0}

represents the thickness of the protective layer inside the beam.

ρ_{g}

represents the density of steel bars.

C_{g}

signifies the cost of reinforcement.

C_{p}

signifies the cost of concrete per unit volume.

C_{z}

is the cost of longitudinal reinforcement.

S_{a}

represents the total area of steel bars, and

ζ

represents the consumption rate of steel bar processing.

n_{k}

represents the number of transportation times.

c_{a}

represents the transportation cost of the beam. After completing the model work, the optimization objective is set as prefabricated components, and the function constraints include strength and structural performance. The IDA is used as the optimization model. The optimization parameters are input into the algorithm model until the maximum iteration output is reached, which is the final optimal construction parameters, including steel bar configuration, prefabricated component size, steel bar configuration, etc., to ensure the lowest cost.

Construction optimization of prefabricated buildings based on BIM

In prefabricated construction, the data integration capability and visualized 3D effects of BIM technology can usually be utilized to obtain the engineering quantity parameters of each prefabricated component and the structural constraints between prefabricated components by establishing a 3D model, ensuring the construction scheduling plan.¹⁹ The research incorporates the DA into traditional BIM technology to find the optimal model of building modules during modeling. Therefore, a prefabricated building construction scheduling model based on optimized BIM technology using DA is constructed, as shown in Figure 4.

Figure 4.

Predictive building scheduling optimization.

From Figure 4, the model first takes component data, resource data, and construction information as inputs for the data. Secondly, the BIM is used to establish a 3D model, obtain structural constraints, and prepare a preliminary schedule. Finally, given the structural and resource constraints, genetic algorithm is used for optimization to produce the shortest duration and optimal scheduling plan. This model is based on multiple resources and logical sequence constraints to effectively manage the assembly process, which can optimize the assembly scheduling period of prefabricated buildings, ensure that the project schedule meets the schedule requirements and has high stability. After the model is established, BIM is used to represent the structural components like walls, columns, beams, and slabs in the project, showing the structural relationships between different types of components and determining the order of construction, as shown in Figure 5.

Figure 5.

Schematic diagram of prefabricated component structural constraints.

From Figure 5, when the horizontal component is a prefabricated board, the support of the wall needs to be set as a vertical structure prefabricated component before construction can proceed.²⁰ According to the construction, prefabricated columns and prefabricated walls are both vertical structural components, and there is no need to emphasize the sequence of the two during the construction process. However, when the column is located at the corner, the construction will prioritize prefabricated columns, followed by prefabricated wall construction. If prefabricated panels are selected as horizontal structural components, prefabricated columns should be used as support before proceeding with the next construction arrangements. If prefabricated beams are used as horizontal structural components, prefabricated columns also need to be used as supports for the next step. When both prefabricated beams and prefabricated panels are used as horizontal structural components, corresponding construction is carried out on the prefabricated beams and connected prefabricated panels, and then the next prefabricated panels construction are carried out. The relevant information and constraint relationships of the model are input into the genetic algorithm for iteration, as shown in Figure 6.

Figure 6.

Schematic diagram of the genetic algorithm process for scheduling problems.

From Figure 6, the algorithm needs to first transform the objective function of the model into the strain function in the genetic algorithm, as shown in equation (10).²¹

F (f (x)) = {\begin{cases} f_{x} - C_{\min}, f (x) > C_{\min} \\ C_{\max} - f_{x}, f (x) < C_{\max} \end{cases}

(10)

In equation (10),

C_{\min}

is the current minimum estimated value.

C_{\max}

is the current maximum estimated value. Next, the resource constraint conditions are input. The resource constraint conditions take into account the disturbances faced by construction in various scenarios such as tower cranes and workers, ensuring the effectiveness of the mathematical model. Constraints are considered from tower crane resource quantity and tower crane command resource quantity, as shown in equation (11).

{\begin{cases} \sum_{i = 1}^{n} R_{1 B i} + \sum_{j = 1}^{m} R_{1 w j} + \leq R_{1} \\ \sum_{i = 1}^{n} R_{2 B i} + \sum_{j = 1}^{m} R_{2 w j} + \leq R_{2} \\ \sum_{i = 1}^{n} R_{3 B i} + \sum_{j = 1}^{m} R_{3 w j} + \leq R_{3} \\ \sum_{i = 1}^{n} R_{4 B i} + \sum_{j = 1}^{m} R_{4 w j} + \leq R_{4} \end{cases}

(11)

In equation (11),

R_{1}

represents the quantity of tower crane resources.

R_{2}

is the quantity of tower crane command resources.

R_{3}

is the quantity of lifting worker.

R_{4}

is the quantity of steel reinforcement workers. The structural constraints to be satisfied by the selected area are input, as shown in equation (12).

\begin{array}{l} Max {C_{w i y}} \leq S_{B i}, i = 1, 2, . . . . . ., n \\ \forall x, y \in E_{i} \forall u, v \in E_{j} \end{array}

(12)

In equation (12),

E_{i}

is the set of immediate tasks for the board.

E_{j}

is the immediate task set for the wall. Given these two constraints, a series of operations for genetic algorithm are carried out.²² The probability of selecting individuals included in the input information is shown in equation (13).

p_{i} = \frac{f (x_{i})}{\sum_{j = 1}^{n} f (x_{i})}

(13)

In equation (13),

p_{i}

is the probability of being selected.

f (x_{i})

stands for the fitness of the

i

-th individual.

n

stands for the total number of individuals. The selected individual generates a new individual through crossover operation, as shown in equation (14).

x_{n e w} = (x_{i 1}, x_{i 2}, . . ., x_{i k}, . . ., x_{j n})

(14)

In equation (14),

x_{n e w}

represents a new individual. The information of the new individual is modified and the variation is introduced, as shown in equation (15).

x_{n k}^{'} = x_{n k} + δ, w i t h p r o b a b i l i t y μ

(15)

In equation (15),

δ

represents a random small variation.

μ

represents the mutation rate.²³ By continuously repeating these steps, genetic algorithms can gradually improve the quality of the solution after multiple iterations, approaching the optimal solution. After satisfying the set maximum genetic generation, the final optimal solution is determined and the decoding operation is performed to output the optimal objective function value. In addition, in performance testing, the study introduces F1 unimodal function and F2 multi-modal function to verify the algorithm’s performance. The mathematical expression of the F1 unimodal function is shown in equation (16).

F 1 (x) = \sum_{j = 1}^{n} x_{j}^{2}

(16)

In equation (16),

x_{j}

represents the variable value of the

j

-th dimension, usually within a certain range, indicating the dimension of the variable, that is, the input of the function is a

n

-dimensional vector. The F2 multi-modal function is shown in equation (17).

F 2 (x) = \sum_{j = 1}^{n - 1} [100 {(x_{j + 1} - x_{j}^{2})}^{2} + {(x_{j} - 1)}^{2}]

(17)

In equation (17),

x_{j}

represents the

j

-th dimensional variable value.

x_{j + 1}

represents the

j + 1

-dimensional variable value. They are used together with

x_{j}

to calculate the function value.

Analysis of optimization results for prefabricated building construction

Analysis of optimization results for IDA

Next, the proposed technology is tested through an improved combination of IDA and BIM technology. BIM technology has powerful data management capabilities and can systematically integrate various data during the construction process, including component information, construction progress, and resource allocation. The application interface of the system is shown in Figure 7.

Figure 7.

BIM prefabricated building construction interface.

According to the prefabricated construction interface in Figure 7, it provides BIM technology with rich functions for construction personnel, including construction progress simulation, resource optimization configuration, and collision detection. At the construction site, construction personnel can view the construction progress in real time through the BIM software interface, detect progress deviations timely, and make adjustments. By optimizing resource allocation functions, construction resources can be arranged reasonably. The collision detection function can detect potential problems during the construction process in advance, avoiding construction conflicts. In addition, in complex structural construction, BIM technology can clearly display the spatial relationships between components. To verify the performance of the IDA, the basic DA is compared with other intelligent algorithms on benchmark testing functions. The population size is 40, the maximum number of iterations is 600, and the dimension is 30. Next, the study takes F1 to represent unimodal functions and F2 multi-modal functions to test the convergence accuracy and local optimum ability of the model, with a maximum of 600 iterations. The search range of F1 function is [0, 100,000], while that of F2 function is [25,0]. The smaller the search value, the higher the search accuracy. The comparison results are shown in Figure 8.

Figure 8.

Comparison of intelligent algorithm iteration curves. (a) F₁ Curve chart. (b) F₂ Curve chart.

F₁ represents a unimodal function, which is applied to evaluate the convergence and accuracy. From Figure 8(a), compared with the basic DA, improved Nelder-Mead simplex Flower pollination algorithm (NS-FPA), and Collaborative particle swarm optimization algorithm (CPSO), the proposed IDA had better convergence, improving convergence accuracy by about 10³ times. In the later stage of convergence, the IDA curve was relatively steep, demonstrating good search vitality. F₂ represents a multi-modal function, which is used to evaluate the model’s ability to escape from local optima. From Figure 8(b), compared with other intelligent algorithms, the IDA had higher optimization accuracy and advantages in escaping local optima. The smooth IDA curve in the later stage indicates better stability during the iteration process. In multi-modal function testing, the optimal result of the research model was 17.54, while the standard DA, Nelder-Mead, NS-FPA, and LMA algorithms were 72.5, 52.5, and 20.3, respectively. Overall, the research algorithm is closer to the target value, and the optimal search performance is the best. Next, we will introduce the Zero Conductivity Transition1 function (ZDT1) and the Zero Conductivity Transition1 function (ZDT2) to compare the testing complexity of different algorithms, as shown in Table 1.

Table 1.

Comparison of testing complexity for different technologies.

Types of benchmark functions	Algorithm type		DA	LMA	NS-FPA	IDA
ZDT1	Model iteration convergence	Best	0.0034	0.0025	0.0024	0.0023
		Mean	0.0227	0.0128	0.0012	0.0011
		Var	0.0061	0.004	0.0037	0.0031
	Uniformity	SP	0.0069	0.0066	0.0066	0.0061
	Diversity	MS	0.7251	0.6912	0.6485	0.6412
	Running time(s)	10.045	9.245	8.156	7.925	7.045
ZDT2	Model iteration convergence	Best	0.0041	0.0039	0.0036	0.0035
		Mean	0.0125	0.0121	0.0146	0.012
		Var	0.0015	0.0013	0.0012	0.001
	Uniformity	SP	0.004	0.0037	0.0037	0.0035
	Diversity	MS	0.7242	0.7142	0.6912	0.6715
	Running time (s)	20.645	18.245	16.415	12.125	9.454

According to the test results, the IDA had better overall performance and shorter running time. In the ZDT1 benchmark, its time consumption was 07.045 s, which was lower than that of traditional DA (0.0034 s) and better than that of NS-FPA (7.925 s). In addition, in tests such as model iteration convergence and diversity, the research method has lower values, indicating better performance. To demonstrate the optimization performance under prefabricated beams, the main beam is set at 7,500 mm for experiments, as shown in Figure 9.

Figure 9.

Iteration curve of prefabricated beam.

From the test results in Figure 9, compared with similar technologies, the IDA had a better overall optimization effect, mainly manifested in the significantly lower cost of prefabricated beams calculated after training, and better economic efficiency. Specifically, compared with before optimization, the total project cost decreased by CNY 1,0382.21 after algorithm optimization. After meeting various constraints during construction, the demand for precast concrete and steel reinforcement in the entire project significantly decreased, further reducing project cost. This indicates that the precast beam structure used has outstanding economic advantages. The stiffness to weight ratio of the optimized beam decreased, indicating that the prefabricated beam reduced the amount of materials used after optimization, further reducing the optimization cost. To verify the optimization effect of the IDA on prefabricated columns and overall structures, a comparative experiment is conducted using the second layer prefabricated columns as an example, as shown in Figure 10.

Figure 10.

Optimization training iteration results of prefabricated columns.

From the training results in Figure 10, the IDA also had the lowest final cost, which was significantly better than similar technologies. The total cost decreased by CNY 3,2478.4 after optimizing the research algorithm. Under various construction conditions, the cost of optimized prefabricated components was significantly decreased, making the project more economical. Moreover, the DA was improved to optimize the construction of the entire structure. Compared with similar technology optimization effects, the proposed algorithm had a significantly lower cost compared with the original, greatly reducing the construction cost and also reducing the time required for producing prefabricated components. In summary, the IDA can effectively optimize the prefabricated structure. The IDA can effectively optimize prefabricated structures. The optimized overall structure, while meeting various construction conditions, also reduces the amount of construction materials used, lowers cost, and ensures construction quality. In addition, the study compares the excellent heuristic search algorithm (Attraction Exclusion Optimization Algorithm, AROA) and conducts experimental analysis using the second layer of prefabricated columns. The results are shown in Figure 11.

Figure 11.

Comparison of multi-constraint searchability of second layer prefabricated columns.

In Figure 11, the second layer of prefabricated columns is still selected for comparison. According to the test results, the proposed IDA model had better overall optimization and search performance. The best balance was achieved on material quantity, cost, and project time. The final construction time for this process in IDA was 62.25 min, while the AROA was 65.25 min. The corresponding two algorithm costs were CNY 2354 for IDA and CNY 2433 for AROA. However, the two have similar resource consumption, with IDA and ARO being 188 and 189, respectively. The IDA has better application effects in construction. Finally, the reliability of different technologies is tested. To verify the reliability of the experimental results, 15 repeated experiments are conducted on the optimization objectives of prefabricated beams and columns. The optimization costs of the IDA, Collaborative Particle Swarm Optimization Algorithm (CPSO), and Standard Dragonfly Algorithm (DA) are compared. The mean, standard deviation, and coefficient of variation (CV = SD/Mean) are taken as error indicators, and the chi square test is conducted. The results are shown in Table 2.

Table 2.

Cost Comparison of repeated experiments.

Optimization objectives	Algorithm	Mean (CNY)	SD	CV (%)	χ²	p
Precast beam	IDA	20,320	312.5	1.54	-	-
	CPSO	20,523	586.7	2.86	28.73	<0.001
	DA	20,610	893.2	4.33	63.85	<0.001
Prefabricated columns	IDA	20,120	265.8	1.32	-	-
	CPSO	26,800	645.9	2.41	1372.48	<0.001
	DA	30,280	1214.2	4.01	1002.35	<0.001

Note. p represents comparison between other algorithms and IDA.

According to the test results in Tables 2, in the second layer of prefabricated beams, IDA saved CNY 203 compared to CPSO and CNY 290 compared to DA. In prefabricated columns, the average cost of IDA compared to CPSO was CNY 20,120, which was lower than that of CPSO (CNY 26,800) and DA (CNY 30,280). In the stability analysis of prefabricated beams, the CV of IDA (1.54%) was only 35.5% of DA, indicating that its cost fluctuation range was significantly smaller. In addition, in prefabricated columns, the SD of IDA (265.8) was 79.2% lower than that of DA, indicating that it had the lowest sensitivity to initial parameters and better stability during the training process. Finally, in the chi square test, there was a significant difference (p < 0.05) between the test results of IDA and CPSO/DA. It can be found that IDA has the best stability and optimization effect in overall optimization.

Optimization analysis of prefabricated building construction based on BIM

To verify the optimization effect of BIM improved by DA on prefabricated building construction scheduling under resource constraints, a comparative analysis is conducted, taking a residential land project as an example. The project is an urban residential housing project, and 12 construction areas are set up in the project plan. 12 construction vehicles are arranged to deliver materials daily according to the set 12 routes, with an average construction period of 0.9652 working days for the components. The BIM technology is used to optimize vehicle scheduling, as shown in Figure 12.

Figure 12.

Comparison of vehicle output per unit time.

From Figure 12, the original plan for the entire project was to set up 12 routes corresponding to 12 vehicles for transportation operations, and the maximum efficiency of vehicles per hour could reached 47, and the minimum was only about 10. This indicates that the allocation of other vehicles during the scheduling process was not reasonable. After using the research model for scheduling optimization, the number of transportation vehicles was reduced to 7, and the scheduling route was re-planned. This significantly increased the transportation efficiency of each vehicle while reducing the number of vehicles, and met the requirements of engineering construction, further reducing the construction cost and project cycle. In addition, before optimization, unreasonable vehicle scheduling led to traffic congestion on construction roads and had crucial impacts on the construction of prefabricated components. However, after introducing swarm intelligence algorithms for dynamic scheduling, the above problems were improved and the efficiency of operations was greatly enhanced. To verify the optimization of BIM technology for prefabricated building construction scheduling under structural constraints, the study selects the construction process of some wall and slab components in the project for analysis, as displayed in Table 3.

Table 3.

Construction table of components under structural constraints.

Wall and slab components	Tower crane (unit)	Tower crane commander (person)	Lifting worker (person)	Reinforcement worker (person)	Construction time (working days)
1	2	2	5	3	0.0354
2	1	1	3	1	0.1278
3	1	2	3	5	0.0657
4	2	2	4	4	0.0875
5	3	4	6	5	0.0548
6	2	2	4	7	0.1049
7	3	4	5	4	0.0390
8	3	4	4	5	0.0520
9	1	2	1	2	0.0587
10	1	1	3	3	0.0571
11	2	2	5	4	0.1574
12	3	4	4	3	0.8541
13	3	4	7	5	0.0657
14	1	2	1	4	0.0459
15	1	2	1	1	0.0225
16	1	1	3	2	0.0338
17	2	2	5	3	0.0941
18	2	1	2	1	0.0562

According to the construction process and conditions of prefabricated building sites, the lifting process of prefabricated components requires tower cranes and other machinery to carry out lifting work in coordination. During this process, tower cranes need to command and issue instructions for the lifting process, and lifting workers need to locate, correct, and install prefabricated components. From Table 3, under the condition of satisfying structural constraints, the construction scheduling plan based on the improved BIM technology reasonably and effectively shortened the construction period with limited resources. Traditional methods rarely consider the structural constraints of prefabricated components when formulating construction plans. The construction scheduling plan based on BIM technology can sort the construction scheduling sequence of prefabricated components reasonably according to their attribute characteristics, shorten the construction period, and expand the advantages of prefabricated buildings. Based on the above content, to verify the construction optimization of the overall project based on the improved BIM technology for prefabricated building construction scheduling model, a Gantt chart based on the improved BIM technology in the selected area is analyzed, as shown in Figure 13.

Figure 13.

Gantt chart based on BIM construction scheduling model.

From Figure 13, the same colored areas represent prefabricated panels and prefabricated units in the same construction area. The numbers on the rectangle correspond to the numbers of the prefabricated components. According to the Gantt chart, the shortest construction period for assembly construction in the selected construction areas in the study was 0.645 days. The traditional construction scheduling method was sequential construction, without considering component information, and only scheduling from one direction to another based on the position of prefabricated components. According to the schedule and construction progress simulation process in 5.2.1, after summing up the construction time of prefabricated components, the traditional mode of sequential construction had a construction period of 0.9652 working days. The designed method saved 0.2631 working days, or 6.31 h. Therefore, the data results demonstrate the effectiveness of the designed model.

Discussion

At present, the optimization of prefabricated buildings faces a collaborative bottleneck between structural design and construction scheduling. For example, although traditional BIM technology can visualize progress, it lacks an intelligent optimization engine. However, metaheuristic algorithms often focus on single objective optimization, but their integration and depth in BIM models make it difficult to meet the requirements of deep integration and coordinated construction in prefabricated building construction. Therefore, an optimization technology for prefabricated building construction that combines IDA and BIM is proposed. It deeply integrates IDA target optimization technology with BIM technology, using a dynamic mapping machine to convert the optimized component parameters (size and reinforcement) of the DA into a BIM recognizable format, thereby optimizing the cost and management of the construction process.

From the experimental data, the IDA performs excellently in structural optimization, outperforming other similar technologies in key indicators such as cost and carrying capacity. For example, after optimization, the total cost of prefabricated beams decreased by CNY 10,382.21, the total cost of prefabricated columns decreased by CNY 1,6287.59, and the ultimate bearing capacity of welded joints increased by 6.24%. Compared with CPSO and standard DA, IDA has more advantages in cost stability and optimization accuracy, with a smaller range of cost fluctuations and lower sensitivity to initial parameters. In addition, in terms of optimizing construction progress, the improved model based on BIM significantly shortened the construction period and saved 0.2631 working days, or 6.31 h, while meeting resource and structural constraints and reducing construction costs. Compared to traditional methods, this model makes more rational use of limited resources and improves work efficiency. By combining the IDA with BIM technology, the technology addresses the shortcomings of traditional BIM technology in multi-objective optimization and coordinated construction management. In particular, the multi-strategy optimization capability of IDA and the data management capability of BIM technology complement each other, achieving integrated optimization from structural design to construction management.

The proposed technology has good application effects in the actual construction process of prefabricated buildings. Based on the multi-strategy IDA, the bottleneck of structural optimization is overcome, and dynamic mapping rules are used to achieve integrated design and construction, effectively shortening the construction period, reducing costs, improving construction quality, and promoting the intelligent development of the construction industry.

Conclusion

Aiming at the structural optimization of prefabricated building concrete components, an intelligent construction plan was proposed for improvement. A modular optimization model was built based on the DA. To improve the training performance of the model, strategies such as elite reverse strategy were introduced to optimize the algorithm, thereby enhancing its search performance and training effectiveness. The experimental results showed that when using the improved algorithm to optimize the prefabricated component structures, the cost of prefabricated beams was reduced by CNY 10,382.21, and the total cost of prefabricated columns was reduced by CNY 1,6287.59. In addition, after optimization construction, the bearing capacity of prefabricated welded joints in the ultimate bearing state was increased by 6.24% compared with cast-in-place beam column joints. A construction scheduling optimization model based on improved BIM technology using the DA was proposed to address the optimization problem of prefabricated building construction scheduling. When satisfying resource and structural constraints, the test results proved that the improved BIM technology construction scheduling optimization model could fully and reasonably utilize resources, improve the work efficiency of the construction site, and reduce cost. The construction scheduling plan developed based on this model was more detailed. Compared with the traditional construction period, this model saved 0.2631 working days, or 6.31 h, shortening construction period for prefabricated buildings. Therefore, from the above results, the proposed technology has significant optimization effects in frame structures dominated by beams and columns (such as residential projects), with good adaptability. By optimizing the size and reinforcement through IDA, the cost can be significantly reduced. However, this technology also has shortcomings, such as inadequate optimization for non-standard nodes and irregular beam column connections. The cast-in-place process is not integrated, and its adaptability to dynamic changes (such as weather disturbances) is weak. In addition, high-precision BIM modeling and algorithm parameter tuning capabilities are required for technology, and the promotion cost for small and medium-sized projects is high. Therefore, in future technology, it is necessary to expand the optimization of irregular nodes, including shear wall connections. In addition, the dynamic model combined with cast-in-place technology can enhance real-time response to changes and improve technical adaptability.

Footnotes

ORCID iD

Hongxia Min

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Shulan

Zhiwei

Hongbo

. Network model analysis of quality control factors of prefabricated buildings based on the complex network theory. Buildings 2022; 12(11): 1874.

Shengxi

Mengqi

Long

. Exploring the impact mechanism of interface management of prefabricated construction projects. Sustainability 2022; 14(21): 14440.

Yuan

Chang

Chen

, et al. Simulating and optimizing precast wall lifting in prefabricated building construction. Eng Construct Architect Manag 2024; 31(6): 2437–2456.

Zhenmin

Haina

. Improving the construction accuracy of precast components in prefabricated buildings by analyzing relevant factors from the perspective of supply chain: a system dynamics model. Architect Eng Des Manag 2024; 20(2): 230–248.

. Image recognition method of surface defects of prefabricated concrete members in prefabricated building. Int J Mater Prod Technol 2024; 68(1-2): 29–49.

Zhao

Yangze

Hongyu

. An approach for monitoring prefabricated building construction based on feature extraction and point cloud segmentation. Eng Construct Architect Manag 2023; 30(10): 5302–5332.

Wei

Lei

. A novel risk assessment model for prefabricated building construction based on combination weight and catastrophe progression method. Teh Vjesn 2023; 30(6): 1959–1967.

Qiang

Yunchao

Renxiang

. Analysis on supply chain risk factors of prefabricated buildings using AHP-DEMATEL-ISM model. Teh Vjesn 2023; 30(5): 1379–1386.

Yange

Tianyu

Zheng

. A quantitative method for assessing and monetizing the failure risk of prefabricated building structures under seismic effect. Buildings 2022; 12(12): 2221.

10.

Xue

Yizhou

Siyu

. Information visualization method for intelligent construction of prefabricated buildings based on P-ISOMAP algorithm. Int J Emerg Elec Power Syst 2022; 24(1): 73–89.

11.

Yang

Fang

Luo

, et al. A BIM-based approach to automated prefabricated building construction site layout planning. KSCE J Civ Eng 2021; 26(4): 1–18.

12.

Dong

Tian

Frank

. Application of ultra-high-performance concrete in prefabricated buildings. Mater Test 2021; 63(12): 1174–1183.

13.

Xiaoli

. Discussion on the necessity of BIM and prefabricated building whole project cooperation from the perspective of adaptive psychology. Psychiatr Danub 2021; 33(S8): 497–498.

14.

Yanwu

. Design of fabricated building wall under nano-magnetic fly ash and its prevention and effect on heavy metal pollution. Sci Adv Mater 2021; 13(12): 2424–2433.

15.

Xiaobing

Dalin

Sabah

. A RBF fuzzy logic neural network algorithm for construction resource scheduling. J Intell Fuzzy Syst 2021; 41(4): 4937–4945.

16.

Meijie

Yuming

Boya

. Influencing factors of resilience of PBSC based on empirical analysis. Buildings 2021; 11(10): 467.

17.

Simeng

Yanqiu

Nerija

, et al. Sensitivity analysis for carbon emissions of prefabricated residential buildings with window design elements. Energies 2021; 14(19): 6436.

18.

Sriraman

Manoj

Agarwal

, et al. Event-triggered control for exponential synchronization of reaction–diffusion fractional-order Clifford-valued delayed neural networks and its application to image encryption. Neurocomputing 2025; 648(1): 130604.

19.

Wellendorf

Tichelmann

Uhl

. Performance analysis of a dynamic test bench based on a linear direct drive. Archives of Advanced Engineering Science 2023; 1(1): 55–62.

20.

Purohit

Dave

. Leveraging deep learning techniques to obtain efficacious segmentation results. Archives of Advanced Engineering Science 2023; 1(1): 11–26.

21.

Aryavalli

SNG

Kumar

. Futuristic vigilance: empowering chipko movement with cyber-savvy IoT to safeguard forests. Archives of Advanced Engineering Science 2023; 1(8): 1–16.

22.

Wang

Jahanshahi

Volos

, et al. A new RBF neural network-based fault-tolerant active control for fractional time-delayed systems. Electronics 2021; 10(12): 1501.

23.

Rajchakit

Sriraman

Boonsatit

, et al. Exponential stability in the lagrange sense for Clifford-valued recurrent neural networks with time delays. Adv Differ Equ 2021; 2021: 1–21.