Optimisation of construction mode of residential houses based on the genetic algorithm under BIM technology

The application of BIM technology in optimisation of structural design will improve the work efficiency and quality, which is mainly reflected in the following three aspects [14,15,16,17]:

The technology of construction simulation, on the one hand, can check out the collisions and conflicts in the design scheme, facilitate the engineers to optimise the design before the actual construction work and effectively reduce the occurrence of design alterations in the construction process. On the other hand, by adding information of the construction process and schedule to the BIM model, it can realise the simulation of process and progress in construction, which is beneficial for managers to make reasonable construction plans, accurately grasp the construction process, efficiently utilise construction resources and scientifically arrange construction sites. At present, there are many software programmes with functions of construction simulation, such as Navisworks software developed by Autodesk company which is often used to check the collision between components and the simulation of construction schedule [18]. Considering the common functions of software and the data link with modelling software, this paper chooses Navisworks software that is suitable for Revit to carry out collision detection and construction simulation for the BIM model.

Navisworks software is used to complete the collision inspection and construction simulation of the optimised 3D structure model. Among them, the workflow of collision inspection is shown in Figure 1:

Add files to be checked. The first step is to add NWC format files exported by Revit software to Navisworks software. The file with the NWC format is one of the three native file formats of Navisworks software, which can reduce the access time.

Fig. 1

In this paper, an optimised architecture system combining BIM technology and immune genetic algorithm is constructed, as shown in Figure 2, in which the cooperation between BIM and immune genetic algorithm is mainly accomplished through the docking of MATLAB and Navisworks. Firstly, according to the design drawing, BIM is used to accurately model the building structure, and then, the model is imported into Navisworks to complete the calculation and analysis of the building structure. Then, the engineering data required for structural design and cost are generated into Excel file A, and the immune genetic algorithm program is run in MATLAB to optimise the file A and output the result, which is Excel file B. Finally, file B is imported into Navisworks, and the structure is analysed and calculated again, where a new design scheme is generated and exported. The derived scheme of structural design can continue to guide the construction and other work by using the BIM model.

Fig. 2

The scheduling scheme of construction affects the length of the assembly period; so, it is necessary to determine a construction scheduling scheme that meets the resource constraints of personnel and equipment and has the shortest assembly time at the assembly stage. This kind of problem is regarded as resource-constrained project scheduling problem, and its purpose is to find out the scheduling scheme with the shortest construction period (or the lowest cost). A complete engineering project is composed of many works with the predecessor and successor, and the determination of the scheduling scheme must meet the predecessor and successor constraints among all the project, as well as the upper limit of engineering project resources [19, 20].

In the assembly stage of prefabricated building construction, hoisting of prefabricated components is the main work, which should be assumed that all resources are available before the construction of the standard floor [21]; the hoisting components are transported to the construction site after being manufactured and qualified by the prefabrication factory, which is different from the construction of cast-in-place structures. The assembly stage is less affected by the environment and climate, and the construction of the standard layer is repetitive; so, it can be assumed that the construction efficiency of each layer is the same, and the assembly period can be expressed by the product of the assembly time of each layer and the number of assembly layers.

Therefore, the establishment of the BIM model in this paper is based on the following hypothesis:

All activities and required resources are ready before the start of the project;

Assuming that a project is made up of n+2 tasks, that is, V = { 1, … , n+2}, in which activities 1 and n+2 represent the tasks at the beginning and the end, respectively, which only represent the logical relationship of the tasks. There are a limited number of renewable resources in these activities, of which the maximum capacity of the kth resource is R_k (k = 1, 2, … , R). In addition, at any time, the sum of the usage of the kth renewable resource by various activities cannot exceed its upper capacity limit R_k.

The time required in activities i of the project is d_i(i = 1,2,…, n + 2); the amount of consumption of the kth renewable resource is r_k, and it is necessary to reasonably schedule the order of each work under the condition of meeting the constraints of time and resources, so as to obtain the scheduling plan with the shortest assembly period S_n+2 = Σd_i.

First, the objective function of the minimum chemical period is established: (1) minSn+2 \min S_{n+2}

Then, Eq. (2) is used to express the processor activity constraint relation of activities i and j: (2) st.Sj+dj≤Si,SSi∈P(i) st.S_{j}+d_{j}\leq{}S_{i,S}S_i\in{}P(i)

Eq. (3) represents the resource constraint of activities, that is, the total demand for renewable resources k by all activities at any time shall not exceed its upper limit R_k: (3) ∑i∈A(t)rik≤Rk \sum_{i\in{}A(t)}\mkern1mu{}\mkern1mu{}r_{ik}\leq{}R_k

Eq. (4) indicates that the time is a discrete value; (4) t=0,1,…,Sn+2 t=0,1,\ldots{},Sn+2

According to Eq. (5), the start time of all construction activities is not negative: (5) Si≥0,i=1,2,…,n+2 Si\geq{}0,i=1,2,\ldots{},n+2

Among them, p(i) is the collection of all processor activities in activity j; T (t) is the set of all activities that are going on at time t, and r_ik is the amount of resource k that activity i uses at time t.

The structural model of this project is established by Revit software, and the manufacturing process is shown in Figure 3.

Fig. 3

The specific steps are as follows:

Click ‘add new project’ to create a template file of project. Usually, the system will create the template file that Revit comes with by default. If there is no special requirement, the default template file can be used.

According to the characteristics of the model, the natural number coding method is adopted to uniformly number all processes in the construction and assembly of the standard floor with natural numbers. If there are m processes, each process is numbered as 1, 2, 3, … , m according to technological constraints. The length of chromosome is m, and the number of procedure 1~m represents the gene value, where the information expressed by each gene is the sequence of processes corresponding to the current gene.

Take the order of 2-1-7-4-3-6-5 as an example; in the case of resource constraint, it is assumed that process 1 and process 7 need the same resource R1. If they can meet the constraint conditions of R1 resources at the same time, then process 1 and process 7 can be constructed in the meantime. If the constraint conditions of R1 resources are not met, the activity of process 1 should be implemented first, and after the activity of process 1 is finished, the activity of process 7 can be carried out.

To make the fitness value of each chromosome indicate the quality of the solution, transform the minimisation target into the maximisation target by Eq. (6): (6) Fit=1F(x) \text{Fit}=\frac{1}{F(x)} where F(x) is the minimum completion time, and according to the relationship of processor, successor and resource, different scheduling schemes are generated with sufficient constraints, that is, different sequencing methods of processes.

Acquisition of fitness value: If the 1/F(x) value of the procedure is greater than the current 1/F(x), update the value of 1/F(x) until the maximum value of 1/F(x) is obtained after the processing of all the procedures in the chromosome sequence.

Roulette wheel strategy: assuming that there is N individual, the fitness value of each individual (x_i) is calculated, and then, the probability of each fitness value P(x_i) is calculated according to Eq. (7), design one and the number of individuals N Roulette wheel of the same sector, sector area and P(x_i) Is proportional to the value, where a random number p ∈ [0,1] is generated. If the p value falls into the position pointed to by a certain sector interval, the corresponding individuals in this area are selected to be inherited to the next generation, and the number of replacement crossover variation individuals is calculated by generation gap probability. (7) P(xi)=f(xi)∑j=1Nf(xi) P\left( x_i\right) =\frac{f\left( x_i\right)}{\sum_{j=1}^{N} f\left( x_i\right)}

Partial-mapped crossover method is adopted. The process of crossing is shown in Figure 4. First, select the parent chromosome from the population; second, generate a chromosome length satisfying 0 ≤ k₁ ≤ k₂ and cut the chromosome into two segments; then, the positions of the captured two segments are exchanged, and finally, the conflict of coding repetition is solved.

Fig. 4

MATLAB 2017a is used to program, and the genetic algorithm is used to solve the model, and the population size NP = 80 is set. maxgen = 200; Pc = 0.8; Pm = 0.2; generation gap probability Pe = 0.9; after 130 iterations, the shortest construction period is 8 days. Within the 8d construction period of a standard floor, arrange tasks of daily construction according to resource constraints, where the work in the first day is arranged for measuring unreeling, binding of cast-in-place steel bars, hoisting of external wall, formwork support of composite beams and plates of Unit 1 and measuring unreeling of Unit 2. The work schedule of every day is clear that the construction of a standard layer is carried out every 8 days, until the construction of all assembly floors of the project is completed.

Set the population size NP = 50; maxgen = 200; Pc = 0.8; Pm = 0.2. After 200 iterations, the optimal solution of the calculated construction period and cost is obtained, and its distribution decreases nonlinearly. In addition, there is an inverse relationship between the target of construction period and the cost target, where the shorter the construction period, the higher the cost. If managers want to shorten the manufacturing time of prefabricated components, they need to introduce high resource cost. Otherwise, they need to extend the time of producing. The corresponding values of objective function are shown in Table 1:

Solution 1 has the optimal duration that d1=5.5d, but the highest resource cost is c1=845,657. The optimal resource cost is solution 8 with c14=61,766 yuan, but its duration of construction is d14=6.9d, which is the longest. Different solutions in the optimal solution set correspond to different construction periods and resource costs, and decision makers need to choose the optimal solution according to the preference of construction period and resource costs.

To verify the rationality of the model optimisation, the Naviswork is used to check the collision of the optimised structural model and simulate the construction process. Accurately locate the location of collision through collision inspection and timely solve the design blind spots that are not easy to find, thus effectively improving the design quality and the work efficiency of designers. Using simulation of construction can solve the problems and contradictions that may occur in the actual construction as early as possible, so as to reduce risks, save costs and shorten the construction period. The simulation process of construction in this project is shown in Figure 5.

Fig. 5

In the process of simulation, with the help of virtual technology, the simulation of technological process and the refined presentation of some special engineering components are realised, and compared with real objects, managers can optimise the construction plan to the maximum extent, which is of guiding significance to the node control of the constructor and the project management of the owner. In addition, as a bridge connecting the design and construction stages, the construction simulation function will play a great role in the disclosure work in the design and construction stages. The constructors can intuitively feel the process of design and construction in each component through the simulation test, effectively improve the dullness brought by traditional drawings and guide the work of construction accurately.