Research on the New Construction Industrialization Platform Based on Computing Technology

Building information modelling (BIM) is based on building project drawings, scaled down to a certain proportion, and then establishing three-dimensional graphics of construction projects, using advanced information technology, the combination of reality and reality, to establish a consistent building information base for the actual building project [22]. BIM contains not only the geometric contours of the building components and component attributes, but also the component objects of the reinforcement, costing, and other information. Intuitively seeing the 3D information model of the building, integrating architecture, structure and costing into one, it greatly improves the degree of information integration of construction projects and provides a platform for information sharing and exchange. Applying BIM technology to the industrialization architecture of new buildings is a major trend in the development of new construction industrialization and integration.

For new construction industrialization, the BIM model alone is not sufficient.The BIM model only provides the basis and conditions for building industrialization, but cannot provide real-time dynamic data of the building during operation and maintenance. Through the introduction of Internet of Things technology, to open the data barriers, the BIM model and the real-time dynamic data of the equipment in the building to bind one by one, to achieve real-time operation and maintenance control of the building. In the intelligent manufacturing of building industrialization, digital twin technology is the integration of real-time dynamic data with the BIM model, which is reflected in the accurate mapping of the BIM model with the facilities and equipment and the integration and sharing of building dynamic data.

The four-block sampling method is to divide the building panels into four rectangular blocks with one parent dividing line and two child dividing lines, to discharge strips of the same direction and length in each block, and to discharge the same kind of rectangular pieces in the strips. According to whether or not there is a constraint on the number of each rectangular piece allowed to appear in the sampling method, the corresponding sampling algorithms can be divided into unconstrained sampling algorithms and constrained sampling algorithms [24]. The unconstrained sampling algorithm has no constraints on the number of rectangular pieces allowed to appear in the sampling method, and the constrained sampling algorithm limits the number of rectangular pieces allowed to appear in the sampling method to no more than its upper bound. In this paper, we design the constrained rectangular piece sampling algorithm as follows:

Assume that the upper bound on the number of rectangular pieces of type i is b_i , where i∈{1,⋯,m}. When b_i is taken to be a sufficiently large number, the constrained nesting algorithm can be morphed into an unconstrained nesting algorithm. Let the horizontal edges of plates, blocks and strips be lengths and the vertical edges be widths. It is useful to set the dimensions of plates and rectangular pieces to integers, which does not affect the generality of the algorithm in this paper, because when the dimensions are not integers, the plates and rectangular pieces can be scaled up to integers according to the scale.

Let v_i be the value of rectangular pieces of type i, s_X(i,x) be the value of horizontal strip x×w_i, and n_x(i,x) be the number of rectangular pieces of type i contained in horizontal strip x×w_i. Then there are: (1) nX(i,x)=⌊ x/li ⌋ (2) sX(i,x)=min(nX(i,x)bi)vi

For vertical strip l_i×y, the value s_Y(i,y) and the number of rectangular pieces contained in it n_y(i,y) can be found in the same way.

For block x×y (length: x, width: y , x≤L, y≤W), let the value of the block be f_x(x,y) when the horizontal strip is discharged from the block, and the number of strips x×w_i contained in the block be g_x(i,x), then we have: (3) fX(x,y)=max(∑i=1msX(i,x)gX(i,x)) (4) s.t.{ ∑i=1mgx(i,x)wi≤y,gX(i,x)∈Ngx(i,x)nX(i,x)≤bi

The above model is a bounded knapsack problem, which is solved by counting the number of rectangular pieces of each type contained in block x×y, and noting that the number of rectangular pieces of type i is n_X(i,x,y). The same can be done to find the value of the block when the block is discharged with vertical strips, f_Y(x,y), and the number of rectangular pieces of type i contained in the block, n_Y(i,x,y). Let the value of the optimal block be f(x,y) and the number of rectangular pieces of type i contained in the block be n(i,x,y); if f_X(x,y)>f_Y(x,y) then the block emits horizontal strips and f(x,y) = f_X(x,y), n(i,x,y) = n_X(i,x,y), otherwise the block emits vertical strips and f(x,y) = f_Y(x,y), n(i,x,y) = n_Y(i,x,y).

Take the lower left corner of the plate as the coordinate origin, the horizontal edge as X axis, and the vertical edge as Y axes to establish a right-angle coordinate system. Set four pieces of row sample way of the parent dividing line coordinates for x, two sub-dividing line coordinates were y₁, y₂, of which x, y₁, y₂ are integers. Let the value of the four pieces of sampling method be V: (5) V=max0≤x≤L{ max0≤y1≤W[ f(x,y1)+f(x,W−y1) ]+max0≤y2≤W[ f(L−x,y2)+f(L−x,W−y2) ]}s.t.n(i,x,y1)+n(i,x,W−y1)+n(i,L−x,y2)+n(i,L−x,W−y2)≤bi,i∈{ 1,⋯,m }

The objective function of model (5) is to maximise the value of the scheduling approach and the constraint is that the number of each rectangular piece in the scheduling approach does not exceed its upper bound.

In order to achieve the maximum utilisation of building panels and relatively high efficiency in the process of rectangular piece scheduling, this paper proposes an iterative neighbourhood search algorithm for solving the optimal discharge order of small rectangles and their scheduling scheme. The discharge order of rectangular pieces is regarded as a certain arrangement of rectangular piece numbers, which is used as an encoding, and the blank rectangle filling algorithm is used to decode the sampling scheme and the sampling utilisation rate [25].

Denote a₁a₂⋯a_n as some arrangement of small rectangles numbered from 1 to n, i.e., the individual code, which indicates the order of the to-be-emitted, where a_i denotes the small rectangle to be emitted for the i rd time R_ai. In the iterative neighbourhood search algorithm, two types of neighbourhood operators are used, the exchange operator Swap(i,j) and the insertion operator Insert(i,j) based on the random two-points (i,j). Swap(i,j) denotes exchanging a_i with a_j, and the other positional elements remain unchanged, i.e., a₁a₂⋯a_i⋯a_j⋯a_n becomes a₁a₂⋯a_j⋯a_i⋯a_n. Insert(i,j) denotes inserting a_i into the original a_j’s position, and the relative positions of the other small rectangles remain unchanged. At i<j, a₁a₂⋯a_i⋯a_j⋯a_n becomes a₁a₂⋯a_i–1a_i+1⋯a_ja_ia_j+1⋯a_n and at i>j, a₁a₂⋯a_j⋯a_i⋯a_n becomes a₁a₂⋯a_ia_ja_j+1⋯a_i–1a_i+1⋯a_n.

The process of the iterative neighbourhood search algorithm is to use the above two kinds of neighbourhood operators to continuously iterate, alternately searching the solution space, i.e., after a certain number of consecutive searches by one kind of neighbourhood operator (known as the maximum number of local searches lsNum), another kind of neighbourhood search operator will be used to continue searching for the solution consecutively. At the same time, in order to reduce the invalid search and prevent falling into the local optimal solution, when the number of times that a kind of operator can not improve the current optimal solution in a continuous search reaches the upper limit of the number of times nimprvNum , it will also be replaced with another kind of neighbourhood operator to continue to search for the solution.

In the iterative neighbourhood search algorithm, in order to control the search time of the algorithm, a global maximum number of searches gsNum is set, and the whole algorithm is terminated when the total number of neighbourhood operators executed is greater than gsNum, outputting the optimal solution and the ranked sample utilisation rate.

Notations x₀ and x_b denote the initial solution, the current best solution, respectively, and logistic variable SwapOpr = True denotes the choice of exchange operator Swap(i,j) and logistic variable SwapOpr = Flase denotes the choice of insertion operator Insert(i,j).

“Random selection of operator execution point (i,j)” i.e. selection of two points i and j for exchange operator Swap(i,j) or insertion operator Insert(i,j). Global search termination is based on the number of total neighbourhood searches exceeding gsNum or the current sampling utilisation reaching 100%. The local search is terminated when the number of local searches exceeds lsNum or when the number of consecutive searches using a certain operator fails to improve the optimal solution reaches nimprvNum.

Structural construction process is a dynamic and complex system, combined with the characteristics of structural construction safety risk management and application needs, define the digital twin five-dimensional model for structural construction safety analysis, as shown in Equation (6), and propose the digital twin framework for intelligent analysis of structural construction safety as shown in Figure 3.

Figure 3.

Structure construction safety analysis framework

Where PH is the physical construction site, VH is the virtual construction site, Rs is the construction safety intelligent analysis system, DD is the twin data service platform, and CN is the connection between the parts. The connections between different elements in the framework have different meanings, and CN is the general term for connections with different meanings.

The physical construction site is the data source that drives the operation of the digital twin model and is the basis for building the corresponding virtual digital twin model [26]. According to the construction safety risk indicator system, the task of physical construction space modelling is to extract the physical entity information related to the safety risk factors by configuring the information collection and transmission devices. Specifically, the physical elements of the construction site are operators, tower crane equipment, prefabricated components, and the site environment.In addition, the level of safety risk management needs to be measured by the policy and institutional situation on site. Therefore, the physical construction site model is described as: (7) PHM={ PP,PM,PC,PE,PS } Where PHM represents the physical construction site model, PP represents the physical construction site operator, PM represents the physical construction site equipment, PC represents the physical construction site key components, PE represents the physical construction site environment, and PS represents the physical construction site management level.

The prerequisite for realising the intelligent assessment of safety in the construction process is to ensure the dynamic perception of the virtual and the real. In the real construction environment, the first step is to arrange the sensors to monitor the structural performance, and the fusion of the measured data from the sensors and the simulated data from the twin model drives the intelligent diagnosis and scientific prediction of the structural safety performance. However, the sensor arrangement is optimised to facilitate data transmission and processing at a later stage. The target parameters measured by the sensors are set as: (8) z=g(x1,x2⋯,,xn) Where x_i is the monitoring value collected by the sensor directly on the structure, and defining δ₁,δ₂⋯δ_n as the error obtained x₁,x₂⋯,x_n on the sensor side, then Δz is the z error caused by δ₁,δ₂⋯δ_n. Therefore: (9) z+Δz=g(x1+δ1,x2+δ2⋯,xn+δn) is obtained from the Taylor expansion: (10) Δz=∂g∂x1δ1+∂g∂x2δ2+⋯+∂g∂xnδn

Then the maximum error is: (11) Δzmax=±(| ∂g∂x1δ1 |+| ∂g∂x2δ2 |+⋯+| ∂g∂xnδn |)

The objective function of sensor arrangement can be derived from the maximum error, i.e. Y = Min(Δz_max). According to the experience of engineering arrangement of sensors in space structure, the sensors are generally arranged in the critical parts of the force and the most sensitive positions to damage.

In addition, according to the structural reliability analysis, the guaranteed structural safety performance should comply with R – S ≥ 0, where S is the action effect of the structure and R is the resistance of the structure. However, during the construction process, the structural performance is easily disturbed by various uncertainty disturbing factors, and the operation of the construction system also has significant dynamic and non-linear characteristics. Therefore, the structural effect and resistance are changing, and it should be ensured that the safety performance is assessed by R(t) – S(t) ≥ 0. The intelligent control centre requires real-time analysis of the whole tensioning process to establish the time function (S(t)) of the structural action effect and the time function (R(t)) of the structural resistance. It is necessary to mine the historical information of the construction process, real-time information is collected and simulated to get the future trend, and a three-dimensional integrated data model is established, namely: (12) D=(Da,Db,Dc)T

In turn, the safety performance of the structure is intelligently analysed in real time.

For the verification results of building construction safety, this paper proposes the verification evaluation coefficient EV value, by which the EV value can not only indicate whether the recognition effect is correct or not, but also indicate the stability of recognition. It is expressed as: (13) EV=EVt−max[EVf]max[EVt,EVf],EV∈[−1,1] Where EV_t represents the EV value of the true key building block that is tagged, and the value is represented separately when there are multiple key building blocks. EV_f denotes the EV value of each of the other components that are not tagged. Obviously, when EV<0, it means that the indicator cannot identify the tagged key component, and when EV>0, it means that the indicator can correctly identify the tagged key component, and the larger the value of EV means the better the identification and the stronger the pointing.

The purpose of designing the optimisation algorithm for rectangular piece nesting in a new construction industrialization platform is to improve the intelligent manufacturing efficiency as well as the utilisation of building panels during the construction of rectangular pieces. For this purpose, this paper splits the large rectangle of 150*150 into multiple small rectangles of the same size, and there are height lower bounds for both strip width and optimal solution, which will thus be used as a perfect data set.

In this paper, a constructive heuristic algorithm (BFP) is chosen as the comparison algorithm. Both algorithms have the same parameter settings, and the algorithm implementations are implemented on the same machine configuration using Python language.On the constructed perfect dataset, 100 tests are performed on different numbers of instances with different numbers of rectangles, and both algorithms are evaluated by strip utilisation and running time. The results are shown in Table 1; the best results are highlighted in bold, and the average performance of all the instances is given in the last row. Figure 4 shows the utilization box-and-line plots for the two algorithms.

Figure 4.

The utilization ratio of the two algorithms

Specifically, for instances with 10 rectangles, both algorithms can achieve an optimal utilisation rate of 1.0, mainly due to the fact that the optimal solution is also easily selected by the greedy rule with a small solution space. The average and worst utilisation of the algorithms in this paper are 0.908 and 0.737, which are 7.08% and 15.52% higher than the BFP algorithm, and the standard deviation is also much lower than that of BFP. With the increase in the number of rectangular pieces, this advantage will be more obvious, which shows that this paper’s algorithm provides better scheduling utilisation and stability for the intelligent manufacturing of rectangular pieces for new construction industrialization buildings. From the point of view of runtime efficiency, the average time for this paper’s algorithm to optimize rectangular parts sampling is 25.18ms, which is about 1.7 times longer than that of the BFP algorithm. This is because the preprocessing of the packing state and the model inference are needed for the optimisation of the rectangular part sampling, which consumes more time than the BFP algorithm, but the results obtained are better than the BFP algorithm. Therefore, from a comprehensive point of view, this paper combines the iterative neighbourhood search optimization algorithm with the rectangular piece sampling algorithm with certain feasibility and efficiency, which can provide a reference for the intelligent construction and efficiency optimization of the new building industrialization.

In order to verify the application feasibility of the iterative neighbourhood search algorithm proposed in this paper in the optimization of rectangular pieces, the rectangular pieces of a building materials steel structure processing plant are selected as an example to carry out an instance analysis. The processing plant needs to cut a batch of rectangular plates for the assembly of steel components because the project requires a large number of components, a variety of types, so from the order selected a batch of 12 types of each of the 10 pieces, a total of 120 rectangular pieces as the experimental object. Table 2 displays the size and number of rectangular pieces, with the raw material plate being 2000mm wide and not limited in height.

Optimisation is carried out using Genetic Algorithm (GA), Tent-Genetic Algorithm (TGA), Tent-Genetic Annealing Algorithm (TGSA), Grey Wolf Algorithm (GWO), Tent-Grey Wolf Algorithm (GWO), Discrete Sparrow Search Algorithm (DSSA), Tent-Discrete Sparrow Search Algorithm (TDSSA), and this paper’s algorithm, respectively, and the optimisation results will be comparatively analysed to Verify the effectiveness of this paper’s algorithm in optimising the sampling of rectangular pieces when there are more types. The parameter settings of the algorithms used for the comparison experiments are identical.In order to prevent the influence of algorithmic chance, each algorithm is run 50 times and compared to the raw material utilization rate and scheduling time after optimization. Table 3 shows the comparison results of optimal scheduling for different algorithms.

Based on the iterative neighbourhood search algorithm, the utilisation rate of building materials rectangular plate sampling is 96.79%, and the sampling time is 489.5s, which is 4.28% higher than that of the TDSSA algorithm, which is the second-best performance, and 37.1s increase the scheduling time. Therefore, the algorithm of the present paper is better than GA, TGA, TGSA, GWO, and TGWO in terms of utilisation rate and scheduling time, even though there is a slight increase in scheduling time compared to the TDSSA algorithm. Although there is a slight increase in sampling time compared to TDSSA, it is significantly better than TDSSA in terms of sampling utilisation. Therefore, the neighbourhood iterative search optimisation algorithm proposed in this paper maintains high stability in optimising the sampling problem for a large number of rectangular pieces, and its application to the new construction industrialization platform can realise the intelligent sampling optimisation of the building panels and provide technical support for the saving of building material costs.

The method proposed in this paper aims to improve the safety of construction by dynamically identifying critical components during the construction process, so the following example is designed to verify whether the method can accurately identify the critical components during the construction process. The validation method is to simulate the damage of a member by changing the cross-sectional area and modulus of elasticity. In a certain construction phase, the cross-sectional area and modulus of elasticity of one or several members are reduced directionally while all other conditions remain unchanged, and the member whose parameters have been changed will definitely be the damaged member, which is also the critical member in that phase. Then, the method of this paper is fully automated to identify these members, and if the construction safety intelligent analysis framework supported by Digital Twin can accurately identify these members, it means that the method of this paper is feasible.

The steel roof cover of a convention and exhibition project in T city, Fujian province is used as the research object, and one of the planar chord-bearing structures is selected as the computational model. The span is 50m, the droop is 3.5m, and the number of braces is 10. The top chord beam is composed of H-beam Q355B, while the spacers are composed of round steel pipe Q245B, and the lower chord cable is composed of prestressing strands with high strength and low resistance. The boundary condition of this string-supported beam structure has a hinged bearing at one end, and a sliding bearing at the other end that releases the restraint along the long axis to release the temperature action, so the effect of temperature action can be disregarded in the analysis. It is assumed that the roof slab system has not fully formed at this stage of the analysis, so the wind load effect is not considered at this time.The unit numbering of the chord-supported beam structure and the load arrangement for validation calculations are shown in Fig. 5.

Figure 5.

Verification example component number diagram

Based on the intelligent analysis method of structural construction safety established by the integration of digital twin technology in the previous section, the key components under different working conditions in the construction process can be accurately identified from the eigenvalue indexes of stress intensity coefficient (SI), stress variation coefficient (SV), displacement variation coefficient (DV), ground state key coefficient (GK), dynamic key coefficient (DK), etc., respectively.

In the analysis process, the structural self-weight condition is taken as the base state of the structure, and the construction process arrangement of the roof (i.e., the structural arrangement of the constant load) is taken as the state of the structure at the first moment of the analysis. According to the load code, the constant load in this study is based on 120kg/m². Since the construction process requires temporary construction loads of 60kg/m² to be temporarily arranged in the half-span area, this phase is used as the first moment to analyse the state of the structure. Firstly, the twin data before and after this stage are generated based on the twin model using finite element software calculations, and then the twin data and the sensed data on-site are input into the data processing centre together, and a series of calculations such as eigenvalue extraction is carried out automatically by the D-S fusion algorithm. And the results of fusion decision-making are output to the application management platform, where managers or operators are able to visually and dynamically view the key components analysed and identified, thus realizing the intelligent management of construction safety.

Figure 6 shows the identification effect of each eigenvalue index on the key stress components under the action of temporary construction load. Through the analysis results, it can be found that the T24 and T23 units of the upper steel beam corresponding to the chord support beam under the condition of half-span temporary construction load are the key safety components, that is, the variable cross-section of the steel beam is the most unfavourable position for safety under this condition, which needs to be paid attention to. The EV values calculated by the D-S fusion algorithm in the digital twin intelligent analysis framework are 0.914 and 0.758, respectively, which show very obvious directionality. In addition, T33 and T17 member cells are also highly critical, which suggests that the tension cable deformation in the span is another potentially critical member. Although a single eigenvalue can also present this result to a certain extent, it can be found from the distribution of SI~DK that the accuracy and directionality of a single eigenvalue are far less than the result of data fusion judgement by digital twin, which indicates that the introduction of digital twin technology for data fusion in the new construction industrialization platform can effectively improve the utilisation rate of the data and the relevance and efficiency of decision-making.

Figure 6.

Security key component recognition results

In summary, the key safety components under temporary construction loads are the variable section of the steel beams and the mid-span tension cables, respectively, and attention should be paid to the monitoring and maintenance of the corresponding steel beams and tension cable components and the necessary stress prediction during this construction process. In fact, similar conclusions can be drawn based on professional mechanical analyses and test results, but this method simplifies the complex professional mechanical analysis process using the new construction industrialization platform. It also shows that the new construction industrialization platform constructed in this paper can provide a scientific decision-making basis for managers without the need for professional experience, reduce the threshold of application of professional technology, improve the degree of intelligence of construction safety analysis, and provide platform support for promoting the high-quality development of new construction industrialization.

To achieve dynamic cost management of construction, it is necessary to manage and monitor progress.On the one hand, effective control of the construction progress can effectively save the construction period, thus reducing the cost of labour, machinery and equipment, and at the same time, can be rewarded for early completion. On the other hand, increasing the progress of the dimensions of the control can be integrated to determine the resource input programme, and real-time control of the amount of resource input to reduce the waste of resources. For a bank in Fujian Province, for example, a comprehensive service building project, the use of this paper establishes a new construction industrialization platform to carry out project cost management analysis. Because manpower, materials, and machinery costs in the project construction account for a large amount, Table 4 shows the specific application of benefits as an example.

In terms of front-line construction personnel costs, the total overrun was 10053658 yuan, mainly because the actual labor unit price was higher than the quota and the cumulative labor price difference was higher than the cumulative saving amount per working day, but the fixed number of working hours was reduced from 392417 to 284152. Complete information on labor personnel, improve management efficiency, shorten the time of personnel counting, and improve the efficiency of labor resource utilization. In addition, the project team that completed the project ahead of schedule can be transferred to the new project in advance, which can save labor management costs to a certain extent, and the cumulative labor overrun is about 10 million yuan. Using the new building industrialization platform to reduce waste through strict control of the quota picking, the project is equivalent to saving 10802400 yuan in material savings and at least about 4843160 yuan in mechanical costs after the completion of the whole project. In addition to the part of manual overrun, the project carried out through the new building industrialization platform can save about 5 million yuan after completion, can effectively improve the efficiency of labor resource utilization and management, and also provide support for promoting the intelligence of the project.

One environmental monitoring terminal is arranged with the upper wall at the entrance of the curing box at the engineering site to collect the temperature and humidity information of the curing box, and another environmental monitoring terminal is arranged at the wall of the first-floor façade of the construction site to collect the environmental information of the construction site. The results of the temperature and humidity monitoring of the engineering equipment are shown in Figure 7. The real-time equipment information collection function of the new construction industrialization platform helps project managers quickly understand the temperature and humidity changes of the engineering equipment and the environmental changes of the construction site. Then, according to the construction site temperature and humidity, reasonable project progress adjustment fully guarantees the service life of engineering equipment and the working environment of the construction site personnel in order to enhance the efficiency of the project to lay a solid foundation.

Figure 7.

Temperature and humidity monitoring of engineering equipment