Construction cost management strategy based on BIM technology and neural network model

Abstract

The construction of construction projects is an important industry of national social and economic development, and price management control is an important part of construction projects, and has become an important factor for major construction companies in China to manage construction projects. At present, the internal construction price management is not the best, nor the most ideal. Few investments exceed the budget, mainly due to defects in effective construction price management, lack of advanced technology and lack of prospects for prepayment, which make it difficult to match the actual and expected results of construction project price management. The actual results are always unsatisfactory. In this paper, the engineering cost estimation model is studied, and the neural network comprehensive prediction model is established to improve the accuracy and application technology of the prediction model. By using the building of BIM technology and neural network model, and effectively using the price advantage of ICT, it is used in the construction industry, and the cost is strictly controlled, so as to bring huge profits to the enterprise and promote the development of the enterprise.

Keywords

BIM technology neural network model construction engineering cost management

1 Introduction

The construction of construction projects is an important part of the social and economic development of our country, which has many effects on the social development of our country, and promotes the economic development of our country through the influence of the construction industry itself and other departments. At the same time, multinational companies enter our construction market, making the construction market more intense to some extent, which forces our construction companies to control construction prices in all aspects of the construction chain during the development of the construction industry. Price management control, which is an important part of construction projects, has become an important factor in the project management of major construction companies in China. Price management is not only related to improving the competitiveness of Chinese companies in the market, but also has a significant impact on the development of Chinese companies. At present, the internal construction price management is not ideal, and the situation of investment exceeding the budget is also rare. This is mainly due to the defects of effective construction price management, the lack of advanced technology and the lack of advanced prospects, the deviation of the actual and expected results of construction engineering price management in China, the construction information model is a new science and technology, highly integrated information and digitization, which contributes to the preliminary design. The medium-term construction management and the subsequent liquidation of the project provide an effective platform for information sharing and a means of rapid communication for the construction participants.

Theoretical research shows that the application of BIM technology in building pricing practices can effectively improve the utilization rate of building resources and avoid waste and inefficient work. Provide a rapid channel for communication and communication among stakeholders and a solid foundation for programme delivery at all stages of construction. The speed and accuracy of calculation are two important indexes to evaluate the effectiveness of cost calculation method. Engineering and rapid and accurate engineering cost estimation methods are particularly important, not only to speed up the selection of project balances, but also to guide and facilitate the establishment of an engineering cost information database.

In this paper, the engineering cost estimation model is studied, and the neural network comprehensive prediction model is established to improve the accuracy and application technology of the prediction model. By using BIM technology to make effective use of its engineering price advantage and make use of these advantages in construction industry and cost management, it can bring great benefits to enterprises and promote the development of enterprises.

2 Related work

As noted in the literature [1], the complexity of the project is one of the main factors contributing to inaccurate cost estimates, and project-based industries face significant challenges in controlling costs and completing project budgets. It is therefore important to determine the complexity of the different cost estimates and to determine the cost of best practices to improve their accuracy. The results of the study show that risk, the size of projects and outputs and the time frame are complex factors affecting cost estimates, and further attention is needed to reduce the unintended cost impact. Document [2] comprehensively analyzes the factors that affect the cost of the project, and points out that project design, scientific engineering management and attention to the completed project are the three main aspects of cost control. According to the literature [3] the factors that affect the project price include bidding factors, design factors, building factors, market factors and natural factors, and reasonable control of engineering cost in the design stage is a key priority for scientific control of engineering cost. [4] the literature, the composition of the project cost is analyzed, and the conclusion is that the construction cost is the most important part of the project cost, and the most important is the most unstable factor, such as the design factor, the building factor and the contract factor. Document [5] the project bidding stage, the project management stage and the project measurement stage, which have a significant impact on the project price in the design stage, are regarded as a stage of the project, and they affect the whole process project. From 30% to 80% of the construction cost in the construction stage will be affected by the design stage. Therefore, the cost control in the design stage is very important to the cost control in the whole process. In this connection, the Advisory Committee notes that cost control in the engineering design phase is largely dependent on cost control in the construction phase.

Literature [6] suggests that at present, Autodesk, Bently and Graphisoft three giant software companies, different BIM software has been developed and disseminated, and in repair and modification of the growing. As a result, some progress has been made in developing basic technologies, data sharing and related standards. According to the literature [7], the technological innovation and replacement of the construction industry is mainly due to the major changes in the construction industry. Therefore, the adoption and implementation of dual-use technology can bring great benefits to engineering companies. In particular, it is recognized and disseminated that software requires relatively loose exchange standards and models so that different software can operate each other and help to incorporate model design into different software, joint models and digital and computer models to determine the final version. The literature [8] develop a method based on the IFC management model to facilitate project stakeholders to develop construction plans, use digital technology to simulate assumptions and reduce user interaction with the system.

Literature [9] analyse 55 historical sample data sets were collected from 55 completed projects and 11 project features were selected to build a multi-source neural network model and a core radiation function. respectively, they randomly sampled 45 sample data sets to train the network model, while the remaining 10 sample groups were used for empirical prediction. The results showed that the parameter-selected neural network digital prediction model could be used to estimate the engineering cost. Document [10] selected an index system that affects engineering prices, taking into account years of experience in engineering price practice and theoretical analysis of engineering price documentation system. This paper systematically analyzes the methodological basis of BP neural network and the reasons for planning engineering costs, and develops a network prediction model. Engineering case data need to be collected first to verify the simulation. The results show that the prediction of BP neural network construction cost is very accurate and rapid. Document [11] Based on 13 detailed technical studies, a model was developed to estimate the cost of building bridges. The factors that affect the cost of the superstructure of the bridge are determined, and then the variables are divided into two different types of multiple linear regression analysis (MLR) and network regression analysis. In order to obtain the best estimation model, artificial nervous system analysis shows that the accuracy of neural network is higher.

3 Characteristics of 3BIM technology and its application in construction cost management

3.1 BIM Technology connotation

IFrom the technical point of view, BIM technology depends on computer three-dimensional digital technology, which makes computer digital. The computerization and full expression of construction entities and the provision of accurate information and data on construction works, as innovative modeling tools in three aspects of construction engineering, strong data storage and processing capabilities will provide it that will lead to significant changes in engineering projects, the benefits of which will promote further modernization of industry and thus compensate for the shortcomings of traditional engineering construction costs, which are not limited to the budget and settlement process, but include all phases and projects within the system, involving a new form of communication and cooperation between the parties.

The BIM model is a complete description of the entire project process, covering data, processes and resources at all stages of the project’s life and facilitating universal application by project participants. Project data technology allows collection, correction, data processing and exchange throughout the project period, enabling coordination and exchange of distributed data between projects. Schematic diagram of BIM model as show in Fig. 1, BIM software schematic as show in Fig. 2, BIM model involvement as show in Fig. 3.

Fig.1

Schematic diagram of BIM model in engineering phase.

Fig.2

BIM software schematic.

Fig.3

BIM model involvement.

BIM main features of the model are as follows:

(1) Integrity: that is, whether the project information is complete. BIM model contains all information about project design and construction, maintenance information, etc., and fully shows the relationship between the components of 3D space.

(2) Relevance: The components of the BIM model are not independent of each other, but interrelated, with changes in one part and changes in the other.

(3) Unity: on the one hand, the BIM data of a project are unified; on the other hand, the data are analysed and unified with computer intelligence at all stages BIM the project.

3.2 Application of BIM technology in cost management

3.2.1 Building data accumulation

The accumulation of building information is crucial for a price consulting firm and is one of the core values of its good construction engineering. However, the current emphasis on this issue is quite common and relies heavily on experienced engineers to accumulate data and perform various cost analyses. However, the high turnover rate of company staff creates great pressure. The loss of empirical data can cause irreparable damage to the company. In the current information society, the Prime Minister has vigorously promoted the Internet+. It also provides us with an opportunity to enable us to effectively deal with the difficulties of past projects, such as a large amount of information and data analysis, which leads to the effective accumulation and analysis of information. The implementation of the project creates an environment for information exchange among participants, which will not only provide a three-dimensional project model, but also allow the storage of classified data and the free flow and communication between participants. Therefore, the congestion of information and engineering is avoided, it is convenient for professional staff to update and modify information data in time, and the delay in updating information is avoided, which affects the access to information. According to the project or time stage review BIM the relevant information stored in the specific project process. After completion of the project, staff members have direct access to all available information on the project, audit and documentation using BIM software. Corresponding BIM powerful data storage and analysis functions facilitate data storage and processing of engineering projects and enable staff to engage in complex data processing. Diagram of the quota design process as show in Fig. 4, Flowchart of cost management as show in Fig. 5.

Fig.4

Diagram of the quota design process.

Fig.5

Flowchart of cost management.

3.2.2 Realization of whole process cost management

BIM achieved real price management throughout the fiscal year. According to the data of the BIM model, the approximate value of the project can be obtained, and the indexes such as unit price can be taken into account to obtain the estimate of the project. According to the data provided by the BIM model, the project stage budget problem has the project component and the project quantity index, and the estimated value of the project is obtained by reviewing the project standard index. Differences in programme design lead to different budget estimates. This method is more detailed than the budget of the construction drawing and the data of the double cylinder model developed at the present stage, and the workload is more accurate, which provides accurate information for the construction drawing. Based on the B IM model, this stage can obtain a complete list of work, show all the components listed in the list and avoid calculation errors. The bidder compared the amount of work with the checklist in order to lay the foundation for the smooth progress of the bidding process–the signing stage of the contract: to develop an original BIM model containing contract information. According to the data of the BIM model and the comparison with the signed contract, and can be adjusted according to the model during the execution of the project to facilitate the timely modification of the contract and liquidation. Post-stage architectural design: BIM model includes component changes and information, financial statement change phase liquidation that is immediately registered in the IASB system and is the basis for the audit of the financial statements: a BIM model based on data for all phases of the above-mentioned project, which contains information various stages of the process and is classified according to the actual conditions of the project so that the end process can be successfully passed. BIM Application of cost management as show in Table 1, Main work of participants as show in Table 2.

Table 1
BIM Application of cost management at all stages of the project

Investment decision-making Planning design Bidding Construction phase Completion phase

Programme selection Budget estimates Precise list Cost analysis Settlement

Limit design Unbalanced quotation Human, Material, Machine Analysis

Quota Collection

Change management

Payment management

Investment decision-making	Planning design	Bidding	Construction phase	Completion phase
Programme selection	Budget estimates	Precise list	Cost analysis	Settlement
	Limit design	Unbalanced quotation	Human, Material, Machine Analysis
			Quota Collection
			Change management
			Payment management

Table 2

Main work of participants

Owner	Designers	Construction	Cost personnel
Control costs	Simulation programmes, decision-making	Contract, Payment Settlement	Reduction in manual computing
Project change	Quick budget estimates, budget	Construction cost control	Investment planning
Visa adjustments		Improved management	Contract management
Completion settlement			Claims, cost analysis

3.2.3 BIM based engineering calculations

The bid control price for the preparation of the project budget is determined according to the calculation of the amount of work. And the function of calculating cost is more advantageous than traditional means to reduce the influence of manual calculation of engineering quantity and obtain more data. Reliable statistical analysis, engineering volume statistics, etc., BIM model helps builders to better understand the total amount of bids and help staff work.

The accuracy of preparing bid price for bid control has a positive impact, which improves the efficiency and accuracy of unit building bidding, and provides basic information for consultation and subsequent construction.

Figure 6 shows the relationship between design and measurement software, and many national software companies related to engineering and pricing are making every effort to develop data interfaces for the relevant models to achieve software design. According to the national standard engineering on design and construction interface, from design to quantification, this technology is increasingly accepted by engineers.

Fig.6

Flowchart of software introduction and calculation.

Because of the boundary design and collision check in the design stage, the design work has been further improved and more feasible. By preparing workload documents and managing bidding prices, construction staff can better achieve their objectives, including construction plans. Request for proposals:

(1) Number of models. The model is based on the map of the building request for proposals, which includes the above design software, redesigns the measurement model, and finally determines the parameters for the measurement model containing spatial, physical and other information.

(2) Design parameters: as one of the conditions for automatic application of concrete structural components, it is necessary to use the strength parameters of concrete structures as the basis for calculation and excavation.

(3) The concept of inventory and fixed quantity engineering is critical and is based on different types of measurement models: concrete, bridge slabs, piers, foundations, etc. can be applied automatically. Fine can be achieved by manually inputting impossible things.

(4) The BIM model facilitates the calculation of aggregate quantities and automatically lists the required quantities and bid prices. Due to the technical convenience of the BIM, basic works, such as the calculation of engineering quantities, are carried out electronically. can be completed quickly, and the time saved can be used best, thus helping to comply with the recommended working conditions and contract conditions, and helping the construction unit to better complete the task.

3.3 Title page

BIM application of technology has a great impact on the work of all parties involved in the construction, and has changed to varying degrees.

As far as owners are concerned, the development of BIM models before the implementation of the project, the use of digital analog technology to detect the technical and cost problems of the project early, pre-planning and cost control; the development of a progress plan and its integration into the dual-service trade union information system can not only ensure the normal progress of the project, but also make timely modifications to avoid duplication of work. During the liquidation phase of the project, that is, the BIM model, the materials can be accurately counted by manpower, mechanical equipment, etc., and the efficiency of liquidation between the parties can be improved.

The task of the designers is to assist the owners in making decisions and, in the Maritime Council, to use the model simulation function of the powerful Maritime Council to simulate various design options, select the best options and refer to the options of the United Nations system organizations. At the same time, according to the information provided by the BIM model, the rapid estimation of engineering quantity and cost and the model of budget are compiled, and the foundation for bidding is laid.

In terms of contracts, the builder may, on the basis of information provided by the Maritime Council, establish a contract-type two-cylinder model that accurately describes the content of the construction cost of the building unit related to its own unit to ensure that it can be paid promptly. After the construction and cleaning of all stages of the project, reduce the size of the project and reduce errors. For cost control, by accurately classifying and processing the data, and strictly controlling the engineering materials according to the information in the BIM model, controlling the cost, in order to ensure that the building materials are not wasted and the construction process is orderly.

For developers, the use of international standard industry classification standards, so that developers get rid of the time-consuming calculation, this complex machine calculation replaced manual calculation, and fundamentally changed the developer’s working methods, no longer focus on quantitative calculation, but on investment planning, contract management, claims, cost analysis and so on.

4 Construction of cost prediction model based on neural network

4.1 Determination of network topology

BP main feature of neural network is a multi-level feedback network, which is a system of reverse learning, which is a neural network model and is widely used by departments. The industrial neural network contains a large number of cartographic relationships, and its main learning rule is the maximum deceleration, which is achieved by continuous learning. The adjustment of the reverse internal weight and the network valve value finally minimizes the error of the entire neural network.

The number of basic neurons has a significant impact on the expected performance of the network, because too few neurons will reduce the accuracy of the network and reduce the ability to analyze nonlinear laws, and too many neurons may lead to overestimation of the accuracy of neurons. Increase network computing load and reduce network operation efficiency. Rules for determining the number of underground nodes: if the function to be adjusted is extremely unstable and the connection weight needs to be increased, the number of underground nodes should also increase; if high accuracy is required, the number of underground nodes should also increase; In the aggregation process, a small number of nodes can be added or phased out at the beginning. The research work carried out by scholars according to these standards is usually based on empirical evidence, and the number of bottom nodes is determined according to the number of input and output points.

To a three-δ neural network, N is the subnumber of input points of N = 5, implicit node 2 NN = 1 = 11, and the number of output points depends on the objective data type of sample M = 1.

4.2 Determination of functions and parameters

The construction cost prediction model determined in this paper runs on the MATLAB operating platform and makes full use of the MATLAB toolbox to develop the prediction model of the network. By MATLAB the platform, scholars can complete the network design and learn the nervous system needed to use the network tool box directly. The neural logic BP network of MATLAB instruments have three main functions: NEFF, SIM and TRAIN.

newff function:

$\begin{matrix} net = & newff (PR, [S 1, S 2, \dots, SN], \\ {TF 1, TF 2, \dots, TFN}, BTF, BLF, PF \end{matrix}$ (1)

train function: $[net, tr] = train (NET, X, T, Pi, Ai)$ (2)

sim function: $y = sim (net, x)$ (3)

When developing a comprehensive neural network prediction scheme, parameter setting, training and prediction functions are the most important basis, and only with a full understanding of the use of these functions can training be prepared to ensure that the neural network can function properly and play a predictive role. On this basis, select appropriate trigger and training functions.

4.3 Network learning algorithm process

The standard learning method of neural network BP algorithm is as follows: step 1: determine the topology of the network, assign random numbers (-1.1) to each connection value for a period of time, determine the error function E, and provide ISPLON with a calculation accuracy value and a maximum number of studies M..1

Step 2: select the output sample randomly, count the selected number as K, and record the corresponding expected output. $x (k) = (x_{1} (k), x_{2} (k), \dots, x_{n} (k))$ (4) $d_{0} (k) = (d_{1} (k), d_{2} (k), \dots, d_{m} (k))$ (5)

Step 3: calculate the input and output of each neuron in the hidden δ. ${hi}_{n} (k) = \sum_{i = 1}^{n} w_{ih} x_{i} (k) - b_{h}, h = 1, 2, \dots, p$ (6) $h 0_{n} (k) = f ({hi}_{h} (k)), h = 1, 2, \dots, p$ (7) ${yi}_{0} (k) = \sum_{h = 1}^{p} w_{h 0} h 0_{h} (k) - b_{0}, o = 1, 2, \dots, m$ (8) ${yo}_{0} (k) = f ({yi}_{o} (k)), o = 1, 2, \dots, m$ (9)

According to the expected output and actual output of the network, the error function is calculated to δ the partial derivatives of each neuron in the output δ_o(k). $\frac{\partial e}{\partial ω_{ho}} = \frac{\partial e}{\partial {yi}_{o}} \frac{\partial {yi}_{0}}{\partial ω_{ho}}$ (10) $\frac{\partial {yi}_{o} (k)}{\partial ω_{ho}} = \frac{\partial (\sum_{h}^{p} ω_{ho} {ho}_{h} (k) - b_{0})}{\partial ω_{ho}} = {ho}_{h} (k)$ (11) $\frac{\partial e}{\partial {yi}_{o}} = - (d_{o} (k) - {yo}_{o} (k)) {yo}_{o} (k) @ - δ_{o} (k)$ (12)

Step 5: the output δ by using the connection weight from the hiddenδ to the output δ_o(k)δ of the output calculation error function of the hidden δ on the partial derivatives of each neuron in the hidden δ_h(k). $\frac{\partial e}{\partial ω_{ho}} = \frac{\partial e}{\partial {yi}_{o}} \frac{\partial {yi}_{o}}{\partial ω_{ho}} = - δ_{o} (k) {ho}_{h} (k)$ (13) $\frac{\partial e}{\partial ω_{ih}} = \frac{\partial e}{\partial {hi}_{o} (k)} \frac{\partial {hi}_{h} (k)}{\partial ω_{ih}}$ (14) $\frac{\partial {hi}_{o} (k)}{\partial ω_{ih}} = \frac{\partial (\sum_{i = 1}^{n} ω_{ih} x_{i} (k) - b_{h})}{\partial ω_{ih}} = x_{i} (k)$ (15)

$\begin{matrix} \frac{\partial_{e}}{\partial {hi}_{h} (k)} = & \frac{\partial (\frac{1}{2} \sum_{i = 1}^{n} d_{o} (k) - {yo}_{0} {(k)}^{2})}{\partial {ho}_{h} (k)} \\ \frac{\partial {ho}_{h} (k)}{\partial {hi}_{h} (k)} @ - δ_{h} (k) \end{matrix}$ (16)

Step 6: Use the error function to δ the partial derivative of each neuron in the output δ_o(k) and the output of each neuron in the hidden δ to modify the connection weight ω_ho (k). $V ω_{ho} (k) = - μ \frac{\partial_{e}}{\partial ω_{ho}} = μ δ_{o} (k) {ho}_{h} (k)$ (17) $ω_{ho}^{N + 1} = ω_{ho}^{N} + η δ_{O} (k) {ho}_{h} (k)$ (18)

Step 7: Using the δ of neurons in the hidden δ_h(k) and input correction connection weights for each neuron in the input δ.

$\begin{matrix} V ω_{ih} (k) = & - μ \frac{\partial_{e}}{\partial ω_{ih}} = - μ \frac{\partial_{e}}{\partial {hi}_{h} (k)} \frac{\partial {hi}_{h} (k)}{\partial ω_{ih}} \\ = δ_{h} (k) x_{i} (k) \end{matrix}$ (19) $ω_{ih}^{N + 1} = ω_{ih}^{N} + η δ_{h} (k) x_{i} (k)$ (20)

Step 8: calculate the global error. $e = \frac{1}{2 α} \sum_{k = 1}^{α} \sum_{o = 1}^{m} {(d_{0} (k) - y_{0} (k))}^{2}$ (21)

Step 9: analyze whether the expected output error meets the requirements, and complete the algorithm if the output error meets the requirements. The algorithm ends the algorithm when the number of training courses in the last network reaches the digital requirement. Instead, select a new learning sample and export it to the previous stage of restart.

5 Evidence-based analysis

5.1 Simulation application of cost prediction model based on neural network

The simulation analysis of the building cost prediction model is based on the two-tube network and neural network methods, the analysis is carried out on the MATLAB, is a matrix laboratory based on matrix operation, has first-class digital capability, is a computer platform designed for visual interactive programs, the platform goes beyond the tedious traditional non-interactive language programs, such as C language and Fortran systems, and greatly improves the efficiency of operation. There are many installed toolkits in the platform, users can use these toolkits as needed, and the interface is easy to see. The language is clear, easy to use, and the use efficiency is very high.

Start the MATLAB platform, after calling the neural network development tool network in the MATLAB, introduce the first 10 sets of sampling data, repeat the network learning, constantly adjust the neural network prediction to optimize the performance of the network and optimize the confluence process of application training.

Input the index data of the other two sets of sampling data into a neural network trained in prediction, then process the prediction by MST function, and then compare with the actual sample value to calculate the relative error between the actual value and the predicted value.

Results are listed in Table 3, the error range is±10%, which reaches the highest error level in the investment estimation of technical feasibility study and can be applied to price prediction.

Table 3
Error analysis table

Forecast Actual value Relative error/%

5028420 5194560 3.20

14962470 16152640 7.37

Forecast	Actual value	Relative error/%
5028420	5194560	3.20
14962470	16152640	7.37

5.2 Application analysis of traditional multiple linear regression model

If the change of one variable leads to the change of another variable, the relationship between them can be expressed in a rough straight line. The regression of two variables is called regression linear analysis. Multi-linear regression analysis is a research method. It is used to study the relationship between variables and another historical data. The change of one variable leads to the corresponding change of another variable, the first variable is called an explanatory variable or an independent variable. The general elements of regression analysis based prediction of explanatory variables are as follows: analyzing a set of sample data, simulating the linear functional relationship between autonomous variables and corresponding variables, statistically verifying the reliability of the proposed functional relationship, analyzing all independent variables, selecting important and non-important variables according to the resulting mathematical relationships, and predicting the value of variables with known independent variables. Sample as show in Table 4, Error analysis as show in Table 5 and Table 6, Comparison of errors between different prediction models as show in Table 7.

Table 4
Sample tables

Variable Project x1 Floor/m2 x2 Floor height/m x3 Number of δs/δs x4 Structure type x5 Depth/m y Cost/yuan

1 560.0 2.8 5.0 1.0 1.5 3088400.0

2 950.0 2.9 9.0 2.0 1.9 10277100.0

3 680.0 3.0 6.0 2.0 1.5 4906200.0

4 1458.0 2.8 12.0 3.0 2.2 21377196.0

5 397.7 3.5 12.0 2.0 2.4 5736424.8

6 293.5 2.8 18.0 3.0 3.1 6454358.5

7 620.0 2.8 21.0 3.0 6.5 15924700.0

8 484.2 3.2 6.0 2.0 5.0 3510450.0

9 640.0 2.8 8.0 2.0 1.3 6152320.0

10 420.0 2.8 30.0 3.0 5.2 15393840.0

11 480.0 2.8 9.0 2.0 2.2 5194560.0

12 1120.0 2.8 12.0 2.0 2.2 16152640.0

Variable Project	x1 Floor/m2	x2 Floor height/m	x3 Number of δs/δs	x4 Structure type	x5 Depth/m	y Cost/yuan
1	560.0	2.8	5.0	1.0	1.5	3088400.0
2	950.0	2.9	9.0	2.0	1.9	10277100.0
3	680.0	3.0	6.0	2.0	1.5	4906200.0
4	1458.0	2.8	12.0	3.0	2.2	21377196.0
5	397.7	3.5	12.0	2.0	2.4	5736424.8
6	293.5	2.8	18.0	3.0	3.1	6454358.5
7	620.0	2.8	21.0	3.0	6.5	15924700.0
8	484.2	3.2	6.0	2.0	5.0	3510450.0
9	640.0	2.8	8.0	2.0	1.3	6152320.0
10	420.0	2.8	30.0	3.0	5.2	15393840.0
11	480.0	2.8	9.0	2.0	2.2	5194560.0
12	1120.0	2.8	12.0	2.0	2.2	16152640.0

Table 5

Error analysis table

Forecast	Actual value	Relative error/%
4207609	5194560	19.0
17549290	15752640	11.4

Table 6

Error analysis table

Forecast	Actual value	Relative error/%
5572724	5194560	7.28
17490156	15752640	0.03

Table 7

Comparison of errors between different prediction models

Model	Forecast	Actual value	Relative error%
BIM+Neural Networks	5028420	5194560	3.2
	14962470	15752640	7.37
	4207609	5194560	19.0
Multiple linear regression	17549290	5194560	11.4
Neural networks	5572724	15752640	7.28
	17490156	15752640	11.03

The models are: $y = β_{0} + β_{1} x_{1} + β_{2} x_{2} + \dots β_{n} x_{n} + ɛ$ (22)

For n group of observations, it can be expressed as: ${\begin{matrix} y_{1 = β_{0} + β_{1} x_{11} + β_{2} x_{12} + \dots β_{n} x_{1 n} + ɛ_{1}} \\ y_{2 = β_{0} + β_{1} x_{21} + β_{2} x_{22} + \dots β_{n} x_{2 n} + ɛ_{2}} \\ \dots \\ y_{n = β_{0} + β_{1} x_{n 1} + β_{2} x_{n 2} + \dots β_{n} x_{nn} + ɛ_{n}} \end{matrix}$ (23)

The multi-linear regression model must satisfy the independence hypothesis, the offset hypothesis (such as the positive number hypothesis), the discontinuous correlation hypothesis (such as the multiple linear assumptions when estimating the model parameters) and so on.

When a regression factor test is performed, it may occur that the linear relationship between the factor and the response variable is small, sometimes because the assumption condition is not satisfied, so it is necessary to test and modify the regression hypothesis, for example, by difference test, normal test, multifunctional interoperability test and sequence test.

The multiple linear regression equation should be:

$\begin{matrix} y = & - 9223993.407 + 15517.804 x_{1} \\ + 536762.295 x_{3} + 523725.352 x_{5} \end{matrix}$ (24)

After standardization: $y = 0.839 x_{1} + 0.692 x_{3} + 0.156 x_{5}$ (25)

5.3 Traditional BP neural network model

The design parameters of the design stage, that is, the engineering characteristics, should be analyzed as an index system that affects the construction cost, that is, the building area, the building height, the standard area, the structure stratification, the structure type, the plane shape, the seismic strength level, the foundation type and the buried depth.

This means that although the topology is different, the theorem is the same and is based on the Kolmogolov theorem. According to this theorem, the number of N = 9 input points, the number of hidden nodes is 2 NN = 1 = 19, and the output points are determined by the objective data type of M = 1.

Open the MATLAB platform, use the neural network toolbox in the MATLAB to create the network, input the first 10 groups of sample data into the network for repeated learning, and adjust continuously to optimize the prediction system of the neural network. The training process is as follows: the training performance curve shows that the relative error of network convergence decreases until the precise requirements of the steps are met.

For the input of the index data of the other two sets of sampling data, the neural network trained by prediction is first used, then the prediction is processed by MST function, and then compared with the actual sample value, the relative error between the actual value and the predicted value is calculated.

5.4 Comparative accuracy analysis

In order to compare the accuracy of three different prediction models using the same set of sampling data, the results of the three models are briefly compared.

In summary:

(1) The prediction results of the multiple linear regression model, with a maximum relative error of 19.0%, are based on widely used statistical analysis. If the relationship between variables is not linear, or if the functional relationship between variables is not accurately understood, it is impossible to determine the functional performance of the regression equation, and the factors that affect the uncertainty of variables should be fully considered. Because of its adaptive learning ability, neural network can approach any form of functional relationship, especially nonlinear relationship, and can better adapt to the dynamics and uncertainty of building cost.

(2) On the whole, BP neural network prediction model is more accurate than multiple regression model, but the range is small and the return is unstable. The relative error between prediction and actual value is 7.28% and 11.03%. Neural network can rapidly develop any form of functional relationship with core radiation function, but the requirement of sample data is higher. Data sensitive points can bring the network to the minimum, and too many data input carriers can lead to excessive estimation.

(3) The BIM neural network prediction model is the most accurate of these three models. It makes up for the defects of linear regression model with high accuracy, simplifies the neural network prediction model, improves the calculation speed and improves the popularization ability.

6 Discussion on construction cost management strategy

6.1 Enhanced training of cost managers

The level of price managers is the key to dynamic building cost management and determines the accuracy and completeness of dynamic building cost management. Dynamic management of the pricing process requires a lot of work and simplified management, which requires participation in the real pricing process. Builders need to abandon outdated and inappropriate price management concepts and actively learn different advanced dynamic control methods to control and master prices. Construction companies must improve the level of construction by providing better training, in addition to organizing expert meetings on pricing to discuss competitive bidding prices. Price can also be reflected in practice, can arrange old workers to work with new workers, this is not only to improve the enthusiasm of employees, but also to effectively help improve the technical level of new employees, and improve the engineering application, dynamic cost management and professional talent reserve level of enterprises.

6.2 Strengthening the dynamic management and control of the whole process of cost work

For the dynamic management and control of construction costs, it is necessary to focus on the management of all construction stages, which requires comprehensive dynamic management and control, as well as good management of each construction stage, from the policy formulation stage of the project to the final stage of the project. The decision-making stage requires the senior management of the company to accurately analyze the necessity of construction, carry out scientific planning, and compile construction estimates and budgets so that the construction process can proceed smoothly and ensure the quality and cost-effectiveness of construction. The design phase requires the designer to conduct a comprehensive engineering study, taking into account the relevant standards and specifications and the reality of the project, in order to effectively reduce the cost of the project. The construction quality can guarantee the dynamic management of the project cost, the most critical stage is the construction stage, in which the constructors must collect all the cost data related to the construction and statistics the relevant information in a timely and comprehensive manner. Provide reliable database for dynamic cost management and ensure accurate construction data. The completion phase of the project, including quality audit and financial settlement, construction funds and costs incurred throughout the project, is compared and reviewed, mainly based on construction contracts to ensure the accuracy of settlement data and the reliability of dynamic engineering cost management.

6.3 Enhanced use of information technology

An important feature of the engineering pricing process is the quantity of data required to improve the application of information technology in price management to ensure the accuracy of statistics to carry out the necessary statistics. The use of information technology well meets the needs of dynamic control, utilizes the benefits of information technology automation and knowledge, strengthens innovation in information management, and enables departments to effectively integrate work types and staff. Effectively reduce the pressure on construction staff. BIM technology allows the simulation of project construction from the decision stage to the start-up stage. accurately estimate the material usage and workload involved in the whole process and provide reliable data sources for construction works. In addition, binoculars technology provides effective action based on early warning. It is very important to reduce the construction risk, reduce the construction cost and promote the dynamic cost control in the whole process, which is helpful to solve the various problems in the construction process.

7 Conclusion

As an innovative scheme in the field of construction, price management and control of construction projects can help to reduce costs and improve the economic efficiency of construction companies. Construction companies must actively implement dynamic price management and control, and improve their price management and control. The cost of the project varies from the decision stage of the project feasibility study to the business stage of the final acceptance. In this process, a series of dynamic factors affect the cost of these products. Therefore, from a scientific point of view, it is feasible to develop a project cost prediction model using the combination of double cylinder and neural network method. Its advantages and values have been recognized by the general public. Its application will promote a new industrial revolution, which will create many convenient conditions for engineering production, improve efficiency and control their own costs. While growing and improving, it is also necessary to realize its value and create extraordinary values for society.

References

Halfawy

Mahmoud

and Froese

Thomas

, Component-based framework for implementing integrated architectural/engineering/construction project systems, Journal of Computing in Civil Engineering 21(6) (2007), 59–64.

Eastman

Chuck

, Teicholz

Paul

, Sacks

Rafael

, et al. BIM Handbook (Second Edition), New York: John Wiley & Sons (2011), 12–16 .

Cetal

Eastman

, BIM Handbook: A Guide to Building Information Modeling for Owners, Managers, Designers, Engineers and Contractors, New York: John Wiley and Sons, 2008.

Bilal

, Building information modeling framework: A research and delivery foundation for industry stackeholders, Automation in Construction 18 (2009), 357–375.

Ernstrom

and Hanson

, The constructors guide to BIM[M], New York, AGC of America, 23(2) (2005), 110–116.

Phair

Matthew

, Software used for Freedom Tower gets updated, Building Designand Constructio 1(4) (2005), 19–19.

Tulke

and Hanff

, 4D Construction Sequence Planning-New Process and Data Model, CIB-W78 24th International Conference on Information Technology in Construction, Maribor 23 (2007), 11–22.

Ospina-Alvarado

A.M.

and Castro-Lacouture

, Interaction of processes and phases inproject scheduling using BIM for AECFM integration, Construction Research Congress 13(2) (2010), 939–947.

Tauscher, Mikulakova, Beucke, König. Automated generation of construction schedulesbased on the IFC object model, Computing in Civil Engineering 12 (2009), 666–674.

10.

Liston

Kathleen Mc Kinney

and Fischer

Martin A

. Generating, evaluating and visualizingconstruction schedules with 4D-CAD tools, Automation in Construction 7(6) (1998), 433–447.

11.

James, D. Goedert, Pavan. Integrating construction process documentation into Building information modeling, Construction Engineering and Management 12(7) (2010), 509–515.

12.

Shahandashti

S.M.

and Ashuri

, Forecasting engineering news-record construction cost index using multivariate time series models, Journal of Construction Engineering and Management 9(9) (2013), 1237–1243.

13.

Trost

S.M.

and Oberlender

G.D.

, Predicting accuracy of early cost estim-ates using factor analysis and multivariate regression, Journal of Construction Engineering and Management 129(2) (2003), 198–204.

14.

Mair

and Shepperd

, The consistency of empirical comparisons of regression and analogy-based software project cost prediction, in: Empirical Software Enginee-Ring 55 (2005), 102–110.

15.

Kim

G.H.

, An

S.H.

and Kang

K.I.

, Comparison of construction cost estimating models based on regression analysis, neural networks, and case-based reasoning, Building and Environment 39(10) (2004), 1235–1242.

16.

Barsoum

S.H.

and Georgy

M.E.

, Artificial neural networks model for parametric estimating of construction project costs, Journal of Engineering and Applied Science 52(6) (2005), 1050–1066.

17.

Hsiao

F.Y.

, Wang

S.H.

and Wang

W.C.

, Neuro-fuzzy cost estimation model enhanced by fast messy genetic algorithms for Semiconductor hookup construction, Computer-Aided Civil and Infrastructure Engineering 27(10) (2012), 764–781.

18.

Elfaki

A.O.

, Alatawi

and Abushandi

, Using intelligent techniques in construction project cost estimation: 10-Year Survey, Advances in Civil Engineering 20(14) (2014), 32–42.

19.

Naik

M.G.D.

, Kumar

, Construction project cost and duration optimization using artificial neural network, in: Architectural engineering conference, Birth and Life of the Integrated Building 22(5) (2015), 433–444.

20.

Sridarran

, Keraminiyage

and Herszon

, Improving the cost estimates of complex projects in the project-based industries, Built Environment Project and Asset Management 7(2) (2017), 173–184.

21.

Feng

T.F.

, Liu

X.S.

, Zhong

, et al. Research of combination forecasting model based on improved analytic hierarchy process, EURASIP Journal on Wireless Communications and Networking 5(1) (2018), 1–8.

22.

Winalytra

, Budi

A.S.

and Andreas

, Cost estimation model for igirder bridge superstructure using multiple linear regression and artificial neural network, Applied Mechanics and Materials 4619(1762) (2018), 142–149.

23.

Shi

H.W.

, Li

W.Q.

, The integrated methodology of rough set theory and artificial neural-network for construction project cost prediction, in: Second International Symosium on Intelligent Information Technology Application, IEEE Computer Society 32(4) (2008), 15–22.

24.

Cheng

M.Y.

, Tsai

H.C.

and Sudjono

, Conceptual cost estimates using evolutionary fuzzy hybrid neural network for projects in construction industry, Expert Systems with Applications 37(6) (2009), 4224–4231.

25.

Koo

, Hong

and Hyun

, The development of a construction cost prediction model with improved prediction capacity using the advanced CBR approach, Expert Systems with Applications 38(7) (2011), 8597–8606.

26.

Ashuorth

Allan

, Pre-Contant studies: Development, Economics, Tendering and Estimating, Oxford: Blankwell Publishing, 12(2) (2012), 45–53.

27.

Ciark

F.D.

and Lorenzoni

A.B.

, Applied cost Engineering, New York: Mareed Dekker Inc, 22(4) (2005), 12–19.

Construction cost management strategy based on BIM technology and neural network model

Abstract

Keywords

1 Introduction

2 Related work

3 Characteristics of 3BIM technology and its application in construction cost management

3.1 BIM Technology connotation

3.2.1 Building data accumulation

4 Construction of cost prediction model based on neural network

4.1 Determination of network topology

4.2 Determination of functions and parameters

5.1 Simulation application of cost prediction model based on neural network

Table 3 Error analysis table Forecast Actual value Relative error/% 5028420 5194560 3.20 14962470 16152640 7.37

5.4 Comparative accuracy analysis

6 Discussion on construction cost management strategy

6.1 Enhanced training of cost managers

6.2 Strengthening the dynamic management and control of the whole process of cost work

6.3 Enhanced use of information technology

7 Conclusion

References

Table 3
Error analysis table

Forecast Actual value Relative error/%

5028420 5194560 3.20

14962470 16152640 7.37