Knowledge based approach for productivity adjusted construction schedule

https://doi.org/10.1016/S0957-4174(01)00029-XGet rights and content

Abstract

The nature of the construction industry always resists the adoption of management techniques that is successful in other disciplines. Unlike the manufacturing industry each construction project presents a unique planning problem to project planners and managers. The construction process is carried out at a new location and materials, labor and equipment are transported to a new job site. The construction is one of few processes in which the product is built in an open environment subject to various weather conditions and other environmental factors which affect labor as well as equipment productivity. This paper presents a knowledge based approach for analysis of different factors that affect construction project duration and generates normal schedules and productivity-adjusted schedules. Project mangers can be greatly aided by such automated system by simulating different factors that reflect anticipated and unanticipated project conditions and perform several schedule alternatives. An analysis of alternative schedules by using the integrated knowledge based system is presented. The knowledge based system interfaces to estimating software, spreadsheet and scheduling soft wares in order to receive data and generate schedule reports.

Introduction

Forecasting of reliable project completion date presents continuing challenges to construction schedulers. This reliability is dependent upon the accuracy of estimating individual activity duration. Generally, activity duration is estimated as the planned quantity of work divided by average crew productivity. Average crew productivity is available from various standard handbooks. However, this average crew productivity must be modified to reflect the actual job, site, local conditions and other uncertain variables; otherwise schedule overrun is unavoidable. In the study carried out by Cohenca et al., 1989, Laufer and Cohenca, 1990 on the top 400 construction firms in the United States mean schedule variance between actual and planned construction time has been found to be 2.5 weeks, and the mean man hour variance 8.4%. In Britain, the National Economic Development office (NEDO, 1983) found that 50% of projects overrun their planned times by 1 month or more. In New Zealand, Soeterik and Foster (1976) reported that the time overruns on contract completion were 20% on average. In addition, the traditional scheduling tools such as Ms Project and Primavera require complete input of activity description, precedence relationships, durations and allocations of resources for each activity and do not consider the uncertainty related with each activity in estimating project durations. Because forming of schedule manually is very costly and time consuming the scheduler does not perform several schedule alternatives to evaluate time integrated cost that indicate the timing of capital needs and can not perform several schedule alternatives by simulating different productivity factors that have an impact on the duration of the project.

A knowledge based approach offers a new methodology and technique, which has the potential to over come the shortcomings and difficulties of conventional scheduling tools. More over the nature of construction planning and scheduling problems are well suited to the application of knowledge-based expert system. This paper presents an integrated knowledge based system for alternative schedule analysis and schedule generation. The knowledge based system supports an automated on line crew and construction method description and crew size selection for generation of normal activity duration. It also supports an automated productivity factor selection for simulating anticipated and unanticipated project conditions and then generates a productivity-adjusted schedule.

Section snippets

Application of expert system in construction scheduling

Knowledge based expert systems have been a subject of considerable research in civil engineering in recent years, especially in the area of construction engineering and management. This section briefly describes several previous works that used knowledge based systems for construction planning and scheduling.

Construction Planex (Hendrickson et al., 1987) uses a bottom-up approach to generate a construction plan. It generates element activities for the building components, aggregates the element

Structure of knowledge base

The structure of the knowledge base has the following components as shown in Fig. 1.

  • 1.

    The knowledge based system is implemented on object oriented environment by using LEVEL 5 OBJECTR for MicrosoftR WindowsTM release 3.6 by Information Builders Inc. LEVEL 5 Object has an integrated array of powerful tools: Graphical User interface development, forms and display builders and has capability to chain more than one knowledge base together. It accesses to all common data base formats and SQL

Knowledge acquisition

An important part of a building knowledge based system is acquiring the knowledge needed to achieve a desired level of performance. The system's knowledge was acquired through two sources: (1) literature review; and (2) on site survey

  • 1.

    Literature review was conducted in three directions:

    • 1.1.

      literature about construction planning and scheduling (O'Brien, 1993, Pilcher, 1992, Clough and Sears, 1994).

    • 1.2.

      Literature about factors that affect productivity on construction site and methods of quantifying the

Knowledge library classes

Knowledge library classes contain construction methods, crew daily output, productivity levels and weather information. They have predefined instances and serve as database. The library classes will not be modified or adjusted during scheduling and productivity analysis processes. However, all library classes can be changed from instance editors to reflect the changes in construction location and time. The library classes are categorized into three groups: crew library class, productivity

Crew library class

Crew library class contains attributes such as: single crew ID, single crew description and single crew daily output as shown in Fig. 2. The crew library class is displayed in crew display, after project quantity and cost are input by user or imported to project class from estimating software or Ms Excel. Such arrangement provides the scheduler step-by-step and orderly way of scheduling. First, the system displays project data from which the scheduler learns the amount of work of each task.

Productivity level library class

Productivity level library class contains productivity levels for the different productivity factors under different conditions. Productivity level library objects have attributes: F1, F2, F3, F4 that represent very good, good, average and poor, respectively. For example, if the specification is very good then F1 of the specification is 98%, if the specification is good then F2 of the specification is 90% etc. The productivity levels are obtained from the foreman delay survey conducted in the

Weather library classs

Weather library classes contain average temperature, humidity and average rain and their corresponding productivity levels. The productivity level related to cold weather and hot/warm weather is obtained from Koehn and Brown (1985) mathematical equations. The values of these mathematical equations for cold and hot weather for 15 years averages is imported from Microsoft ExcelTM to weather the Library class. The two non linear equations, which express construction productivity as a function of

Selection classes

Selection classes contain scheduling parameters, type and number of crew and productivity factors selection. Each type of crew and productivity factors selection are represented by compound attributes and the compound attributes are attached to series of crew selection, and productivity factors selection displays as shown in Fig. 3, Fig. 4. The compound attributes also have attached methods, demons and production rules which are part of the different category of process modules. When user

Project classes

Project classes contain quantity, cost and schedule attributes. The values of schedule attributes are determined by process modules. The process modules are the engine of the integrated knowledge based system. Each process module consist of a production rules, methods, demons and system commands. There are two types of process modules that perform different types of processes and these are: (1) the schedule process module; and (2) the productivity analysis process module.

Schedule process module

The schedule process module performs the following processes:

  • 1.

    Calculating activities duration.

  • 2.

    Generating activities sequences. Based on the knowledge extracted from schedule experts The system uses activity constrained sequence relations. Activity constrained relations determine the sequencing of activities based on activity type. For example, ‘place formwork’ always precedes ‘place concrete’.

  • 3.

    Assigns relationship and lag time for each activity instance.

  • 4.

    Assigns each activity instance selected

Productivity analysis process

Productivity simulations for adjustment of normal activity duration are performed as follow. First user selects productivity factors that reflect expected project conditions from productivity displays. Then the productivity process modules obtain the corresponding productivity levels from productivity library class and then normal activity durations are adjusted. The combined effect of productivity factors associated with activity productivity is considered independent and hence a

Conclusions

The traditional estimating and scheduling takes many man hours of detailed effort: first the scheduler must translate the design information to activity lists, and formulate the network, calculate cost and duration in order to prepare input data for traditional scheduling tools such as MS Project or Primavera for generating schedules reports. This is costly and time consuming to be used to search for alternative schedule analysis. Moreover, the traditional scheduling tools do not have a

Acknowledgements

The writer would like to thank to site Eng. Nurhan Erk and to the many foremen who regularly and carefully filled in the forms daily over the period of 9 months.

References (32)

  • A.N Mohamed et al.

    An integrated knowledge based system for alternative design, material selection and cost estimating, Expert Systems With Application

    (1998)
  • H Adeli

    Expert system in construction and structural engineering

    (1988)
  • H Adeli

    Knowledge engineering

    (1990)
  • H Adeli

    Knowledge engineering

    (1990)
  • H.N Ahuja et al.

    Simulation model to forecast project completion time

    Journal of Construction Engineering and Management

    (1985)
  • N.B Benjamin et al.

    Simulation effects of weather on construction

    Journal of Construction Division, ASCE

    (1973)
  • R.I Carr

    Simulation of construction project duration

    Journal of Construction Division, ASCE

    (1979)
  • R.H Clough et al.

    Construction contracting

    (1994)
  • D Cohenca et al.

    Factors affecting construction planning efforts

    Journal of Construction Engineering and Management

    (1989)
  • C Hendrickson

    Expert system for construction planning

    Journal of Computing in Civil Engineening

    (1987)
  • J.P Ignizo

    Introduction to expert systems: the development and implementation of rule based expert system

    (1991)
  • N Kartam et al.

    Intelligent planning of construction projects

    Journal of Computing in Civil Engineering

    (1990)
  • E Koehn et al.

    Climatic effect on construction

    Journal of Construction Engineering and Management

    (1985)
  • A Laufer et al.

    Factors affecting construction planning

    Journal of Construction Engineering and Management

    (1990)
  • Level Object for Microsoft Windows (1995). Reference guide manual, release 3.6. Information Builder...
  • R.E Levitt

    Artificial Intelligence Techniques for Generating Construction Project Plans

    Journal of Construction Engineering & Management

    (1989)
  • Cited by (5)

    View full text