Development of an inference structure for automated data collection systems for construction operation

An important prerequisite for establishing cause-effect relationships is the availability of large amounts of high-quality data that have been collected in a consistent and uniform manner. Manual methods are often subjective. The automation of data collection is an essential step to eliminate subjectivity. The limitations of current sensor technology warrant development of an inference structure that can handle uncertain input. An inference structure is developed, based on the Bayesian Theory and using surrogate variables. Surrogate variables are work elements that help to predict production activities. The Bayesian approach is chosen after comparing it with other methods of decision theory on choices under uncertainty. Two types of construction are considered--masonry and compaction. Data collection methodology is discussed in detail. An experiment was conducted to validate the likelihood data collection method. The Cochran method was used to test the repeatability of the data collection method. The results of the Cochran test showed that all the surrogate variables except one are suitable to use in the model. Sensitivity analyses were carried out to address three aspects--accuracy of the model to predict productive activities, effects of the number of sensors on the probability, and the effects of the reliability of sensors on the final probability. The model was found to be sufficiently accurate. The effect of increasing the number of sensors did not follow the anticipated results. Consideration of reliability of the sensors enhanced the model.