Predictive Model for Siding Practice in Panelized Construction

Abstract

Offsite construction offers an opportunity to standardize processes and better predict schedule requirements when compared with onsite construction. This predictability allows for balanced labor distribution and accurate time estimation. This research investigates the exterior wall siding practice at a panelized home manufacturing facility in order to predict future productivity of this operation based on a time study and known panel design characteristics such as wall length, wall height, and number and size of openings in the wall. The siding workstation is currently a bottleneck in the wall production line. In this area vinyl siding, window and door trim, and other exterior finishes are added to the panels. The case study plant uses a radio frequency identification (RFID) system to track panel locations and the amount of time spent at each station. This system tracks the time the panels spend in the siding area, but not the amount of time that is necessary to complete the work required. This discrepancy results in difficulties identifying the idle time and the working time within the total duration. By applying data science procedures of classification and association applied to lean manufacturing concepts, such as value-added activities and waste minimization, the research in this paper establishes a model to predict the labor requirement for each panel at early design stages. Using the developed model, the case study factory is able to quantify the idle versus working time that panels are subject to at the exterior wall siding workstation, as well as the ratio of value-added activities to non-value- added activities.