There are relatively few available construction equipment detectors models thatuse deep learning architectures; many of these use old object detection architectures like CNN (Convolutional Neural Networks), RCNN (Region-Based Convolutional Neural Network), and early versions of You Only Look Once (YOLO) V1. It can be challenging to deploy these models in practice for tracking construction equipment while working on site. This thesis aims to provide a clear guide on how to train and evaluate the performance of different deep learning architecture models to detect different kinds of construction equipment on-site using two You Only Look Once (YOLO) architecturesYOLO v5s and YOLO R to detect three classes of different construction equipment onsite, including Excavators, Dump Trucks, and Loaders. The thesis also provides a simple solution to deploy the trained models. Additionally, this thesis describes a specialized, high-quality dataset with three thousand pictures created to train these models on real data by considering a typical worksite scene, various motions, varying perspectives, and angles of construction equipment on the site. The results presented herein show that after 150 epochs of training, the YOLORP6 has the best mAP at 0.981, while the YOLO v5s mAP is 0.936. However, YOLO v5s had the fastest and the shortest training time on Tesla P100 GPU as a processing unit on the Google Colab notebook. The YOLOv5s needed 4 hours and 52 minutes, but the YOLOR-P6 needed 14 hours and 35 minutes to finish the training.ii The final findings of this study show that the YOLOv5s model is the most efficient model to use when building an artificial intelligence model to detect construction equipment because of the size of its weights file relative to other versions of YOLO models- 14.4 MB for YOLOV5s vs. 288 MB for YOLOR-P6. This hugely impacts the processing unit’s performance, which is used to predict the construction equipment on site. In addition, the constructed database is published on a public dataset on the Roboflow platform, which can be used later as a foundation for future research and improvement for the newer deep learning architectures.

Interpersonal communications during civil infrastructure systems operation and maintenance (CIS O&M) are processes for CIS O&M participants to exchange critical information. Poor communications that provide misleading information can jeopardize CIS O&M safety and efficiency. Previous studies suggest that communication contexts and features could be indicators of communication errors and relevant CIS O&M risks. However, challenges remain for reliable prediction of communication errors to ensure CIS O&M safety and efficiency. For example, existing studies lack a systematic summarization of risky contexts and features of communication processes for predicting communication errors. Limited studies examined quantitative methods for incorporating expert opinions as constraints for reliable communication error prediction. How to examine mitigation strategies (e.g., adjustments of communication protocols) for reducing communication-related CIS O&M risks is also challenging. The main reason is the lack of causal analysis about how various factors influence the occurrences and impacts of communication errors so that engineers lack the basis for intervention.

This dissertation presents a method that integrates Bayesian Network (BN) modeling and simulation for communication-related risk prediction and mitigation. The proposed method aims at tackling the three challenges mentioned above for ensuring CIS O&M safety and efficiency. The proposed method contains three parts: 1) Communication Data Collection and Error Detection – designing lab experiments for collecting communication data in CIS O&M workflows and using the collected data for identifying risky communication contexts and features; 2) Communication Error Classification and Prediction – encoding expert knowledge as constraints through BN model updating to improve the accuracy of communication error prediction based on given communication contexts and features, and 3) Communication Risk Mitigation – carrying out simulations to adjust communication protocols for reducing communication-related CIS O&M risks.

This dissertation uses two CIS O&M case studies (air traffic control and NPP outages) to validate the proposed method. The results indicate that the proposed method can 1) identify risky communication contexts and features, 2) predict communication errors and CIS O&M risks, and 3) reduce CIS O&M risks triggered by communication errors. The author envisions that the proposed method will shed light on achieving predictive control of interpersonal communications in dynamic and complex CIS O&M.

Public institution facility operations and maintenance is a significant factor enabling an institution to achieve its stated objectives in the delivery of public service. To meet the societal need, Facility Directors must make increasingly complex decisions managing the demands of building infrastructure performance expectations with limited resources. The ability to effectively measure a return-on-investment, specific to facility maintenance indirect expenditures, has, therefore, become progressively more critical given the scale of public institutions, the collective age of existing facilities, and the role these institutions play in society.

This research centers on understanding the method of prioritizing routine work in support of indirect institutional facility maintenance expense through the lens of K-12 public education in the state of Arizona. The methodology documented herein utilizes a mixed method approach to understand current facility maintenance practices and assess the influence of human behavior when prioritizing routine work. An evidence-based decision support tool, leveraging prior academic research, was developed to coalesce previously disparate academic studies. The resulting process provides a decision framework for prioritizing decision factors most frequently correlated with academic outcomes.

A purposeful sample of K-12 unified districts, representing approximately one-third of the state’s student population and spend, resulted in a moderate to a strong negative correlation between facility operations and student outcomes. Correlation results highlight an opportunity to improve decision making, specific to the academic needs of the student. This research documents a methodology for constructing, validation, and testing of a decision support tool for prioritizing routine work orders. Findings from a repeated measures crossover study suggest the decision support tool significantly influenced decision making specific to certain work orders as well as the Plumbing and Mechanical functional areas. However, the decision support tool was less effective when prioritizing Electrical and General Maintenance work orders.

Moreover, as decision making transitioned away from subjective experience-based judgment, the prioritization of work orders became increasingly more consistent. The resulting prioritization, therefore, effectively leveraged prior empirical, evidence-based decision factors when utilizing the tool. The results provide a system for balancing the practical experience of the Facility Director with the objective guidance of the decision support tool.

Implementing Building Information Modeling (BIM) in construction projects has many potential benefits, but issues of projects can hinder its realization in practice. Although BIM involves using the technology, more than four-fifths of the recurring issues in current BIM-based construction projects are related to the people and processes (i.e., the non-technological elements of BIM). Therefore, in addition to the technological skills required for using BIM, educators should also prepare university graduates with the non-technological skills required for managing the people and processes of BIM. This research’s objective is to develop a learning module that teaches the non-technological skills for addressing common, people- and process-related, issues in BIM-based construction projects. To achieve this objective, this research outlines the steps taken to create the learning module and identify its impact on a BIM course. The contribution of this research is in the understanding of the pedagogical value of the developed problem-based learning module and documenting the learning module’s development process.

This research focuses on assessing the impact of various process mapping activities aimed at improving students' abilities to plan for Building Information Modeling (BIM). During the various educational activities, students were tasked with generating process maps to illustrate plans for hypothetical construction projects. Several different educational approaches for developing process maps were used, beginning in the Fall 2015 semester. In all iterations of the learning activity, students were asked to create level 1 (project-specific) and level 2 (BIM use-specific) process maps based on a previously published BIM Project Execution Planning Guide. In Fall 2015, a peer review activity was conducted. In Spring 2016, a collaborative activity was conducted. Beginning in the Fall 2016 and Spring 2017 semesters, an additional process mapping activity was conducted aimed at separating process mapping and BIM planning into separate activities. In Fall 2016, the BIM activity was conducted in groups of three whereas in Spring 2017, the students were asked to create individual process maps for the given BIM use. To understand the impact of the activity on students' perception of their own knowledge, a pre-and post-activity questionnaire was developed. It covered questions related to: (i) students' ability to create a process map, (ii) students' perception about the importance of a process map and (iii) students' perception about their own knowledge of the BIM execution process. The process maps were analyzed using a grading rubric developed by the author. The grading rubric is the major contribution of the work as there is no existing rubric to assess a BIM process map. The grading rubric divides each process map into five sections, including: core activity; activities preceding the core activity; activities following the core activity; loop/iteration; and communication across the swim lanes. The rubric consist of two parts that evaluate (i) the ability of students to demonstrate each section and (ii) the quality of demonstration of each section. The author conducted an inter-rater reliability index to validate the rubric. This inter-rater reliability index compares the scores students’ process maps were when assessed by graduate students, faculty, and industry practitioners. The reviewers graded the same set of twelve process maps. The inter-rater reliability index was found to be 0.21, which indicates a fair agreement between the graders. The non-BIM activity approach was perceived as the most impactful approach by the students. The assessment of the process maps with the rubric indicated that the non-BIM approach was the most impactful approach for enabling students to demonstrate their ability to create a process map.