Level:
Intermediate
TCM Section(s)
7.6. Risk Management
Venue: 2024 AACE International Conference & Expo
Abstract: Recently, AI-schedule risk analysis (AI-SRA) has emerged as a groundbreaking approach to construction project risk management. [1,2] Unlike traditional quantitative schedule risk analysis (QSRA),[3] AI-SRA leverages machine learning models, trained on extensive historical schedule data, to directly predict activity durations distributions based on data embedded in the schedule. As previously shown, AI-SRA surpasses traditional QSRA in both the accuracy of end-date forecasts and activity duration predictions. [4,5] Despite its proven effectiveness, practical implementations of AI-SRA within project organizations have so far been scarce and the practicalities of rolling out this new process are unstudied.
Here, a comprehensive study of how to implement AI-SRA successfully on megaprojects is presented. AI-SRA roll-out on five projects across different sizes and sectors shows that, generally, three challenges need to be overcome: Lack of trust in AI models; the unintuitive nature of AI-SRA results; and AI-SRA differing from tried and tested processes. These challenges can be overcome by clear steps towards understanding AI, visualizations to aid comprehension of the results, and demonstrations of equivalence of AI-SRA to traditional processes. As a case study, the specific application of these principles on a major UK rail project using AI-SRA is discussed.
TCM Section(s)
7.6. Risk Management
Venue: 2024 AACE International Conference & Expo
Abstract: Recently, AI-schedule risk analysis (AI-SRA) has emerged as a groundbreaking approach to construction project risk management. [1,2] Unlike traditional quantitative schedule risk analysis (QSRA),[3] AI-SRA leverages machine learning models, trained on extensive historical schedule data, to directly predict activity durations distributions based on data embedded in the schedule. As previously shown, AI-SRA surpasses traditional QSRA in both the accuracy of end-date forecasts and activity duration predictions. [4,5] Despite its proven effectiveness, practical implementations of AI-SRA within project organizations have so far been scarce and the practicalities of rolling out this new process are unstudied.
Here, a comprehensive study of how to implement AI-SRA successfully on megaprojects is presented. AI-SRA roll-out on five projects across different sizes and sectors shows that, generally, three challenges need to be overcome: Lack of trust in AI models; the unintuitive nature of AI-SRA results; and AI-SRA differing from tried and tested processes. These challenges can be overcome by clear steps towards understanding AI, visualizations to aid comprehension of the results, and demonstrations of equivalence of AI-SRA to traditional processes. As a case study, the specific application of these principles on a major UK rail project using AI-SRA is discussed.