Identification of Critical Pavement Response Location Using a Gradient Descent Approach

Abstract: In mechanistic-empirical pavement design, the magnitude of certain responses is correlated to the development of different pavement distresses. Critical mechanistic responses are computed at different analysis points in the pavement structure to mitigate these distresses. Selecting appropriate analysis points is important in capturing the critical location, where the pavement experiences a maximum response in the subsurface. Responses calculated at these analysis points are employed in a performance model. Missing the critical response location results in a thinner structure, since the design is based on a lower response. Historically, these analysis points were selected using engineering judgement or by creating a dense grid to calculate response at multiple locations within the loading footprint of the aircraft. In this work, a gradient descent optimization algorithm was used to iteratively search for the critical response location at a specific depth using an elastic response model. Furthermore, changes in the critical location were investigated for different material properties and response types. While the procedure is sensitive to the seed location, results showed that a multi-start approach accurately identifies the critical response location for a given aircraft, pavement structure, and response type. Based on preliminary findings, the proposed procedure could be used to select analysis points based on identified critical locations to program vehicle databases for pavement design software.