Site Map

Search

PEER Director Mosalam Presents Future Vision
at 2018 PEER Annual Meeting

Mosalam

PEER Director Khalid Mosalam

Opening remarks from PEER Director Khalid Mosalam at the 2018 PEER Annual Meeting started with a brief history of EERC and PEER, and a summary of PEER and its mission. A few highlights of the ongoing research efforts and progress at PEER, including mega-projects and recently funded TSRP and lifelines projects, as well as key events that took place last year, were presented. The presentation ended with some envisioned future directions.

A vision of the overarching theme of PEER’s future focused on natural hazards, extreme events, and resiliency initiatives. The vision made use of a “spiral” technology transfer model. Building upon the early research on Performance-Based Earthquake Engineering with many developments achieved in the past 21 years, PEER is aiming towards Resilient Design for Extreme Events.

The presentation also highlighted recent findings from a recently completed PEER Blind Prediction Competition using a shaking table experiment of an innovative design of a two-column bridge bent. Such blind predictions are useful tools for independent verification and validation of computer models. The success of this competition spurred interest to continue such efforts in future years to use experimental facilities in other PEER core campuses, and to close the gap between computational model prediction and experimental observations.

Advances in data science and machine learning were presented and give rise to PEER’s future interest to embark on more activities related to sensor network and data-driven structural health monitoring (SHM). A vision of Data to Decision (D2D) was illustrated with the following four elements: 1) data-driven damage assessment, 2) vision-based damage assessment, 3) sensor development for SHM, and 4) development of a seismic performance observatory (SPO) which is a structural information/damage database that was recently used in the PEER reconnaissance efforts after the magnitude 7.1 Mexico Central Region earthquake on September 19, 2017.  Motivated by recent success of the use of deep transfer learning for image-based structural damage recognition, a future release (sometime in the second half of 2018) of a new PEER Hub Image (PHI) challenge, analogous to the ImageNet challenge, was announced.

In the PHI challenge to be held annually by PEER, each participating team will be asked to complete several multi-classification tasks and a localization task. More than 20,000 labeled images will be provided as training data to the contestants. Detection performance will be evaluated on test images, for which labels will not be provided. Prediction results by the contestants will be submitted together with a brief report including their used algorithms/methods as their challenge submittals. Prior to the announcement of the PHI challenge, call for uploading images to SPO website together with an image labeling web application will be posted in the PEER website.

With these initiatives, PEER expands its current Performance-Based Earthquake Engineering to other extreme events. Combining well-established engineering research with the power of data sciences research and the broad reach of social sciences research will invigorate PEER’s mission.