The overarching goal of the project was to commercialize a visible optics guided diagnostic technology to mitigate the adverse effects of gas leakage, reduce financial loss, save lives and overcome existing limitations of current technologies by implementing a portable and inexpensive state-of-the-art leak detection solution. The primary activities in this program were (1) customer discovery and (2) customer validation, directed towards an informed prototype development. The hydrogen gas production and operation facilities were determined as the primary point of entry based on perceived needs for cost and energy savings in related markets, prevention of environmental hazards, the outcomes of the I-Corps customer discovery, and to a certain extent the focus of certain government agencies towards accelerating the growth of hydrogen power for the next decade. However, the technology being agnostic to the chemistry of the gases, it can be ported for sister gas-based facilities such as carbon dioxide leak detection in carbon capture and storage, as well as poisonous gas leaks in fabrication facilities, to identify a few. 

Some of the key Intellectual Merit related outcomes of the project were:
1. Through the customer discovery process, certain challenges related to the proposed technology were identified, most notably the need to leverage data recorded from the diagnostic tool to improve the accuracy of the predictive and detection methodology.
2. The use of vector validation would assist in mitigating cross-correlation of images that might induce some error in the vector field. This function would help to select the valid x and y displacement limits from a scatter plot and helps eliminating vectors with spurious magnitude.
3. Environmental disturbance, human error such as man-in-the-middle, and/or impurities in the air and fluid medium could possibly lead to erroneous leak detection and trigger false alarm. Both semi-supervised learning, and wherever possible, physics-Informed learning models needed to be developed to leverage data from the extensive sets of images, and build robustness into the detection mechanism. This tool could enable quantitative profile reconstructions to correlate physical variations in the surroundings, data errors and process anomalies, with leakages. This approach could be accomplished by extracting high dimensional descriptors from processing of the recorded images.

Some of the key Broader Impact related outcomes of the project were:
1. Sensor based techniques are currently being strongly considered in competing markets, and hence robustifying the visible-optics based approach with AI tools would offer an advantage to the proposed low-cost diagnostic setup.
2. Indranil Roy, now a doctoral graduate from the Department of Mechanical Engineering and Mechanics at Lehigh University, led the effort as the Entrepreneurial Lead and received formal training through courses in Technical Entrepreneurship and Project Management from the Business school at Lehigh. He also received mentoring through the Venture Well E-team grants program through their Pioneer Workshop for customer discovery and deeper engagement.
3. A provisional patent was filed with the aim of filing a utility patent once the prototype is matured and field tested at a facility of one of  the potential customers’ facility.

Last Modified: 05/14/2024
Modified by: Ganesh Balasubramanian