(69a) Designing Efficient Human-Machine Interfaces for Decision-Support Tools: Case Studies in Healthcare and Process Systems

Human-machine interfaces link human decision-making with the underlying processes, measurements from the process, and the algorithms used for data analysis. They enable humans to interact with complex systems and make sense of the information they provide, allowing us to make informed decisions. The type of data feedback dramatically influences our decision-making. Incorrect decisions increase risk exponentially, specifically for safety-critical domains like healthcare and process systems. This article presents the design of efficient human-machine interfaces required at different stages of developing decision-support tools in these two domains.

Healthcare: We focus on Type-1 diabetes, a condition where the pancreas ceases to produce insulin requiring external insulin input to control blood glucose. Developing decision-support systems can ease the cumbersome task of continuous monitoring by patients. However, comprehensive context-driven data is needed to develop the algorithms required for decision-support systems. Required data can be collected from patients or caregivers in free-living conditions. To make this task effortless, we designed a human-friendly interface for daily activity (eating, sleeping, exercising), logging on to smartphones, and collecting accelerometer data from smartwatches.

Process systems: This case study focuses on small and medium-scale food industries lacking digitalization and real-time quality monitoring. To address this issue, new sensors were installed on the output line to monitor and collect real-time product stream data, improving efficiency. The human-machine interfaces were designed to provide clear and easily understandable information about the sensor output and decision support if necessary. The data transfer was enabled using cloud platforms, specifically Smart Manufacturing and Innovation Platform (SMIP) developed by the Clean Energy Smart Manufacturing Innovation Institute (CESMII).

In human-in-the-loop decision-making, one must understand how humans respond to safety-critical operations like start-ups and shutdowns. A plantwide dynamic process simulation of vinyl acetate monomer production is developed, with the human-machine interface providing the necessary inputs to the operator. The simulation allows for safely testing various scenarios and provides an opportunity to study the operator's responses and understand human-in-the-loop decision-making. The human-machine interface is designed so the operator can get decision support during critical start-up and shutdown operations. By tracking operator responses, guidelines for efficient interface design can be developed.

The gap between industry and academia has long been reported and discussed in the literature. As a step toward reducing this divide, a pedagogical human-machine interface and guidance system are developed for a heat exchanger used by undergraduate students. The system uses cloud platforms for sharing data. The interface is designed to help students understand the experiment process dynamics by providing multi-modal information about the experiment in real time. This real-time understanding of the process dynamics helps students make correct manual decisions for successful experiment learning. Furthermore, the interface enables interaction and synthesis with student-designed fault detection algorithms. This provides a realistic setting to deploy algorithms for real-time decision support and better prepares them for careers in industry, where they will encounter similar challenges with real-world equipment and data.

Briefly, the proposed presentation aims to educate viewers on machine interfaces in human-in-the-loop systems, including the importance of considering the needs of individuals based on their level of expertise and modifying designs accordingly.