A Framework for Teaching Safety Critical Artificially Intelligent Control Systems to Undergrads Honeywell Technology Solutions Lab

Format:

Book

Conference/Event

Author/Creator:

Jeppu, Yogananda, author.

Contributor:

Raman, Ramakrishnan

Conference Name:

AeroCON 2022 (2022-06-02 : Bangalore, India)

Language:

English

Physical Description:

1 online resource cm

Place of Publication:

Warrendale, PA SAE International 2022

Summary:

There is an increasing demand to educate students on systems thinking and systems approaches at undergrad and graduate levels in colleges in India. Efforts are being made by industry, academia, and professional societies to join hands to bridge the gap. Specifically, there is significant emphasis on providing wholistic "live" case studies and examples to students to get their "hands dirty" on actual systems. One of the inhibitors on this aspect being faced, in the aerospace domain, is that actual examples are not available in the open literature as they are considered proprietary and/or confidential. This paper illustrates a framework for educating students on systems approaches and systems thinking in a near "live" scenario through a case of safety critical control system embedded with Artificial Intelligence (AI). With the recent advances in AI and increasing demands on embedding AI in complex aerospace systems, certification of such systems poses many hurdles and challenges. Though AI has been used in aerospace industry for optimization problems for quite some time, challenges in certification of safety critical aerospace control systems embedded with AI are actively being explored in the industry. Experiments on such system are required to clearly understand the verification and safety aspects of deploying such systems. The proposed framework is illustrated using the case of an aerospace environmental control system. The case study has sufficient complexity as a near "live" scenario to merit an industry level attention but is again simple enough to be understood by the students. The first step involves developing a physics based model with adequate fidelity to illustrate the various nuances in control systems beyond the theoretical foundations. Subsequently, simulation runs from this physics model is used to train a Neural Network (NN) model as an observer. The NN model is developed to predict divergence from a nominal system behavior and to raise flags for warning and a subsequent control change. The proposed framework brings together various systems thinking aspects for study and teaching of NN in safety critical domain. The added advantage is that these models and framework will be made available in the open domain for the student community to experiment and learn

Notes:

Vendor supplied data

Publisher Number:

2022-26-0025

Access Restriction:

Restricted for use by site license