(195a) Data Science in Process Automation: Evaluating Economic Advantage

Each day businesses make decisions and take action as best they can based on available information and in the face of uncertainty.  The goal of data science and analytics can be seen in terms of improving the amount and interpretation of information such that uncertainty can be reduced and quantified and better decisions and actions may be taken.

This goal – while agreeable – often raises new questions:  How do we select the right data science problem that will return value for our efforts?  How do we determine the performance required of a data analytics algorithm for it to be economically beneficial?  How can data science techniques complement and augment traditional engineering techniques – and avoid having machine learning algorithm return simple mass-energy balance or automation configuration rules?

Honeywell provides an integrated suite of tools to monitor and automate many aspects of plants in the process industry – including refining, oil and gas, pulp and paper, chemicals, power generation, and mining.

In this presentation we will describe an approach in which the economic costs and benefits to the business can be evaluated for given data analytic solution.  This economic framework can be introduced early in the data science analysis process and has proven to be an effective tool in guiding and engaging developers and stakeholders throughout all phases of the project – setting the question, exploratory analysis, model building, communication, and decision making.