Quantum Approach will Tame Big Data to Improve Manufacturing Processes

Big data analytics will improve manufacturing processes

Thanks to digital computers, manufacturers have been able to significantly improve their process efficiencies in the last 30 years.  This was achieved by more accurate control of the production processes, raw materials and the process variables.  However in many complex processes, such as fine chemicals, pharmaceuticals, and mining large variability of raw materials are unavoidable facts of life.  The sheer complexity and the number of variables involved in the manufacturing of bio-pharmaceuticals, for example, leads to big variations in end products causing large rejection rates. 

Here advanced analytics can improve manufacturing efficiency by applying various forms of statistical analysis of data to determine the inter-dependencies among process parameters and their impact on the quality of end products.  Analytic techniques can also help in controlling the quality of an end product when the raw material starts to vary as in the case of mining operations.  Most manufacturing companies collect process data for tracking purposes rather than for improving their operations.  Analytics can identify patterns in collected data that are difficult to discern manually.  Thus, augmenting data collection and their analysis can lead to significant paybacks.  

Quantum computing will help in machine learning and optimization

A first step for optimizing a manufacturing process is to find the relationships between cause and effects.  A powerful analytical tool for machine learning is based on topology, which can find the connections between elements in a complex dataset.  However, finding these connections can be too demanding for a conventional computer.  For example, if there are 200 data points, a conventional approach for analyzing all the topological features would require 2²⁰⁰ processing units, which beyond the capacity of any digital computer.  Today, such problems are solved by using various approximations and shortcuts, whereas, a quantum computer with only 200 quantum bits (qbits) will easily be able to tackle such a problem.  

Thus, quantum based approach exponentially speed up an application with a large number of nodes.  The same approach may be used for analyzing many other systems with a large number of interconnections, such as the U.S. power grid, internal wiring of a brain, and detecting terrorist networks. 

Practical quantum computers are still under development.  The only quantum computer that is available today is quite expensive and not cost effective for most industrial applications.  However, quantum computation as a cloud service may soon be available making it viable for industrial and commercial use.  For suitable applications, early adopters will be able to gain considerable competitive advantage.  Analytics companies should include quantum computation in their portfolios. 

Why is quantum computing so special?

In a classical computer information is encoded in bits, which can either be a 1 or a 0, or “on” and “off.” Quantum computers use qbits (quantum bits), which take advantage of superposition and entanglement.  Superposition means that each qbit can be both a 1 and a 0 at the same time. Thus, three conventional bits may store a single number from 0 through 7, but 3 qbits can store all the numbers 0 through 7 at the same time.  Entanglement means that qbits can be correlated with each other.  That is the state of one qbit may depend on the state of another entangled qbit.  This allows quantum computers to create switches that are more complex than simple on/off, allowing them to process data quite differently from a classical digital computer.  The concepts of superposition and entanglement are counterintuitive and are initially difficult to grasp.  However, theoretical basis of quantum mechanics have been postulated in early 20^th century and are now well proven. 

Today, D-Wave is the only supplier that offers quantum computers for commercial and research purposes.  However, they are priced at around $15 Million each, quite out of reach for most users.  IBM, Google and others are active in this field. IBM is offering a cloud service for a small system with a few qbits to researchers and hobbyists. 

About the Author

Asish Ghosh is a Control Systems Engineer with over 40 years of professional experience.  He held various research, engineering, and consulting positions while working for ICI in England and for The Foxboro Company, and ARC Advisory Group in Massachusetts.  His recent publications include “Dynamic Systems for Everyone – Understanding How Our World Works” a book on system science for engineers and general public.