Big Data analysis techniques have provided tangible benefits and a definite competitive advantage at IBM's world-class Bromont, Que. facility.
BROMONT, Que.—The semiconductor industry is a highly competitive sector defined by fierce competition and technology that shifts from cutting edge to obsolete in a rapidly shrinking timeframe.
Indeed, in this kind of environment maintaining an edge in innovation is key. This is caveat holds true for small and large operators alike, and there are few as large as IBM.
IBM’s facility in Bromont, Que. employs hundreds of engineers and technicians focused on microchip production and delivery. The plant in southwestern Quebec assembles 200 types of products involving 800 manufacturing processes with a new production batch coming onto the assembly line every three minutes. Last year it posted more than $500 million in revenue.
To preserve its leadership in innovation, IBM’s Bromont leadership has decided its time to tackle the vast amounts of data generated every shift. It integrated analytics techniques and tools to improve problem solving efficiency, optimize its operating conditions and lower inspection costs.
IBM gathers intelligence from data mining and analytic techniques and applies it to various processes in the manufacturing cycle.
Finding the hidden factor
Assembling a microprocessor is similar to assembling a car. It involves about 30 successive stages in which raw materials are put together like a jigsaw. However, problems can occur during any stage.
To identify and offset potential issues, a vast amount of data is mined to collect information on raw materials, the environment, manufacturing equipment and throughput.
Rather than analyze variables individually, data mining combines the variables to highlight a specific potential problem, such as if the quantity of parts in a furnace is impacting solder adhesion.
Being able to identify and correct a problem in a matter of days instead of weeks is crucial to guaranteeing on-time delivery and high quality service.
Simulating corrective measures before applying them
To minimize energy costs the plant uses a fairly wide interior humidity range in order to benefit from fresh air as often as possible.
But many products are sensitive to these variations in humidity. The relationship has always been difficult to quantify because humidity data is not linked to production data, and other factors come into play, which creates defects in product characteristics, raw materials and tools.
Over the course of three months, data was gathered on several factors in relation to two types of defects. Data mining helped determine the conditions under which humidity could have an impact on quality.
In this case, the corrective action to be applied to the humidity range incurred substantial energy costs. Scenarios were developed taking into account volume forecasts and factors identified with analytics. These simulations showed that a minor correction to the humidity range would have a positive impact on one of the two types of defects, generating a 160 per cent return on investment (ROI) in the first year.
Inspecting with data
The process for inspecting extremely small items, such as microchips, requires substantial financial investment.
Recently, analytics were used to reduce the inspection processes being carried out in Bromont’s labs. The first step was to look at all the parts rejected by an electrical test. A classification algorithm applied to the test results identified parts’ profiles that had similar results. Certain profiles were strongly associated with inspection results with matches exceeding 97 per cent.
Bromont now uses this profile instead of laboratory inspections, thereby reducing transportation of parts, analysis and quality management time—all critical to customer satisfaction. By applying this technique to a number of other products, Bromont anticipates generating a 150 per cent ROI by the end of 2014.
Data analytics techniques are not new. However, massive quantities of data on manufactured products, equipment and the environment is vital to improving quality. These three examples illustrate that analytics help process data to improve and optimize operating conditions and reduce inspection costs. For a world-class plant such as the Bromont facility, data analytics techniques have provided tangible benefits and a definite competitive advantage.
Matthieu Lirette-Gélinas is a junior business analytics engineer at IBM Bromont