QMT Features: November 2012
Quality and Big Data
How can Big Data analytics improve product quality and reduce product recalls? Nitin Narkhede, general manager, Technology Strategy and Innovation, Wipro Technologies, explains.

Recently, reputable companies have been in the news for voluntary recalls of products that are manufactured between certain dates. The big question most consumers ask is “why don’t the manufacturers know that there is something wrong with the parts or the way the product is put together before they ship the product?”

Every manufacturer knows that product recall increases engineering and manufacturing costs, affecting the bottom-line. In spite of the sophistication of today’s manufacturing industry, according to the Consumer Product Safety Commission (CPSC)’s annual report for 2011, there were 412 voluntary recall cases in the US alone last year, and the US government recovered $3.26 million in penalties from manufacturers for knowingly failing to report problems.

In the past, the concepts of quality measurement, tolerance, quality control and statistical process control (SPC) meant dealing with simple upper and lower control limits for which a process is monitored and controlled. There were hand gauges and tools measuring each dimension and plotting control charts in the factory. Consider what it takes to build some of today’s complex products; products with very high part counts like cars and aircraft, or food, beverages or consumer packaged goods manufactured to match the speed of demand using continuous process manufacturing. With modernization and automation in manufacturing, inspection and logistics, all the machines and objects in factories are intelligent in terms of their ability to support data capture, storage and transmission almost in real-time, offering a new level of process control.

The quality of a product is improved by measures such as: properly capturing design intent; manufacturing process quality; and rigorous testing methodology. In today’s fast paced world enabled by technology, both process quality and in-line testing are part of shop floor operations, enabling manufacturers to study product characteristics in real-time before the product hits the market, avoiding costly product recalls. 

For example, consider a manufacturing facility building automotive engines. A typical engine consists of over 500 parts that are either manufactured in-house by the OEM or supplied by various suppliers.  On this factory floor, there are lines machining major components as well as assembling entire engines. In a scenario where one of the assembly station’s operators has received about 10-15 parts that he is assembling to the main engine and the next station is performing leak test on the engine, real-time analytics would enable immediate leak test results and letting the operator study the root cause of any issues. This would enable to operator to make changes – adjust the machine speed or feed for one of the subcomponents on a machining centre in another work cell that is causing a leak – in a few minutes, rather than the days that it may take otherwise.  Being able to study process variations such as changing robotic arm movements for automated lines to adjust as per feedback from test results or adjusting the sensor on a conveyor to change the location of the next item on the assembly line in real-time along with their impact on the overall system performance enables manufacturers to ensure better product quality and process control.

In the food and beverage industry, all processes - from the cleaning-up of food items, inspections, adding preservatives and air tight packaging to shipping – are automated. The speed of operation is so high that if there is a delay in identifying a food quality or a process issue (e.g. machine issues for packaging), the manufacturer would have already shipped multiple batches of faulty items. This would mean investing a great deal of time and money on tracking, recalling and disposing of the items from store shelves across the country, as well as impacting the brand image, which could also affect the top-line.

The value of controlling any process in real time for quality is clear. In order to do so, a manufacturer needs to create an automated closed loop decision making process that relies on the analysis of data from a combination of data sources such as product data, multi-dimensional tolerance simulation DOEs and unstructured machine data related to monitoring and control of machines, inspection equipment output, control system process calls, dealer data, warranty data, field quality data etc. (See Fig 1).

Dealing with product and process complexity, production volumes and a number of data sources that can provide useful data from the product quality perspective, could mean processing large amounts of data. Collecting, storing, moving and processing this data can become a challenge in itself in terms of infrastructure, network speed and processing capability. Over the past few years, the availability of increased processing and network performance as well as storage capacity at a lower price point is helping to address these issues. However, software is another story. While the emergence of big data technologies is promising, a number of software technologies need to be integrated to create a framework for a closed loop system as described above. Fig 2 shows an example of such a framework.

Commercial machine data platforms like Splunk have the capability to perform machine data analytics in real–time, correlating the machine control data with structured and unstructured data such as testing data and product tolerance mode, service data and field quality data. This provides actionable insights to enable operations managers to take decisions on improving product quality in matter of minutes, as well as influence any engineering changes if necessary. Depending upon the capabilities needed, additional software products may be required to achieve the desired results.

In future, end-to-end visibility and actionable insights with big data analytics and quick feedback control of the overall manufacturing operations as shown in fig. 1, will improve overall product quality in the market to a whole new level and reduce the need for large scale product recalls.l
email: nitin.narkhede@wipro.com
You can now view all QMT Magazine issues on your favourite tablet or smart phone.
Download the free Quality Manufacturing Today App from the Apple iTunes App Store or from QMT Magazine on Google Play.

Rob Tremain Photographer
Click above to see full page display and links to QMT articles.
View Olympus video>>
Vision 2016 ad
Control logo
Nikon logo