analytics for manufacturing

Operational optimization requires improving manufacturing efficiency with data science today. Large corporations are striving hard to consistently make 20% or more of their products totally new or substantially revamped. According to Harvard Business Review, while the need for innovation has continued to grow, the cost of innovation, especially in manufacturing, also has increased. One key driver for this need is increasing energy costs.

With higher fuel costs, there is higher demand for building light weight vehicles. But safety concerns dictate that the new lighter weight designs should be equal (or better) in strength to their heavier counterparts. So a trivial solution such as using thinner gaged metal cannot do the trick. What is needed is to build better quality parts by using advanced manufacturing processes. One example is hot stamping.

Hot stamping process yields parts which are more than 300% stronger than conventionally produced parts (cold stamping!) and have other engineering advantages as well. With three times stronger material, engineers can now choose thinner sheets of metal to build parts and thus save weight.

One mid-sized automotive supplied of stamped parts indicated that in their business they expect the use of hot stamping, to grow to nearly 25% of all vehicle structural parts as opposed to 16% today. However hot stamping is heavily capital intensive and investment choices must be made cautiously. Higher process efficiency is required to obtain good returns on investment.

This is where improving manufacturing efficiency with data science can become critical. Using analytics tools, we can identify which process factors are most critical for improving performance. The performance objective can be anything, here are three examples:

  • Increasing production throughput
  • Reduce energy consumption through better use of waste energy
  • Reduce tool fatigue and wear

If there is enough good data, identifying the most important parameters is a natural step one takes before building predictive models – it even has a name – feature selection. Fortunately data availability is not a problem for process industries. Big data is making waves today in many industries, but manufacturing has traditionally been drowning in data for a long time! However, intelligent use of data requires careful attention. Once the data is identified, predictive models can be built to explore a variety of scenarios that can improve process efficiency.

Many automotive suppliers today who make specialized parts are medium sized businesses. They cannot necessarily support large analytics practices and afford expensive software systems. This is where relying on open source technology and cloud computing can play a decisive role in bringing the innovation of big data analytics to these smaller manufacturers.

Originally published on Wed, Nov 16, 2011 @ 02:29 PM

No responses yet

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

simafore.ai