Rapid Insights and Operational Intelligence for the shop-floor
It is no surprise that the big-data hype is falling into the trough of disillusionment (Lisa Kart, Gartner, January 2015, Big data industry insights) and organizations are struggling to get value from their big data investments. As if to save the day, a McKinsey January 2015 article Getting big impact from big data makes a number of recommendations;
“Visualization tools… are putting business users in control of the analytics tools by making it easy to slice and dice data, define the data exploration needed to address the business issues, and support decision making.”, writes David Court. Earlier in the article he says “…analytics specialists builds models targeted to specific use cases. These models have a clear business focus and can be implemented swiftly.”
In my own work and those with our customers, I find self-service data analytics using models targeted to specific use cases is key to rapid insights – whether it is small data or large data, or even big data. And, an enterprise wide MOM (manufacturing operations management) infrastructure that collects data and events along with the appropriate semantics is a critical foundation for operational intelligence.
V is for Value
Our industry is abuzz with all the promise of big-data and related technologies such as Hadoop, cloud, NoSQL, in-memory database, complex event processing, and others. We all recognize big-data which is often described by the three V’s – volume, velocity, and variety. Some add a fourth V for veracity and even a fifth V for volatility. But seldom is there a mention of the most important V – value. What’s the new insight that brings value?
As such, this blog is not about big-data enablers such as Hadoop and map-reduce. Nor is it an academic overview of the different types of data such as structured vs. unstructured, data-at-rest (historical data) vs. data-in-motion (streaming data), relational vs. NoSQL, and the likes.
Here the focus is on getting insights from small and large datasets with tools that are practical and possible today such as:
- Self-service BI (business intelligence) aka operational intelligence for MOM
- Visual data discovery
- Data mining and advanced analytics
- Volume - large volumes of data; hundreds of thousands to millions and billions of measurements from sensors and automation controls
- Velocity - high velocity data; each sensor collecting data every few seconds, and some with millisecond and microsecond precision
- Variety - from the four key MOM systems covering production, quality, maintenance and inventory; time-series sensor data in production, relational data as work-management transactions in maintenance, lab sample values and free-form text/notes in quality, and others.
When my daughter was one year old, I still remember holding her in one arm and balancing the phone with the other while talking to my travel agent to book airline tickets. Today, she is in high school, and I do my own travel arrangements on Expedia – all self-service – via the internet. I ask Expedia to show me the cheapest way to fly from Boston to Miami; and, then I ask it to show only non-stop flights. And, then filter the results to show flights leaving after 3 PM, and so on.
Why can’t I ask such questions of my plant floor data? In an interactive, ad hoc mode?
Where is my process spending time? Show me –by product, by process cycle, by day, by month, by year…
Where are the downtimes? Show me – by machine line, by product, by day…
What was the best production run? When was it?
An example rapid insights using self-service BI and data analytics comes from Tate & Lyle. It shows how a process engineer engages with MOM data. And, how he can step through an end-to-end use case to locate and eliminate an unnecessary pause of about 20-30 seconds in Step 7 of a manufacturing process. The savings for 250 cycles per day easily translates to about an hour of additional daily production time.
At first glance, yes, these are just charts and trends. The key difference is the interactive user-experience and how quickly you can do this without it becoming a long IT project. The MOM data infrastructure itself provides the foundation to not only collect your process and machine data but also connect them. For example, provide the relevant context to use with self-service BI tools - now available from several vendors. Here is a video of the end user explaining the process. Towards the end of the talk, he explains the money slide with estimated financial savings.
In subsequent posts, I plan to highlight the use of data analytics and related topics in MOM for other manufacturing verticals. Until then, I ask…
What has been your experience, big-data or not, in using self-service analytics for MOM data?
Gopal has been involved in several roles at OSIsoft and has been working with the PI System since the mid-1990s, in software development, technical and sales support and field services. Attached to the Philadelphia office, he is currently a Solution Architect in the Partners & Strategic Alliances Group. Previously, he was a Product Manager with a focus on Enterprise and Asset Integration and PI System data access.
Gopal has a Master's in engineering, continuing education in business administration and is a registered professional engineer in Pennsylvania. He is also active in the MESA Technical Committee and the Education Committee and the MESA Continuous Process Industry Special Interest Group and active in topics such as data mining, energy efficiency, manufacturing intelligence, sustainability, including green and sustainability initiatives in facilities and data centers.