Thursday, June 19, 2014

Improvement is a Good Thing – Right?

It is human nature to want to improve and now, more than ever, we have tools to make this easier.  We live in a day and age of data – think of all the information that we have at our fingertips each day, with a simple swipe of a finger on our smart phones,  a powerful combination – we improve and optimize based upon that wealth of data.

But wait there is a snag… Data is not information; without understanding, it often times can lead us to bad decisions.  We have discussed the need for context and timing before [please click the links if you need a refresher].  Today I am going to focus on Metrics and KPIs.

First let me state that Metrics (or Key Performance Indicators – KPIs) can be very good thing.  They are the key sharing your manufacturing goals and vision.  For this reason, they should be visible and aligned to your overall vision.  By adding this visibility, it drives people to focus on the metric and improve.  So why did I use the word “can”?  Metrics in themselves are a tool, if they are not consistent or align to a vision that is a very bad thing!

Take for example Cycle Time, most people try improving Cycle Time – but is it the right metric for you?  Is it a single cell or an entire line?  I have the fortune of observing many different processes and plants and often see areas of local optimization.  Local optimization is where a particular KPI is applied towards a machine and improved, however in the larger picture these are counter-productive.  Back to the Cycle Time reference, it doesn't do any good to run at a faster pace, if the upstream or downstream processes can’t keep up, or worse yet if you are producing the wrong parts (no demand).  If you haven’t done so, I would highly recommend that you read “The Goal” by Eliyahu Goldratt.

Also keep in mind that Metrics and KPIs do not need to be permanent – they are there simply as a tool to affect a change.  Most companies have a continuous improvement initiative, the underlying metrics should change to support new initiatives and evolve over time.

My next post will focus on the methods for getting data and how based upon the maturity of a plant, how data collection and KPIs will evolve.

Thursday, May 1, 2014

Getting from here to there: Data Collection

In my last post I outlined my views on the “Smart Manufacturing” and all the complexities that exist in the high volume manufacturing environments. That post also outlines a need for the “Smart Manufacturing” Platform that includes various technologies to enable Predictive Space – Predictive Maintenance, Virtual Metrology, Predictive Analytics, Dynamic Sampling and next generation of advanced analysis techniques to find the "hidden signatures". 
The logical question is – where to start? All of the above “smart technologies” heavily rely on information – the form of the Manufacturing Big Data. But data in itself is not information; operations or business metrics, events, and comparative analysis need data in the right operational context.  So before tackling issues with “Big Data” we need to ensure “Right Data”.
In the typical manufacturing enterprise, this information exists in various forms, spread across 15 to 50 standalone plant and enterprise systems, all with different data models, schemas and data velocity:
  • ERP
  • PLM
  • APS/Scheduler
  • Production/WIP Track and Trace
  • Maintenance
  • Quality
  • Warehouse
  • Batch/DCS
  • PLCs
  • Process Historians
  • Etc.
Also, there are numerous business processes where relevant information can’t be directly acquired and only exists in people interacting with systems, equipment and each other.
Traditionally, the current generation of deployed plant systems uses a mix of data items:  automated data collections from Levels 1 and 2 consisting of both control sensor inputs and actuator outputs,  manual data collections for Level 2 recipes (unit procedures and phases), and Level 3 production order routings (steps and operations).
Most manual data collections are entered into random UIs on workstations with a large collection of separate HMI and operations management applications.  Thus, the random collection creates non-value-add steps in the operator’s workflow and creates errors.
Piling this data into yet another database will not resolve semantic and timing differences. For effective execution of plant operations and business processes, the data collections within real-time processes of production and support operations must be accurate in term of:
  • Timing for step/operation execution
  • WIP state
  • Common context under a single plant equipment/process model across all operations
The ideal “Right Data” approach shall handle high frequency data, complex data, old (late arriving) data and data that exists outside (not owned) the system, but is needed for decision making. The system should be able to correlate timing information and apply “well known” time that reflects high level execution context. Where needed, the system should infer, combine, and enrich information based on a model and a context of execution. Manual data collection should be system-driven as part of the execution, prompting user for information in the context of a process or activity using least intrusive data entry mechanism available, ideally using single UI.
Having the “Right Data” is the first step on the road to “Smart Manufacturing”.