Data Collection for Energy and Resource Monitoring
I mentioned in my last article that I was attending a manufacturing conference in Italy the end of August and that there was a lot more discussion about some aspects of green and sustainable manufacturing – at least efficient use of energy.
This is supported by business surveys and comments in the business press reflecting, I assume, the interaction with business folks “in the know” on such matters. A recent McKinsey special topics report titled “The next environmental issue for business” gives some interesting statistics on what matters most to business. The report actually was focused mainly on biodiversity and the importance that holds in the minds and hearts of business. We can get back to that topic in the future (and read the report…it is interesting).
I was intrigued by the more general data given in the McKinsey report on issues of importance to business (and based on a responses of almost 1600 survey takers). The top vote getter was “climate change/energy efficiency” coming in at 43%. Next in line was “waste/pollution/recycling” with 42%. Following that was “water scarcity/water quality/sanitation at 27%. There are 10 other categories of issues ranging from data privacy to global public health. And, “biodiversity” was 10th on the list. Another “environmental” related concern was toxic materials at 14%. (Note: the respondents ranked a number of issues; so, the percentages will not add to 100.)
I was pleased to see the top three as close to our topic of green manufacturing since they deal with, in order, energy we use and its impact, things we throw away/waste and things from the environment used to make our product besides energy – here, water.
Water is often overlooked in all the concern about energy, but not by everyone. Caterpillar has a goal of “hold[ing] water use flat” as they increase their business listed in their 2009 Corporate Sustainability Report. The website gives a short discussion of Cat’s plan to determine the “true cost of water” and includes the following statement:
“Without good data it is impossible to justify the cost of water-saving initiatives.”
They go on to explain how a program in 2009 at one of Caterpillar’s American plants launched a program “to quantify how much water it was using in its different processes, and the costs associated with water use in each process – including water bills, chemicals, labor, maintenance and energy. The project helped the plant identify its most expensive water processes and associated costs and justified the capital expenditure needed to implement savings.”
They plan to extend this program to other Caterpillar facilities in 2010.
Good data … and plenty of it. Ditto for energy, other resources, etc., throughout the factory.
In the manufacturing conference in Italy I attended, the CIRP General Assembly, I presented a paper co-authored with one of my recent graduate students, Dr. Athulan Vijayaraghavan, titled “Automated Energy Monitoring of Machine Tools.” The full reference is “CIRP Annals – Manufacturing Technology 59 (2010) 21–24.” (Let me know if you’d like a copy.)
This paper laid out the immense challenges associated with trying to acquire, store and process the streams of data from a variety of machines in a variety of systems throughout a variety of factories. This is done in the hope of, first, understanding where energy (in this case) and other resources (like water) are used and then how to meet the kind of goals the Caterpillar folks are aiming at. This means understanding the nexus between process operation and resource use to be able to find ways to minimize the use per unit of output. That is, decouple the process and resource equation so we can effectively reduce the “impact/GDP” to reduce overall impact of manufacturing.
We focused on only the machine tool…but the approach can be extended much more broadly.
If you think about making this “connection” between resource consumption and process, you need to first determine the rate of data you need to make the link. For energy, this can range from parts of seconds to hours.
Earlier this year (January 21, 2010 to be exact), I wrote about “temporal vs spatial” aspects of manufacturing. We can create a similar diagram to illustrate this discussion of data rate
demands for tracking energy and resource use. The figure highlights the data rates for machine tools but, for broader sections of the enterprise, you can see the time scale also. The idea is you need sufficiently high data rates to capture the process effects or variability you are trying to associate the use with. Then, we can see how adjusting those parameters or variations can yield savings (without, of course, sacrificing quality or cost).
Here is another illustration from the paper showing the use of energy in the context of the manufacturing process. The objective is to have data rates “tuned” to the process so one can extract such
Time scale of data collection for energy use in the context of manufacturing process
– energy usage per day (lot or batch basis),
– embedded energy during manufacturing a part (piece basis),
– energy used for value-added and non-value-added activities (between productive operations),
– relationship between spikes/troughs and process parameters (details of process operations),
– impact of process parameters on sub-component loads (what’s going on around the machine or line), and
– energy used for machining specific part features (related to part design/geometry and functionality.)
This information depends on vastly differing data rates with sampling times varying from milliseconds to minutes.
The data volumes can be impressive. In the paper we give a sample of the number of energy data values for a “medium sized facility” for a day. This facility is comprised of 25 CNC machines, 10 programmable logic controlled machines and assorted other handling and line equipment. For the CNC machines alone, assuming 8 data sources per machine at a sample rate of 5 hertz (5 times/second), we will have a data stream of 86,400,000 data points each day. If we add the other sources, with reasonable numbers of data collection sites and data rates, we would likely end up with over 100 million data values a day to deal with.
So, what’s the solution?
The paper proposes a structure for, first, standardizing data (for example using MTConnect), implementing a modular, scalable architecture that supports multiple concurrent data streams and sources and, importantly, employs multi-dimensional reasoning tools.
There is more to discuss on this but this is more than we can cover in one posting. I’ll finish the discussion next time.
David Dornfeld is the Will C. Hall Family Chair in Engineering in Mechanical Engineering at University of California Berkeley. He leads the Laboratory for Manufacturing and Sustainability (LMAS), and he writes the Green Manufacturing blog.
Energy Manager News
- Clauses to Consider in Green Leases
- Bahama Yacht Club to Generate Power from Solid Waste
- Duke Energy, USF Launch Solar Battery Research Initiative
- Energy Storage Helps Hotel Reduce Demand Charges by 10%
- EU Smart Campus Pilot Achieves 30% Energy Savings
- Uline to Operate 130 GenDrive Fuel Cell Units from Plug Power
- Los Angeles Shopping Center Installs 504 kW Solar
- SustainCo Wins $575,000 Contract for Energy Management Controls