A casting plant with three identical aging furnaces is operating 24/7 using certain energy consumption and peak demand with data logging logging on 10 minute interval. I am trying to study whether its operating patterns can be deduced. I have used cluter algorithm on the data segregated by peak (8am to 10pm) and non-peak period (10pm to 8am). However, elbow curve showed 2 to 4 variables. However, there were only 2 cluster patterns with the rest having no correlation. The energy consumption is quite steady with small peaks. I am wondering whether there are any other algoritms could be used to deduce the processes.
I have successfully used various means to identify such patterns in plant operational data. These techniques go beyond FFT. There are some examples in my book, Orthogonal Functions (https://www.amazon.com/dp/B07GT8TLDV) I also cover this subject in Curve Fitting (http://www.amazon.com/dp/B01LDUK032). Both eBooks will be free on 8/18 (this coming Tuesday). The accompanying software (and examples) is always free. You can look inside the books to see if you're interested. There's a link to the free software below the Forward (or you can find it with Google). I have attached a spreadsheet using Legendre polynomials to approximate the measured temperatures obtained from four platinum RTDs. I have lots of plant data and spreadsheets and could construct other examples if you're interested in this approach.
Here's a spreadsheet with some 30 second pump data. I have another 200 columns of other measurements from this same time period (temperatures, pressures, current, voltage, etc.). The regression is a hybrid between Fourier and least-squares, using 4 sin() and 4 cos() and LINEST to arrive at 9 coefficients each for temperature and pressure. There are all sorts of tricks available that you won't see in a textbook. You can use SUMPRODUCT(x*y)/SUMSQ(x*x) to project each sin or cos onto the data (like numerical integration) and step through time scaling factors with a button and macro to find dominant frequencies. You can also use the Solver to minimize the residuals (cells O6 and O12) or maximize the R² (cells N4 and N10) calculated by LINEST.
Thank you but sorry for the late response, Dudley. I will update the timestamp and the furnace performance. Wll update you soon on the results.
- Badges
- Science topic
- Similar topics
- Physics
- Condensed Matter Physics
- Solid State Physics
- PEAKS