SMART DOG MINING^TM
It takes a smart dog to find hidden treasures

Big Data Collection and Use

Data is only as good as how it is collected

I am not against the concept of “Big Data”, but I am against the concept of collecting data just because you can.  The data does need to have a use.  I take this position because you also need to verify and confirm your data.  Also you have to monitor and maintain the instruments and systems that collect that data. Without this last point you can get into some unusual issues.

 I ran into an interesting case where lack of verification and maintenance cause data to be in conflict with itself.  I was helping a plant solve a problem with the draft on a rotary kiln.  Two of the engineers were arguing over some data that showed a pressure spike in the duct that seemed related to the upsets.  One was reporting the spike occurred right after the upset, the other was saying right before the upset.

 Both had graphs showing their information.  The arguments got going strong, when it was suggested to bring up the information on the conference room screen.  First one brought up his then the other brought up his, and they were in sync approximately five minutes apart. But it was then noticed that they were both looking at the same instrument reading.

 Shall we say that the consternation was a bit high with several cries of “that’s impossible”.  The I&E engineer and techs said they would look at this immediately and report back the next day.

 Turns out that two separate PLC’s were reading and recording the same data point.  The internal clocks of both were off, one two minutes fast, the other three minutes slow.  The spike was actually occurring right at the upset.

Oh and it turns out the problem was a bad bearing mount on the exhaust fan.  Found by a mechanic who was walking by just as it squealed and the fan stuttered.

Data is only as good as how it is analyzed!

A recent story in the Wall Street Journal (1) highlighted this, where a piece of large mining equipment (a Joy Continuous Miner) was having problems with one of its drive motors.  The operation thought that a control system was at fault.  Repairing it was a major effort.  The equipment vendor (Joy Global) had access to the data remotely, and was able to indicate that the problem was a heat exchange unit, which was much easier to repair.

As I have said before, I am not against the concept of “Big Data”  but I am against the concept of collecting data just because you can.  Once the date has been collected it needs to be collated, analyzed, and summarized, or as we said in the pre-electronic media age - Folded, Spindled and Mutilated (2). 

With the advent of improved instrumentation and faster and more powerful data systems, the collection of the data is easier than ever.  And it looks to become easier and easier to collect in the future.

Not many operations have the luxury of a full time staff that can look at and tell what all the data means.  This becomes even more the case with large and special equipment.  The operation knows how to run it and maintain it (hopefully).  But may not be knowledgably in what some of the instrument readings mean.

Making use of outside sources to help monitor and analyze the data is a way for operations to expand their capability with experts that they would not otherwise have. And may become a key decision in future when purchasing new equipment, the ability to remotely monitor and analyze equipment data.

• http://www.wsj.com/articles/big-data-gives-manufacturers-a-new-revenue-source-1433185734
• Folded, Spindled and Mutilated – an old term from the days of punch cards.

Using Big Data: how good is your data

In mining keeping the plant operating on target is important, this means keeping the plant on grade and meeting recovery targets.  Optimization is beating the grade and recovery targets or increasing through put without losing grade and recovery.  Best is to increase through put and increasing grade and recovery.  But to do this you have to know what you are doing and how everything is working.

Many a plant operator has being running along keeping an eye on his instrumentation when the financial people tell him his actual production is off and where is the product he said he made.  At which point he will start digging through his data.

In most modern plants the one thing you probably have a lot of is data.  But is it any good. As I mentioned in some earlier articles, your data is only as good as how it is collected (https://www.linkedin.com/pulse/data-only-good-how-analyzed-mike-albrecht-p-e-) and your data is only as good as how it is analyzed (https://www.linkedin.com/pulse/data-only-good-how-collected-mike-albrecht-p-e-).   But beyond collecting the right data and having the proper tools to analyze the data, is the data actually any good.  But then what does good data mean.

Having good data will deal with the quality of the data. Key terms are accuracy and precision.  As in the picture at the top of this article, your data can very accurate and precise (the desired state) are some combination thereof. While high accuracy and high precision is the goal, you will often be at the other extreme.  Understanding what effects the accuracy and precision of your data can help you understand it.

Having the correct tools to collect and analyze your data is very important, as is understanding the accuracy and precision of your data. 

In general use the two words precision and accuracy are often considered the same  in technical use they are different.   In science, engineering, industry, and statistics, accuracy is the degree of closeness a set of data is to the quantity's true value (or accepted true value.  The precision is related to reproducibility and repeatability, or how repeated measurements under unchanged conditions show the same results. 

Accuracy and precision are often defined in terms of systematic and random errors. The more common definition associates accuracy with systematic errors and precision with random errors.

The system used to collect your data can be accurate but not precise, precise but not accurate, neither, or both.  An example being if your measurement system contains a systematic error (often called a bias) it can give a set of data that is very repeatable and reproduces consistantly but is actually off by a significant amount (low accuracy high precision).  Once this is found, it can be corrected. 

For accuracy we can distinguish:

·         the difference between the mean of the measurements and the reference value, the bias. Establishing and correcting for bias is necessary for calibration.

·         the combined effect of that and precision.

For precision:

·         repeatability — the variation arising when all efforts are made to keep conditions constant by using the same instrument and operator, and repeating during a short time period; and

·         reproducibility — the variation arising using the same measurement process among different instruments and operators, and over longer time periods.

So, what does this mean.  First are your instruments actually reading what you think they are and on top of that are they the right instruments for what you want.  Next instruments are influenced by weather and age, and need to be recalibrated on a regular basis.  Also they do not last forever.   Keeping on top of your plant requires knowing how wil your instruments are doing so keep an eye on them.