Our Sites

Measure it, graph it, control it

You cannot control what you do not measure

Pareto diagram figure 1

Figure 1 Click image for larger view A Pareto diagram shows that in this ERW mill, nearly half of the manufacturing delays are associated with the hydrotesting step. The other 50 percent are associated with the welder, cutoff, coil splicer, uncoiler, changeovers, and miscellaneous causes.

You have probably heard of statistical process control (SPC), and you probably use it as a tool for quality assurance or quality control. Developed in the 1920s by physicist and mathematician Walter Shewhart, SPC has other uses too. Specifically, it can help to improve productivity and therefore profitability.

Manufacturing tube or pipe so that it conforms to a particular manufacturing specification isn't as easy as it sounds. For example, if you produce electric resistance-welded (ERW) tube and pipe for energy applications, you probably conform to a standard specification and rely on documented mill control practices. However, specifications such as API-5L and API-5CT from the American Petroleum Institute (API) and those of ASTM Intl. (formerly the American Society for Testing and Materials) are not detailed enough, complete enough, or comprehensive enough to form a complete pipe mill quality program. These specifications cannot take into account the characteristics and physical condition of each individual mill, so you must use documented mill control practices such as standard operating procedures (SOPs) and quality control checks and procedures.

Using SPC for improving productivity is a matter of developing a big-picture strategy in three steps: collecting data (taking measurements); arranging and presenting the data (charting) in a useful manner; and controlling (improving) the individual processes that go into tube and pipe production.

In this article, the word measure is synonymous to collecting data; the word control is interchangeable with the word improve; and charting indicates a visual display of the data collected.

What Should You Measure?

Many individual processes in an ERW tube and pipe manufacturing facility can be measured, so the first order of business is to choose which process or processes need your attention. Be aware that it's pointless to attempt to develop charts for every characteristic of the tube- or pipe-producing process. It would be too large a task and result in too much data.

Select It. Your mission is to select information directly (or closely) related to the mill's operating time, throughput, and efficiency. Any element of any process that contributes to the mill's uptime, downtime, material yield loss, or running cost is a candidate for measuring and charting.

Select an element in the process that is troublesome and develop a way to control or improve this element. Be aware that you might not be able to measure the characteristic directly. You might have to find a related one that you can measure, and use this as a proxy for the element you want to measure.

Collect It. You should develop and maintain a historical record for each product—size, grade, associated tooling and mill settings, weld power, customer name, and so on—so each record is available for review and duplication (if needed). Complete records also are necessary to facilitate improvement.

Analyze It. It is imperative to make decisions and take action based on the collected data, but only after you determine that the data collected is representative of typical conditions and actually reveals the necessary facts. Accurate data and careful analysis are necessary for any control system to be effective. Inaccurate or incomplete data can lead to errors in two ways—it can show a change even though the process is stable, or it can show stability even though the process has changed.

It is vital that you act on data and not on intuition. It is not acceptable to arrive at the solution to the problem until data has been presented, the problem has been defined, and every member of the team has had an opportunity to voice his thoughts.

Pareto diagram figure 2

Figure 2 Click image for larger view This Pareto diagram shows a huge improvement. The number of manufacturing delays fell from 100 to 40.

It is disheartening to enter a discussion concerning a specific problem and, before the problem is defined and the data reviewed, one individual begins the discussion with his solution to the problem in question. When one person attempts to make changes solo and there is no basis for the effort, the result is predictable: A large amount of time and energy wasted.

Tracking Time

In any repetitive manufacturing process, it is possible to establish a standard production rate. If raw material flows continuously to the unit, if the finished product does not back up on the outflow table, and if the equipment experiences no mechanical or electrical failures, production runs steadily.

To determine whether every machine runs at the standard production rate, the operator must track every instance of downtime and the cause. Management establishes the smallest unit of downtime that the operator must record.

Downtime Record for an ERW Mill. Most ERW mills have skelp accumulators so the mill can run continuously. A standard amount of downtime is from 4 percent to 12 percent of the total operating time. Operators must keep accurate downtime records.

A Pareto diagram, correctly constructed, shows that relatively few causes or events lead to a disproportionately large amount of downtime. This chart is known as a method for distinguishing the "significant few from the trivial many." The importance of this statement is that for a normal tube or pipe mill, about 20 percent of the problem areas represent 80 percent of the cost impact. Stated another way, approximately 20 percent of the problems deserve 80 percent of the available attention. Most of the remaining 80 percent of the problems either do not have an assignable cause and would be very costly to evaluate, or these problems occur only by chance and do not have a meaningful solution.

Figure 1 and Figure 2 are Pareto diagrams that show before and after conditions in a typical ERW mill.

Control Charts

If the frequency distribution chart is a snapshot of a process, then a control chart is like a movie, a continuous series of time-related pictures of the same process. It is a running record of a specific process. Control charts come in two types: X-bar and R. An X-bar chart provides the average of the measurements, and an R chart shows the range of the measurements of the process. Figure 3 shows the variations in skelp width, the average skelp width variation, and the skelp width variation range.

A control chart for any process has a lower control limit and an upper control limit. If the control chart tracks specifications—the weight (mass) of the tube or pipe, the diameter, and the material thickness—plant SOPs and applicable specifications indicate the upper and lower control limits for the listed characteristics. Figure 4 shows variations in skelp edge condition graded from A (best) to E (worst).

Visual Illustrations

If you are not familiar with statistical methods, you might say, "This is all well and good, but will it work in my plant?" The answer is "Yes!" The control chart and many other statistical tools are suitable for any measurable or quantifiable process or characteristic.

Many styles and types of charts can help to illustrate data. Listed below are a few of these charts and the way in which each one may be used:

  • Pareto diagram: Determines which elements of the problem unit are the most troublesome.
  • Data collection chart: Shows the information that is available already and the information that needs to be recorded.
  • Cause-and-effect diagram: Determines which variation of a factor is likely to be important and which is not.
  • Histogram: Quantifies variabilities. Are the specifications and procedures meaningful? Or should you change your expectations?
  • Scatter diagram: Determines the correlation (if any exists) between two variables.
  • Time sequence chart: Reveals whether the results change significantly over time. Bear in mind that a time sequence chart can track gradual, continuous changes over time or abrupt changes that occur at specific intervals, such as shift changes.
  • Control chart: Shows how tightly you can control a variable, the acceptable limits, and when the variable changes.
  • Frequency distribution chart: Indi-cates that you need to take corrective action and measure and record the action taken.

The charts listed here are available from the American Society for Quality (formerly the American Society for Quality Control). Many manuals and textbooks are available that have complete instructions on how to develop and use these charts.