Steve Beeler

You have a goal…I have a way to get you there.

Control Charts

Lake Michigan Water Levels

by

Lake Michigan shore line

Record high Lake Michigan water levels and dune erosion have been big concerns along the “north coast” in recent years.  As beaches disappeared under the rising water, homes near the shoreline were threatened.

Barge with crane on Lake Michigan

Property owners resorted to armoring the coastline with huge boulders to dissipate wave energy.  In some cases, cranes on barges were required to place the boulders.

I finally got around to doing a statistical analysis on Lake Michigan water levels using data from the United States Army Corps of Engineers.  For years, control charts have been in my operations engineering toolkit.  Why not apply statistical process control to a natural phenomenon?

An X-bar & R chart is commonly used to measure the magnitude of common cause (random) variation and to assess statistical stability (constant probabilities).  For a sample plan, I used one data point per year in subgroups of five.  That is, each data point on the X-bar chart is an average of five years and each data point on the range chart is difference between the highest and lowest value in that five-year period.  With 103 years of data (1918 to 1921), this sample plan provided twenty data points, the minimum required to calculate control limits.

Control chart of Lake Michigan water levels

The range chart (bottom) is in control.  All points are within the control limits and there are no trends.  This means the magnitude of year-to-year water level variation has been constant over the last 100+ years.

On the other hand, the X-bar chart (top) does not exhibit statistical control.  While it is mean centered with no trends, there are multiple points (4 above and 6 below) outside the control limits.  This suggests that the five-year sample plan has underestimated the range.  In other words, the within subgroup variation did not entirely capture the common cause variation in Lake Michigan water levels.

In retrospect, this is not surprising.  While 103 years seems like such a long time, it is not even an instant from a geologist’s perspective.  A sample plan with a longer duration appears to be called for but then there would not be enough subgroups to calculate control limits.

Given that the range chart is stable, I suspect that the X-bar chart (i.e., average water levels) would be too if we had data sampled over a sufficiently long period of time.

Others are interested in the Great Lakes.  Enough, apparently, to warrant a scientific publication, the Journal of Great Lakes Research.  In it, I found an abstract that supports my hypothesis of stable Lake Michigan water levels: “Historical Variation of Water Levels in Lakes Erie and Michigan-Huron” by Craig T. Bishop.

From this abstract I learned that the earliest “reliable” Great Lakes water level data were recorded in 1819.  From historical and archeological evidence, Bishop concludes that “… over at least the last 1,800 years, climate-related variations in maximum mean annual water levels have probably not exceeded those measured on Lakes Erie and Michigan-Huron since A.D. 1819.”

That sounds like Lake Michigan water levels have been stable for a long time.  In 2022, the water is down from record levels and beaches are reappearing.  The spectacular sunsets never left.

Lake Michigan sunset

If you would like to perform your own statistical analysis on Great Lakes water levels, click HERE for a link to the United States Army Corps of Engineers data set.

Great Mementos

by

Shelter in place is providing time for many projects that would not get done otherwise. While decluttering, I found some great mementos. Not just souvenirs, mementos tell a story.  Here are four COVID-19 decluttering favorites:

Mementos Steve Beeler with Dr Deming circa 1989

During the late 80’s, Ford (and the rest of the global auto industry) was desperately trying to catch the Japanese on quality. Part of that effort was to put problem solving resources into assembly plants.

I was on one of the plant quality teams, assigned to Louisville Assembly Plant. We received extensive training in statistical methods: SPC, DOE, etc. During a training off-site at the Dearborn Inn, W. Edwards Deming stopped by for a visit. Yes, that Dr. Deming.

After an impromptu talk, photos were taken. I am just behind Dr. Deming’s right shoulder. On his right are David Kho and Steve Redding. To Dr. Deming’s left are Ben Monhollen and Jim Dottavio. I think that is Dave Johnson at the left edge of the photo.

Mementos Steve Beeler LAP Process Capability Reviewe circa 1991

The next two mementos came from Q1.  Ford’s Q1 quality award was a well-conceived program to “stretch” plants to improve quality methods and systems. A major Q1 stretch was to demonstrate process capability on (I think) 50 significant characteristics. An SC was something important to our customers: dirt in paint, door efforts, box-to-cab fits. Voice of the customer.

All SC’s were documented with flow charts, fishbone diagrams, control plans, reaction plans, and control charts on SPC boards near the production operation. The task was to first achieve statistical control and then reduce variation. The higher the Cp/Cpk, the more points toward Q1.

Here’s a photo of the Brakes SPC board during Louisville Assembly Plant’s process capability review. The SC’s would have been brake pedal travel and parking brake pedal effort.

I am just right of center in the red tie. John Coleman is behind me. It’s Lanny Vincent in the short sleeve shirt looking in. I think that is Vera Linnansalo to his left. It looks like Steve Redding in the light blue shirt in the back with Stu Kendrick mostly hidden behind him.

Mementos Steve Beeler LAP Ranger Box-to-Cab Fit circa 1990

Here’s a summary sheet from the Louisville Q1 Process Capability Review. This SC, Ranger Box-to-Cab Margin Left Side, earned 13 points towards Q1 with a Cp/Cpk of 1.35/1.28. In the background of the image, you can see the Ford Blue SPC board.

I don’t remember the two operators in this photo but I can remember others on the team like it was yesterday: Larry Graham, Bob Bearden, Frank Kindrick, Ed Atherton. Good people, all of them. It was a busy time…we were working towards Q1, preparing for the initial Explorer launch, and building 87 (!) Rangers and Bronco II’s per hour.

Mementos Steve Beeler ISO 9001 Pocket Guide

After all the plants were through Q1, the next stretch was ISO 9001. Somebody convinced Alex Trotman that ISO 9001 registration was a perfect fit for his “Ford 2000” global reorganization. A letter was signed and the two-year clock started ticking.

I found myself on a very capable team responsible for the initial registration of 31 North American stamping and assembly plants.  Project management?  This was it.

A big challenge (and there were many) was communication. How to get one message to tens of thousands of employees?

Building on lessons learned at Oakville Assembly Plant’s single site ISO 9002 registration, Julie Trosen designed a pocket guide. I wrote the content, a “Cliff’s Notes” version of the Vehicle Operations quality manual. Don Riker sold the concept to senior leadership. Carlos Filio translated it into Spanish for our Mexican plants. A PO was approved for 100,000 pocket guides.

The story continues. One by one, other Ford activities adopted the pocket guide…and soon it was global! I found nine versions in a hanging file. I am sure there were more.

And the 31 plants were registered in only 21 months…whew!

Great mementos, to be sure.

Source Inspection for Zero Defects

by

As the highest form of error proofing, source inspection for zero defects is such a powerful concept that it merits a close look.

Source inspection is 100% inspection, not statistical sampling, for causal factors that lead to defects. Defects are prevented through immediate action to remove causal factors. While this may sound complicated, it is not. In fact, examples of source inspection are everywhere around us.

source inspection for zero defects

The logic behind source inspection is pretty simple. Before completing an action of some kind, check for causal factors of defects. If causal factors are present, then the action is stopped before completion, and before the defect occurs.  With the causal factor(s) removed, the action can be restarted and completed without creating a defect.

Source Inspection

Everyday examples are all around us:

Automobiles. A little fender bender in the parking lot is a annoying defect. All vehicles with automatic transmissions have a safety interlock that requires the driver to have their foot on the brake pedal before shifting out of park.

On-line forms. In this case, the defect is missing information on an airline reservation, a catalog purchase, or whatever. Possible causal factors: error of omission or typographical error. Source inspection: if a required data entry field is empty, incomplete, or in the wrong format, the reservation or ordering process will not advance to the next screen until the field is filled correctly.

Thermocouples. A residential natural gas explosion would be a very bad defect. Possible causal factor: no pilot light. Gas water heaters, furnaces, fireplaces, and so on have thermocouples. If the pilot light is on, the thermocouple is hot and gas is allowed to flow. If the pilot light goes out, the thermocouple cools down and shuts off the gas flow.

With the explosion of sensors and connectivity in Manufacturing 4.o, there will be more and more opportunities to apply source inspection for zero defects.

Will source inspection make statistical sampling and control charts obsolete?  Absolutely not.  Use statistical process control to reduce causal factor variability to minimize interruptions and increase productivity.

Quality improvement is a big part of my day job as a Professional Engineer.  Visit my Operations Engineering page for methods and case studies.

 

Control Charts

by

To provide a rational means to categorize (common cause vs special cause) and measure variation, Walter A. Shewhart invented control charts circa 1925 – 1931. W. Edwards Deming took these basic statistical concepts and tools with him to Japan after World War II and planted the seeds for the quality revolution. The rest is history.

Control charts are best maintained at the job station by the production team to provide those closest to the process guidance on when and when not to take action. Once a process is in statistical control (a very fine achievement in itself), a control chart can verify actions taken to reduce variability.

The primary signal of an out-of-control process is a point outside the control limits. Given constant probability (an requirement for process control), the chance of a point outside the control limits is very, very small. Other out-of-control signals include seven points in a row above or below the mean and runs of seven up or down. If these signals are present, then a special cause has visited the process. Look for it right away, before the trail grows cold!

Control Charts

Here is an example of an X-bar & R chart for variables data. An assembly plant body shop was requiring significant overline overtime to make production and the management team was attributing the volume shortfalls to “catastrophic” breakdowns at its constraint. A control chart for hour-to-hour production counts was in control (no special cause events) but with an unacceptably high level of common cause variation. The management team was “recalibrated” to focus their effort on the many small production loss causal factors occurring every hour: speed losses, absenteeism, operator pacing, untrained operators, material deliveries, weld faults, tip changes, etc.

Control Charts

Control charts also work with attribute data when variables data is not available. Here is an example of a U-chart of incomplete operations at an inspection and repair station. This chart has been out of control for quite awhile: the mean number of repairs have shifted up. In retrospect, the team was slow in finding the special cause in this case a defective bumper shield clip.

Control Charts

Out-of-control signals can be good. Let’s say an adjustment has been made to a process running above the specification target. A run of seven points below the averages chart mean is verification that the adjustment has moved the process mean. Or maybe a material spec change has been made. A run of seven points below the range chart mean is verification that the material change has reduced common cause variation. In both cases, new control limits must be calculated to reflect the process changes.

Terminology

Variables Data = quantitative value that can be measured (572 feet, 2403 lbs, 59.6 seconds, etc)
Attributes Data = qualitative value that can be counted (pass or fail, black or white, etc)

Common Cause vs Special Cause

by

All manufacturing and business processes contain many sources of variation. The differences may be so small as to be difficult to detect, but they are there. Variation is at the root of all waste.  Less variation, less waste.  To reduce variation, and effectively solve problems, its sources must first be understood as common cause vs special cause.

common cause vs special cause

Common Cause

Common cause variation is a constant system of chance. The sources of common cause variation are many in number but small in size. A process characterized by common cause variation is in a state of statistical control: its output is stable and predictable. The magnitude of common cause variation can be measured using control charts. The sources of common cause variation can be identified and quantified using Design of Experiments, regression analysis, and other statistical methods. The reduction of common cause variation requires actions on the system.  Example:  the total of two die.

Special Cause

Special cause variation affects processes in disruptive and unpredictable ways. The sources of special cause variation are relatively few in number but are large in size. A process driven by special cause variation is neither stable nor predictable. Special causes can be detected using control charts through out-of-control signals. The elimination of special causes requires local action on the process. Although special causes account for less than 20% of total variation in most processes, they must be removed before common cause variation can effectively be reduced.  Two examples:  an untrained operator or parts from an unapproved supplier suddenly appearing one day and generating defects at an otherwise stable and predictable work station.

Statistical Control

Statistical control is not a natural state. All processes are under relentless attack from special causes.  Confusing common cause vs special cause variation results in one of two mistakes:

(1) The first mistake is to assume variation to be a special cause when it is in fact common cause. The mistake of over adjustment leads to more variation and time wasted looking for a reason for a defect when there is no single assignable cause.

(2) The second mistake is to assume variation to be common cause when it is in fact special cause. The mistake of under adjustment is a lost opportunity to find and eliminate a special cause. Special causes are not always present so it is best to start looking for them right away before the trail grows cold…

To provide a rational means to make the distinction between common cause vs special cause variation, Walter Shewhart invented control charts circa 1930. W. Edwards Deming took them to Japan after World War II and planted the seeds for the quality revolution.

This distinction is not often obvious. A machine alarm and subsequent shutdown could be due to common cause variation.  In the digital age, we have sensors on almost everything.  Don’t just assume that alarm or defect is due to a special cause.  Analyze data over time with a control chart.

Making the correct distinction between common cause vs special cause variation is a big part of my day job as a Professional Engineer.  Visit my Operations Engineering page for methods and case studies.