If you have an intermittent or periodic problem, start counting frequency of occurrence, and then figure out what the order of magnitude is compared to your process.
To solve periodic or intermittent problems in our shops, the first step after identifying the problem is collecting data about “When” and “How often” it occurs. Then, comparing it to the orders of magnitude that occur naturally in your shop can help you narrow down the likely causes.
Relative frequency can be a big help, when you figure out that the frequency has some relationship or equivalence to some aspect of your process. For example, if the frequency is about equal to two occurrences per bar, than it becomes relevant to look at bar ends first, With two ends per bar, or the fact that you might get just two parts out of the first bar end, this tying of frequency to an order of magnitude denominator saves a lot of thrashing about to try to identify root cause.
What are some orders of magnitude that occur in your shop that you should consider for your problemsolving efforts on intermittent or periodic problems? Material Order of Magnitude
Your shop processes have orders of magnitude too. Per Machining Operation
Per Stock Up
Per Production Order
How does this work? In a prior life I had an intermittent customer complaint for a twisted square bar product. The customer was counting bad pieces cut from bars in bundles.The frequency was extremely low, it was not at one per bar or one per ten bars, nor one per twenty bars. It turned out to be approximately, slightly less than “one piece per bundle.” Knowing that the frequency was that low, we were able to eliminate most of our upstream of bundle process steps. They would have generated much higher frequencies – more on the order of multiple occurrences per bar.
Based on our frequency being an approximate order of magnitude of one per bundle, we focused our investigation on the product and process at and after the bundle stage. Which was where our problem occurred-when a single bar end was being twisted by the movement of the last strapping and clip installation as it was tightened for packaging. the balance of the bar was held securely by the prior installed starps, but the tensioning unit grabbed one corner of a bar as it secured the final band around the bars, creating a twist in the end of the bar held under the tension of the clip that locked in that last strap.
Without comparing frequency of occurrence to orders of magnitude in our process, we would probably still be trying to figure out where in our process we could twist just one 14″ segment out of 3,260 feet of bars. We’d be in denial, and eventually lose the customer. If you have an intermittent or periodic problem with your products, start counting frequency of occurrence, and then figure out what the order of magnitude is compared to your process.
The frequency of occurrence of an unexpected characteristic or non-conformance is an important indication of possible root cause.
When you tie frequency of occurrence numbers back to the process you can gain some insight into what just may be the root cause.
“Occasionally I get one of these parts with the drill going off-center.” While this sentence names the problem, it doesn’t give us much insight as to where to look- the drill I guess.
“I have about half a percent of the parts exhibiting an off-center drill feature on my six spindle,” gives us enough information to think that perhaps the first couple of inches of one bar on a stock up was crooked, resulting in the off-center drill.
Converting count to a percentage of production can help you prioritize where in your process to look for the likely cause.
Trying to solve a problem without baseline data is a fool’s errand. It is the contrast between the data of the process, and the baseline data, that makes it possible to identify that a problem exists and to analyze it for root cause.
Most problems are identified because the output departs from the expected.
Brainstorms do not solve problems, they usually just waste resources in a process of aggrandized groupthink.
I call this “the Diff” when I am working with continuous improvement teams. It is the difference between expected and actual.
You cannot have a difference without having an expected or baseline measure of the characteristic to be improved.
Four Measures that I have used in my continuous improvement work include
Frequency of Occurrence,
Location of Occurrence.
Frequency of Occurrence
The difference between expected (or under statistical control) frequency and the rate of occurrence in the current state gives insights into what may be occurring. If it is a small fraction of a percent- it is unlikely that a global change of process is needed. if the rate is in the double digit percentages, it is likely that there is a major change in the process (or needed!)
Simple ratios can also be powerful clues. Defects arriving in 20, 25 or 33 percent of the production point to areas within the greater process where there may be 5, 4, or 3 sub processes- like dies, cavities, or molds. Similarly, a rate of 12.5% on an 8 spindle screw machine tells me not to look at a single tool ( it hits all 8 spindles) but instead to look for one of the 8 spindles (12.5% of the machine’s total production) that might be out of line compared to the others. Costs
My cell phone costs spiked almost 100% in July of 2010. In August, I brought my Dad back home to a nearby assisted living facility. The cell phone cost data was a pretty clear ‘cost’ signal that something had changed compared to prior (baseline) bills- Dad need assistance. ( BTW- Dad’s doing fine!) Duration
Comparison of time to complete 100 ton orders on my mill grew by a significant figure, and follow up indicated a problem at an intermediate shear. Without baseline data, how would I have known that my production time had increased? Location of occurrence
This is another piece of data that when tallied against the baseline of “no occurrences” always leads your thinking. If it only occurs in the threaded area, but not on the original bar surface, what does that mean? Looking for deltas or “Diffs” between your baseline and current process data is a far better way to inform your Problemsolving than Brainstorming.
Interested? The Delta is the Difference