Recently, a company hired me to audit several plants to determine how well its investment in statistical process control was managed. Sam, a plant manager, was expecting me the first day. "We're not using SPC," he announced, leaning back in his chair.
"Why not?" I asked.
"After the training, we tried to get it started. Our production numbers went south because of the time spent in training and meetings. My boss jumped on me with both feet, and since then, SPC has been on hold."
"Is that what you want me to put in my report?"
"Look, the truth is, I don't think SPC has any value," he stated matter-of-factly.
"Sam, what would it take to prove otherwise?"
He stared through the window, then smiled. "You see that guy there? That's Ken. He's been through the SPC training -- all the workers have. You convince him that SPC is worth something, and I'll give it another try."
"SPC won't work without manag-ement's support," I reminded him. "Can I say I'm acting on your behalf?"
Sam nodded. "Sure, anything within reason."
Ken was a bear of a man in his late 20s, with long hair and a bushy, jet-black beard. I asked him if he'd mind using SPC on his process, and I offered to help set it up.
His eyes narrowed. "When my boss comes around, he doesn't ask me how much paper I've pushed. He just wants to know how many parts I've made."
Ken's response left me in no doubt about his attitude toward SPC. I pondered my situation for a moment, then noticed the gaging setup to measure the parts. "What are these parts' most important characteristic?" I asked.
"Would it be OK if I checked them and did the paperwork myself?"
Ken grunted his consent and returned to work.
The lapping operation involved two machines, each with four circular heads that rotated independently as well as around a common center axis. Small rec-tangular parts were rubbed between the rotating heads and a stationary metal bed while bathing in an abrasive slurry. The machines polished the parts' two main surfaces to a fine finish and precise thickness.
I decided that samples from each head represented a rational subgroup and periodically charted five measurements from each head. Ken produced about four or five batches per hour. It wasn't difficult for me to keep ahead of him by plotting the subgroup averages and ranges using the same scale for all charts. Ken made many adjustments to the machines, and I noted them.
By day's end I'd plotted data from 25 subgroups. Ken stopped the machine and showed his first glimmer of interest.
"So, what do your charts show?" he demanded.
I arranged the eight charts so they were all visible. They showed much variability between batches, heads and machines.
"Look here." Ken pointed to one of the machines' charts. "I kept adding more grit to keep the thickness consistent. But I must've added too much because the parts kept getting thinner." He had a point. A pronounced downward trend was apparent, but only on one of the machines.
The next morning when I arrived, Ken was already there.
"I knew it!" he exclaimed, kneeling beside a large bag of grit. "I used new grit only on this machine. This grit is smoother and probably doesn't lap as much off. That's why the part thickness didn't change on the other machine." He frowned. "That means I wasted my time making all those adjustments yesterday. I could've left both machines alone and gotten better parts."
Ken decided to experiment with adding the two grits to the slurry. I agreed to chart his efforts and soon got a strange result: All five readings from one of the eight heads were much higher than any previous result.
Annoyed, Ken closely examined the parts and picked one up. "See how shiny this part is compared to the others?" he asked. "That's because it's thicker." He confirmed this with a micrometer and shook his head. "When this happens, my numbers go straight in the toilet."
"Why does it happen?" I inquired.
"Who knows? It just does."
Ken agreed to plot the charts while I talked to the parts people. I showed Bonnie, the parts operator, the samples Ken gave me. Bonnie measured them. "They all meet the spec," she observed.
"Yes, but they cause trouble for Ken when they vary this much." I explained the situation, and Bonnie was surprised to hear it.
"The print allows us ±0.003", but we can do a lot better," she declared.
I set up a chart and plotted a few points to show her how. "Hey, this is pretty neat!" she exclaimed. "Is this SPC?" I nodded and agreed to return at shift change and explain the chart to the night-shift operator, invoking Sam's name when the supervisor protested about the added work.
When I returned to Ken's work station, I found Melissa, an inspector, checking the parts. Surprised to find Ken plotting charts, she studied them for several minutes, then pointed to the outlier we'd plotted earlier in the day.
"What happened here?" she asked. Ken explained the problem. Melissa nodded and walked away. By lunch, Ken's grit experiments had reduced the variation by about 25 percent. He said he could do even better if he could use just one type of grit.
After lunch, I talked to a grit buyer in purchasing, showing him Ken's charts for the different grits. Normally, the buyer takes whatever grit the distributor delivers. However, the distributor, when called, said it would send whatever grit the buyer requested and also would exchange any from the existing inventory.
Back to Ken, who agreed to sort out the different brands for the exchange. He discovered there was enough of his favorite grit to last until the exchange happened.
By the end of day two, the variation was about half what it was the previous day, and Ken had centered both machines to near the blueprint nominal. His productivity for the day was close to a record, despite the time spent sorting grit in the warehouse.
Day three dawned with Bonnie and Melissa both waiting for Ken to arrive. Melissa had a batch of parts that were within 0.001" in thickness. She asked Ken to run them and let her know how they performed -- and insisted he present her with before-and-after control charts. Ken promised to do so. Bonnie showed him a quality control technique to sort the less consistent parts into groups of similar thickness.
The results were truly astonishing. By noon, Ken had accurately determined the cycle time, slurry flow and grit required to lap the parts to a precise surface finish and thickness. With few adjustments to make, he then had time to plot the control charts himself. He offered to do it before I could ask. Ken's production for the third day set a plant record.
I briefed Sam before leaving. He agreed that SPC was worth a second try.
About the author
Thomas Pyzdek is president and CEO of Pyzdek Management Inc. He has written hundreds of articles and papers on quality topics and has authored 13 books, including The Complete Guide to the CQM. Comments can be e-mailed to him at Tom Pyzdek.