In the mid-1970s I worked in product design and quality control engineering for a well known battery (primary cell) manufacturer, troubleshooting a variety of problems. This was my second stint there. In my previous go 'round, I had come up with several devices related to production line troubleshooting.
Three weeks after I returned, the production superintendent stormed into the office telling me that a 30 percent rejection rate was causing chaos in the AA battery assembly line. My boss intervened, asking whether the R&D guys had any answer to the problem. Apparently not. They were just as baffled as before. This problem turned out to occur sporadically over the past two years, but now it had reached crisis levels since a majority of the AA production was allocated to supply a military contract.
The line had set aside defective cells -- about 50 of them -- and these were delivered to my desk. I sat looking at them for a while trying to figure out where to start. Looking up from the desk, I saw a sea of anxious faces staring in my office windows. I knew I had better do something.
I began testing the cells one by one on a no-load meter. This allowed me to rearranged the cells according to voltage in one of the trays that had compartments in a 10x10 matrix. As I worked my way through the testing, a pattern emerged that approximated a bell curve. This gave me an idea. Whatever was causing the problem appeared to have a normal frequency distribution. This suggested that a statistical analysis might provide a clue. To get a comparison distribution, I commandeered good cells from the line and did the same no-load voltage test. Again, the pattern of good cells showed a normal frequency distribution.
I had a Mettler balance on my desk, and on impulse, I started to weigh the bad cells, noting down the tally on a Frequency Distribution Analysis Sheet (FDAS). This also showed a normal frequency distribution. Next, I did the same for the good cells on a common FDAS and plotted the bell curves. This showed that the defective cells were significantly heavier than the good cells.
I dismantled the best of the good cells and the worst of the bad cells and trotted along to the R&D lab to have the consolidation pellets (MnO2) analyzed. Within a couple of hours the lab came back with a report saying the X identified pellets had about 3.5 percent mercuric oxide (HgO), and the Y identified pellets had 99 percent manganese dioxide (MnO2) with no traceable HgO. X came from the defective cells and Y from the best cells.
Again I took extreme differences of good and defective cells. This time, I had them X-rayed before I took them apart. This showed that the defective cell displayed a soft short at the bottom of the separator. Again, the lab did an analysis for me and found the defective cell had just over 2 percent HgO and the good cell around 98 percent MnO2. By now, the suspicion was emerging that HgO was causing the problem. Meanwhile, most of the day had passed and my anxious boss asked me if I had found anything. I showed him the results, suggesting that HgO was the culprit. He had difficulty accepting this. In the past, recovered dust from the pellet room had always contained percentages of HgO without causing any problems in the MnO2 cells. He asked me to do some more testing. The results ended up the same -- up to 2 percent HgO in the defective cells and 0 percent in the good cells.
The next day, I showed my boss the latest results and suggested we manufacture cells from virgin MnO2 material as an interim solution until we found out what was causing the soft shorts in the defective cells. My chain-smoking boss went through several cigarettes before issuing a directive to manufacture only with virgin material and to quarantine all recovered powder and consolidations containing recovered powder. The plant manager went through the roof, demanding an explanation. When he was shown the evidence his eyes told me he saw the possibility. He told my boss and me we had better be right or we'd both be looking for new jobs!
There was a five-day aging period before sample cells were ready to do a life test. My boss and I waited impatiently for the results. It was the best distribution on record. In fact, the test lab had done a duplicate; the chart traces for both nearly blended as one discharge curve. The consistency was that good.
The problem was handed over to the R&D lab. They had to track down what was happening based on the HgO percentages and the evidence of soft shorts.
Four months later, R&D came back with the finding that a replacement (and cheaper!) separator material was drawing oxygen from the HgO, resulting in microscopic spheres of mercury bridging between anode and cathode, and causing the soft shorts. That separator material was scrapped and the previous material reinstated. The quarantined powder was systematically brought back into production at much diluted quantities in the virgin MnO2 -- and the popularity of Frequency Distribution Analysis Sheets popped up in all sorts of troubleshooting from then on!
This entry was submitted by John Mitchell of Mitchell Research and edited by Rob Spiegel.
Tell us your experience in solving a knotty engineering problem. Send to Rob Spiegel for Sherlock Ohms.