Bottleneck rod machining and assembly

Automotive OEM machining and Assembly of V8 connecting rods

Situation: A rod machining transfer line of age 35 years had a needed daily throughput of 12000 units total on three shift needed to assemble 1500 engines on day shift. From historical data the line had two well known chronic bottlenecks:

The assembly machine which bolted the cap to the body suffered high levels of downtime.
A Blanchard style grinding machine used to grind the rod pin and crank end thicknesses to a consistent finish had a high reject rate. This machine had a long term rejection rate of 300 to 600 pieces per shift. Assuming half are undersized this amounted to scrap of 150 to 300 pieces per shift. The annual dollar value assuming a piece price of $10 amounted to over two million per annum. This problem was reported to have existed for decades, "over 25 years".
Both machines were manually fed and were both perceived to be joint line bottlenecks.

Rod Assembly Machine:

Situation: After observing the machine for a number of shifts and querying the employees it was concluded that a new electronic timing drum which controlled the transfer in and out motions of the machine was faulting.

Analysis: An analysis of the problem found that there were two proximity switches located at each end of the in and out motion of the transfer which were not being made consistently causing the machine to fault. The program was changed to remove one switch from the program.

Outcome: The machine started to run much more consistently. With this success, the maintenance group were motivated to address other reliability issues with the machine. In short order the bottleneck problem had been eased to the point where the line was meeting the throughput requirements. At this point ergonomics and an employees ability to load 4000 rods per shift became the new limiting factor.

Thickness Grinding Process:

Situation: The blanchard style grinder (vintage 40 years old) had three grinding wheels. An in process caliper on the grinder was used to measure the rod thickness in real time and lower the grinding wheels. The table contained two sets of fixtures arranged in concentric circles of 20 cavities per circle (pass). An employee loaded two rods at a time onto the outside pass of the grinder as the table rotated continuously. At the same time, the rods on the existing outside pass were picked flipped and placed in fixtures on the inside circle or pass. With each pass the grinder ground one side of the rod. The caliper on the grinder was used to control the thickness of the crank end of the rod to a specification of plus minus 0.005 inches. The pin end of the rod was ground to a nominal thickness of 1". Rods exited the grinder automatically into an automatic gauge which measured the crank and pin thicknesses. The gauge rejected an average of 300 to 600 rods per shift for over and under thickness conditions. The pin end had a much looser tolerance with the result that there were zero defects for this characteristic.

Analysis 1: An initial problem solving group was assigned to assess the machine and its process. After a number of months, the group concluded that the problem was excessive operator adjustment. Given that rods were being made over and under specification, operators were making frequent up and down adjustments of the caliper attempting to achieve all rods in specification. A padlock was put on the caliper. The problem was declared "fixed" to management and the group disbanded.

There were a number of concerns with this:

All operators were given a key to the lock, which neutered the intention of preventing excessive adjustment.
While rejects were slightly better the reject rate continued to be 100 to 300 units per shift. This was still a "nosebleed" level of waste.

Conclusion: The problem was not really fixed. The excessive adjustment was just a symptom of the real problem. The process was behaving like a sinusoidal wave form with the thickness rising and falling with each up and down operator adjustment.

Analysis 2: A second more intensive investigation was conducted with the employees' involvement.

They believed machine speed was a factor.
The fixturing might be the root cause.

Process speed: To test this, two sets of rods were collected of batch size 200 pieces each 10 rods per fixture at 600 and 800 pieces per hour. Thickness data was collected by 2nd pass grinding cavity Id and tabulated into a spreadsheet. A correlation coefficient of 0.94 was calculated. Since the two sets of data were highly correlated one would conclude that there was no difference in the process which made each pass. Therefore the speed of the machine was not a factor. However it was concluded that since the two sets were highly correlated that this was a strong signal that fixturing might be the issue. The unexpected benefit of this is that employees could now make there unofficial work quota 33% quicker.

Fixturing: To test this a third set of 200 rods were collected from the grinder running at the higher speed. The rods were tracked by inner pass fixture number and measured in a manual air gauge.

Analysis: The crank end thickness data was keyed into a spreadsheet and plotted on a scatter diagram with the grinder second pass cavity position plotted from one to 20 on the x axis. Plus and minus horizontal specification lines were added to the chart. Visually the chart clearly represented the root cause of the problem. Cavity five was below the low specification by 0.001" and cavity 19 was oversize by 0.001". The remaining 18 fixtures tracked in the middle 50% of the specification range.

Subsequent to this finding the grinder was shutdown on a production day. A toolmaker adjusted the two problem fixtures by grinding one and shimming the other in less than eight hours. A check of product, 10 pieces from each fixture confirmed that they were now targeted to the middle of the specification.

Outcome: The blanchard grinder proceeded to run with less than 5 rejects per shift. These rejects were "gauge errors". Scrap and rework were reduced to zero. Employee morale on the line and in maintenance improved markedly. Uptime was reduced because adjustments were all but eliminated. This process was no longer a bottleneck. This solution addressed a problem which had beset the shop for over "25 years" and was estimated to save approximately two million per year.

Note:

All numbers in these examples are order of magnitude to demonstrate each case.
The timeline to address each problem simultaneously was approximately 6 months.
Solution costs were trivial and not material.