The urgency of eliminating the link failures required an immediate redesign on the basis of Stan's calculations. Meanwhile, work on the experiments would continue to determine the validity of the assumed shock load. Stan's memo recommended that the minimum width and depth be increased to .100 inch, thereby reducing the shock induced stress to 28,000 psi. Dan and Stan wanted to increase the width on the inside of the link as this would be the most efficient use of material. It would reduce the eccentricity of loading and increase the strength of the cross section. However, system tolerances indicated that such an addition could result in interferences and the material had to be added to the outside of the link.

Attempts to add material to the height of the link were unsuccessful as interferences would result for any increase in this dimension. A change of material had been considered and rejected because of the necessity of carburizing the pin area. The new link design is shown in Exhibit B-2. During his calculations Stan had also found that the link spring strength was marginal. Therefore, his design called for a spring of .100 inch in width to match the width of the new link and to provide an adequate safety factor. During the week in which Dan and Stan were conducting their tolerance investigation to determine where they could add material, they had the model shop rough machine a few links. Once the final dimensions were established, these links were finished machined, treated and put into life tests, less than two weeks after the investigation had begun.

Four weeks after the task force formation, Dan completed the shock load tests and found normal shock loads of 22.4 lbs and worst case shock loads of 47.7 lbs. The loads at the cycle start and at latch-up were approximately equal.

Dan and Stan were now convinced that overload had caused the link failures, but they decided to extend their life tests before starting the manufacture of new links. Two weeks later, with approximately 10 million cycles on the links (twice their design life) the new design was released for manufacture and immediate replacement of the old links.

With the link problem solved, Dan became concerned that he may have created a new problem. It was possible that the increased link stiffness would increase the pin shock loading and cause pin failures (Exhibit B-2, Detail C). Dan set up an experiment with the strain gauges mounted as close to the pin as possible and found that the new link did not create a measurable increase in force. Dan surmised that the flexibility of the system made the increased link stiffness insignificant.

About one year later, Dan began to question these conclusions as the link pins began to break off. The pins were taken to IBM's metallurgy group who analyzed the problem as a fatigue failure. Dan redid his stress calculations and concluded that the pin was over designed by a safety factor of 3 or 4 and could not be overstressed. Furthermore, if the pins were overstressed a major redesign would be required. The pins are inserted into the pickup head shoe and the cross section at that point was as thin as possible. An increase in the pin diameter would require an increase in the shoe height and therefore an increase in the guide and drum housing dimensions.

Dan tried various ways of analyzing the problem, but the pin always came out with a reasonable safety factor. Also confusing was the fact that the pins failed in the uniform section, just above the .02 in. blending radius rather than at the blending radius stress concentration (Exhibit B-2, Detail C). Metallurgy did not find any machining marks which could act as stress concentrations, but continued to identify the failure as one of fatigue. Dan refused to accept the failure as one of overloading and a professor from San Jose State was brought in as a consultant. After examining the broken pins, the professor concluded that overloading could not have been the cause of failure because the areas of initial failure were much too small. Dan smiled. Upon close examination, Metallurgy found white areas about .0001 inches thick near the break point. These flakes indicated surface decarburization which would result in soft spots that could serve as failure initiation points. Therefore the failure had been one of fatigue, but it was due to a metallurgical imperfection rather than to overloading.

Dan immediately had some pins buffed to remove the surface decarburization and put the parts into life test. When the pins did not fail, the buffing operation became part of the link manufacturing process. Dan also tightened tolerances in the pin areas to prevent binding or eccentric loading and he changed surface specifications in the pin area, requiring increased smoothness to eliminate stress risers and shot peening to introduce compression in the surface (see Exhibit B-2, Notes XVII and XX).

During the following years, Dan has come to the conclusion that a change from AISI 8620 steel to a maraging steel would have been the most elegant solution to both of his problems, but unfortunately, maraging steel had not been developed at the time of his redesign. Since both the pin and the link have now been operating successfully in the field for over four years, no further changes have been made.