Using SPC in the coreroom.

Assuming statistical control can help to eliminate variations that hinder consistent quality control.

To make improvements to any process, that process must first be stable--a recognized truth in all foundry manufacturing technologies. Unless a process is in statistical control, it is impossible to understand the process' problems and find a solution. Consequently, efforts to control the process and then further reduce any variation go hand in hand.

Focusing on the tensile strength of chromite sand cores bonded with conventional oil binders, the following article describes the task of obtaining statistical control and then reducing variation for a better metalcasting operation.

Auditing the Process

Taylor & Fenn, a 160 year old foundry, pours ductile and cast iron, carbon, low alloy, and stainless steel for specialty machines, machine tools and pumps. The reason for trying to get the coremaking process under statistical control was simple. The tensile strength of cores was subject to excessive variation. Those that were too strong wouldn't collapse, leaving hot tears in the casting, while the rigidity retarded gas venting. Others were too weak, which left the castings without the good surface finish desired by customers.

To get a handle on these problems, the sand committee set to work auditing the entire coremaking process from beginning to end. All the sections of the committee--sand management, maintenance and quality control--contributed input and raised concerns, while the sand lab undertook extensive testing.

Soon, a list of all possible variables within the coremaking process was compiled and each one was measured to gage impact on quality fluctuation. Figure 1 shows a variables control chart (x R) of the tensile strengths of cores over a seven-month period. The control chart shows an out-of-control process.

Using the charts, we were able to identify several sources of variation:

* distance between pan and plough in the muller;

* accuracy in weighing sand mixture ingredients;

* mulling time:

* temperature variation in the core oven.

Measurements were recorded of all the above variations. Temperature variation was immediately and significantly reduced to within 20F.

The Quick Fix

All assignable causes of variation were taken into account, and once recognized, they were easily monitored and manipulated. While readying the process for even greater improvements, the simple correction of these factors changed the variation--a natural benefit accrued from having a process under statistical control.

The average variation of tensile properties in a mold of six samples was reduced from 72 psi to 56 psi. Variation around the nominal was cut from |+ or -~35 to |+ or -~26 psi.

Tangible Benefits

The results have had a favorable impact on one of the foundry's main concerns--on-time delivery. More predictability of both process and outcome allows us to more easily meet any given job requirement. Obviously, the better surface finishes please customers and the foundry can deliver a better product.