Time and again, clients approach with a request to help replace a machine that has been producing rejections. Further, they add that since the rejections are extensive, it requires complete inspection of all parts. Inspite of 100 per cent inspection of parts, they still receive customer complaints; and this again takes them back to the original problem.
Before getting into selecting a better machine, the first and foremost action is to study the current process. Most often, studying the process systematically results in identifying the cause. Following this up with a root cause analysis results in eliminating the problem, and a better machine may not be required.
The run chart
A systematic study starts with making a run chart for the dimension, which causes rejections. A simple run chart, which plots the dimension sequentially as components are produced, is similar to the pulse of the process. On this run chart, one must mention the reason for stoppage of work. One must continue to make the run chart for two or three shifts or even longer, depending on the cycle time. Problems in the process surface automatically - including inconsistent input material, cutting tool problems, work holding problems, measurement problems, etc. Several variables affect a process. A run chart is a starting point in identifying the significant variables that affect a process.
The dimension grows and then stabilizes. After some time a similar behaviour repeats. One also notices that this behaviour takes place every time a machine is stopped for a while and then restarted. This is a clear indication of spindle growth during warm up, also indicated in the run chart. Sometimes, I have noticed this phenomenon in mid shift, without any reason. Further investigation showed that the operator had stopped the machine, either for removing chips or for maintenance.
Over a period of time, the dimension increases, this can happen within an hour or in a single shift. This is an indication of tool wear. Offset must be given periodically. At some point of time, the tool would also have to be changed. Run charts serve as reference for future needs, and help in determining when offsets should be given or when tools changed. As the tool wears out gradually, one also finds more variation in the readings. If one observes this behaviour and continues to look at the surface finish, one can have a better judgment of when the insert needs to be indexed.
In aluminum machining, it shows abrupt and erratic change in dimensions. Close observation of the process indicates that a built-up edge had formed. After some time, the built-up edge will break. This causes a shift in dimension or process average. After a while, one more built up edge is formed, and so on. Without doubt, correct cutting parameters is the solution applied.
In some run charts, one observes too much variation. In one case, during a boring operation, the input material was pre-bored on a conventional lathe to reduce cycle time on the CNC. A single cut was taken on the CNC lathe. The behaviour of dimensions was erratic. The operator kept measuring each component. He kept the setting on the lower side of the bore, so that the component would not be rejected and could be polished and saved if found undersize. Whenever the operator measured a job going towards one end of the tolerance, he would give an offset. In spite of all his efforts, it resulted in rejections to the extent of 3 per cent. Productivity continued to suffer as the machine was stopped at regular intervals to give offsets. During that particular shift, 140 components were produced. This is a clear case of a poor process. It was decided to keep a little more material at the time of pre-boring, and an additional semi finish cut was added. Even though it increased the CNC cycle time, 160 components were produced in the shift without any rejections and the run chart appeared. For long periods of time, tool offset was not required and production was very smooth.
Sometimes, a run chart shows dimensions being produced with different averages. This is clearly a case of distribution, which is not normal. On plotting a histogram (frequency diagram) of the readings, the conclusion is clear. This is found to be a very peculiar case. On analysis, it was found that the raw material bars came from two different sources. Even though they had the same chemical composition, there was a difference in microstructure. Without a run chart, it would have been impossible to find out what had happened.
In one interesting case, a run chart was found. There seemed to be too little variation in the process. After some time the inspection equipment was used to measure it. It was found that the instrument was sluggish and defective, and was not showing the right variation. When replaced with a correct instrument, the inherent process variation was clearly visible. Sometimes, even if the instrument is working fine, one happens to find that it is not the correct one. For instance, if one uses a micrometer with a least count of 0.01 mm to measure a component having 0.02 mm tolerance. Similar situations have to be tackled quite frequently on shop floors. In this case, the run chart would show very little variation.
Dimensional deviations often take place due to deflections. These deviations take place at several places. It can be from the machine structure and slides, from work holding, tool deflection, and workpiece deflection. Deflections can also take place due to varying cutting forces arising out of varying workpiece hardness. As cutting tools wear out, cutting forces increase, causing deviations. In a run chart, one may not be able to see the effect of each variation separately, as it becomes part of the random variation of the process. However, sometimes patterns emerge, which tell their own story. It may be necessary to deploy additional tools to study the process.
The run chart is the first step in studying a process. It is similar to a doctor studying the heartbeat of the patient with a stethoscope. Analysing a run chart makes the process engineer observe the process. More often than not, close observation itself throws up the underlying cause and its solution. Only when the run charts do not provide required information are other sophisticated tests performed.