Meaning
Statistical data profiles deviate from the standard bell-shaped curve due to process non-linearities. In polymer injection moulding, a non-gaussian distribution of part weights or cavity pressures often indicates the presence of a systematic process drift or a non-linear material response. Identifying this behavior is essential for selecting correct control charts and setting realistic tolerance limits.
Systematic Influence
Asymmetric data distributions emerge from processes constrained by physical boundaries such as zero thickness or maximum pressure. When monitoring part mass, a non-gaussian distribution occurs if the non-return valve on the injection screw does not close consistently. This behavior produces a tail of underweight parts, showing that the process is not operating under random variation alone but is influenced by mechanical wear.
In these scenarios, the presence of a mechanical bottleneck or thermal cycling in the mould creates a skewed distribution that cannot be modeled by simple standard deviations.
Quality Penalty
Standard statistical process control methods can fail when applied to asymmetric distributions. If a moulder assumes a normal distribution for data that has a non-gaussian distribution, the calculated process capability index will be incorrect. This error leads to either false alarms or undetected defects, as the control limits do not match the real distribution of the parts.
Process Correction
Tracking the skewness and kurtosis of the collected data helps identify the root causes of the asymmetry. Once the non-gaussian nature is confirmed, the process can be corrected by replacing worn non-return rings or adjusting the screw recovery settings. This diagnostic approach restores process capability and ensures consistent quality without relying on unrealistic normal assumptions.
