Interval Lower-Bound QuantMiner for Identifying Equipment Group–Time Interval Conditions Inducing Quality Degradation in Multi-Stage Manufacturing

Purpose: In multi-stage manufacturing, quality rarely stems from a single operation. It arises from the interaction of equipment routing and continuous conditions such as residence time. Conventional association rule mining is interpretable but confined to categorical conditions. Its reliance on discretization and defect count consequents obscures joint equipment and time effects and misses early quality degradation.Methods: We propose the Interval Lower-bound QuantMiner (ILQ-Miner). Each antecedent combines a categorical equipment group template with a continuous residence time interval [L, U]. The consequent is a lower- bound condition on deviation from the target quality level (res ≥θ). A genetic search first restricts the space using frequent templates. It then jointly optimizes the interval and threshold within each template through a bounded fitness function, and prunes redundant rules by an anchor-based strategy.Results: On a semiconductor dataset (Commonality4000), adding residence time conditions improved rule discrimination. Across identical equipment combinations, lift rose from about 1.01–1.17 to 1.17–1.81, and ARI from 0.01–0.14 to 0.15–0.45. Rules with negative ARI fell from 19.49% to 3.21%.Conclusion: ILQ-Miner captures joint equipment and time causes that categorical mining cannot express. The rules offer actionable evidence and support root cause analysis and quality control.