Issue link: https://iconnect007.uberflip.com/i/1539960
68 SMT007 MAGAZINE I OCTOBER 2025 are tedious, time-consuming, and prone to human error. Additionally, debugging without data led to extended system downtime on the production floor during test fine-tuning. The manufacturer adopting this AI/ML solution had multiple production sites worldwide. This solu- tion helped to quickly identify the Cpk scores of their new fixtures. When the data from the design phases is also integrated and extended further dur- ing the "improvement phase" of fixtures, the model can monitor and suggest improvements in qual- ity and yield during mass production in real time. The model used a time series forecasting algo- rithm, which tracks the real-time data and predicts the trend. Further, the model also suggests the lim- its for a target Cpk. This solution helped to reduce the NPI roll- out cycle from two months to three weeks. It also increased the first pass yield from 80% to 95% and reduced retest rates by 5%. Typical Design of AI-augmented Manufacturing Most manufacturing organizations have already implemented digital operations, including Manu- facturing Execution Systems (MES), with defined processes and systems to capture both structured and unstructured data. This is the foundational step towards becoming AI/ML-ready. The complexity of AI implementation can greatly depend on the orga- nization's data security policies. Any AI/ML imple- mentation requires significant hardware infrastruc- C o m p a r i s o n of o l d vs . n ew b u s i n e s s f l ow fo r N P I i nt ro d u ct i o n fo r m a s s p ro d u ct i o n . ▼