Issue link: https://iconnect007.uberflip.com/i/594816
November 2015 • SMT Magazine 91 As shown by Figure 10, each different machine has its own library, which must contain defini- tions of the parts it will handle. A single assem- bly line may have five to six different machines all with their own libraries; across a large factory, the number of separately maintained libraries can run to the hundreds. Just consider the time in- volved, per new part, per machine. An industry average is to spend 15 minutes per new part, per machine type, preparing and testing out the part data. If there are 20 new parts in a PCB that is new to that assembly line, that is five machines in the line. It will take 1,500 minutes (more than two hours) of machine down-time to prepare the ma - chine libraries for that particular product. In an environment of lean manufacturing with small lot sizes, such set-up overheads can mean that the manufacturer spends more time with his assem- bly lines down than up and running. When we examined the nature of the parts data in the machine libraries we discovered that, with the exception of data that cannot be found in a component's datasheet such as color changes and markings that vary with lot code, the machine library content can all be derived from a standardized 2.5D model through the use of machine-dedicated rules. The flow for the solution, together with the requirements at each stage, is as follows. First requirement The flow requires a source library for the component models that links back to the com- mercially purchasable part-numbers used in the supply chain. This enables transferability of a STreAmLINING PCb ASSembLy AND TeST NPI WITH SHAreD ComPoNeNT LIbrArIeS arTIcLe figure 10: There are as many machine-library standards as there are types of machines in the world. figure 11: Mapping purchasable component part numbers in the boM to shared library content.