Tuesday, March 12, 2013

Process Traceability using big data

Process Traceability can be understood as being able to trace every process that happened to a part as it got manufactured. Part Traceability primarily deals with what part was manufactured, and when. Process Traceability, on the other hand, builds on this information, and expands it greatly to understand exactly how the part was manufactured. Process Traceability is a key requirement in Aerospace manufacturing since manufacturing defects can have severe impacts on the quality of parts, and with capable Process Traceability systems we can go back in time and find out exactly how and when the manufacturing defect was introduced into the part. Process Traceability is a key requirement in Aerospace manufacturing since manufacturing defects can have severe impacts on the quality of parts, and with capable Process Traceability systems we can go back in time and find out exactly how and when the manufacturing defect was introduced into the part. Part traceability is done with ERP or MES,the data can reveal when a particular batch of parts was manufatured and what was the heat number (or lot number) of the castings/forgings that went into it. But this is not enough to fully understand what happened when a part was made. To build a full-fledged Process Traceability system, lets see what kind of data we can collect from the shopfloor:
 Identity data is the most basic kind required for part traceability. This data tells us what is being made, how much was made, and when was it made. Examples: Heat ID, Batch ID, Operator ID.
Operational data can tell us what the machine was doing when it had a part associated with it. We can understand the utilization of a device when it was operating on a part, how long the device was in “auto” mode versus “manual” mode, and the different downtimes that device experienced when it was working on the part.Examples: Device uptime, downtime, modes, states.
Process data can help us get into a lot more detail, and can reveal how specific features were generated on the part. For example, in high speed machining of aerospace alloys, its very important to preserve a specific chip velocity when features are being created. With process data, we can precisely know when these deviations happened, and can use it in understanding its impact. Examples: Positions, velocities, acceleration, flow rates.
Environmental data can tell us the impact of the part as its being manufactured. With detailed knowledge of the resource flows associated with the part, we can estimate its environmental impact.Examples: Resource usage, energy consumption, effluents and emissions
The bigger challenge, is to be able to intelligently and efficiently operate on this massive set of data, and find pertinent information associated with a part. Three steps are followed to find the process details using big data:
Part Search: Here we are trying to find all the information associated with a specific part (or family of parts). We start with some identifying characteristic for the part (or the family of parts). This can be a Part ID, a Heat ID, or even the day the part was manufactured. We can also identify parts based on other events, like the first part manufactured after a power outage. Information associated with a part can include all of the five kinds discussed above.
Similarity Search: Here we are trying to find similar parts based on one or more specific parts that have been identified. The idea here is that we have flagged a certain part or set of parts, and we want to scan our historical system and find other parts that share a similar process history. This is very useful when parts are being quarantined after a quality spill and we are trying to find all the other parts that need to be quarantined. If we know the ID of one part, then we can find other parts similar to it based on a variety of criteria, including: heat code, operator who made the defective part, machine condition, and alarm sequence during manufacture.
“Black Swan” Search: This is perhaps the most interesting application of a Process Traceability system, and it looks at identifying “black swans”, or the rare events that are anomalous to the norm. These queries will attempt to find parts that have been manufactured differently from the rest (starkly or subtly). These could reveal potential problems in the production process before it is identified by the customer. Examples include: excessive spindle loads, erratic feedrate override, and anomalous energy consumption during machining.
Reference: http://www.manufacturingbigdata.com/

No comments:

Post a Comment