Inspection of Complex Geometries: The Human Element 10.09.2012
Brief Biography A consumer of metrology Managed all coordinate measurement for production facility making over 1 million airfoils per year United Technologies - Pratt & Whitney Development Operations supported experimental manufacturing, prototyping, and evaluation of measurement systems for global commercial and military operations. A provider of metrology Hexagon Metrology Applications, Service and Technical Support Currently the National Applications and Support Manager for Hexagon Metrology Blade Specialist and a pilot And most importantly, I like to work backwards -??? 2
Let us discuss Working backwards, let s start with the end consumer. Who is the real consumer and what are their needs? How is innovation in metrology changing their experience? What exactly are these innovations? How do we cope with these changes? What is Actionable Information? Blades as a specific example of complex geometry and the Human Element How can we better use the information we get today? 3
Quality of Life: Global megatrends Water Management Mosquito Vector Control 4
Quality of Life: Aerospace Travel powered by Aerospace, Automotive, and Powertrain. Cost reductions even as fuel prices rise Comfort: Innovations used for consumer comfort or cost savings. Courtesy: FAA 5
Quality of Life: Medicine Proton Radiosurgery 6
Quality of Life: Medicine Credit: Nucleus Medical Art/Getty Images In reality, these are not our direct consumers, but we affect them greatly and we should not forget that. Working backwards, our effect on them is carried out by many others, doctors, manufacturers, the direct consumers of our metrology 7
Design Needs Need as much data as possible Acquisition speed is generally not an issue Need both real-time and historical data Need process information to improve design for manufacture Need for data reduction is not as great as others since they have the ability and motivation to do their own analyses 8
Manufacturing Needs Need data that helps make process decisions Acquisition speed is critical Need mostly real time data and some short term historical data Most important consumer of process information Need for data reduction is high, must take action based on timely and understandable information. 9
Quality Needs Both a provider and consumer of data Acquisition speed is dependent on the data consumer Must provide both real-time and historical data Must provide process information in understandable format to Design, Manufacturing and Customers. Must operate with the data at all levels in order to satisfy all consumers Compliance Defects 10
How is innovation in metrology affecting our consumers? The largest change is the speed and volume of data acquisition in the form of non-contact sensors Laser Scanners White Light, Blue Light Scanners 2,2.5 and 3D Vision inspection Constantly evolving software with Point Cloud Engines now capable of processing and visualizing clouds with billions of points Information Overload!!! We need Actionable Information. 11
What is Actionable Information? It is a core value of Hexagon. It is not a 10GB point cloud, nor is it a single CPK value. It is not simply a GO/NOGO decision. Computers can make compliance decisions easily with supervision. It is the precise amount of information that drives the decision to correct or improve a process, part or design. What is the right amount? How do we get to that amount? 12
How do we extract individual decisions from billions of measurements? Reduction: Break the problem into understandable pieces Data reduction Filters that reduce the data volume without losing definition Dimension reduction (3D>>2D>>1D) Region of interest reduction What do we care about? Feature Extraction Causation Determine cause and effect in our processes Isolate these causes and measure their effect. Problem/Symptom/Effect: The part is moving during machining, and a specific hole pattern is using too much tolerance. Cause: Work piece holder is not holding the part steady during machining of OP 20 Decision: Redesign the work piece holder 13
Profile is proposed as a solution Profile is a type of data reduction, seeking to simplify large sets of measurements. It is simple to understand It is easy to implement, It is incredibly useful for diagnostics It is great for compliance decisions It is common in model based definition But it has limited usefulness for manufacturing especially in regards to process control. Why? Profile is not well suited for use in process control because it is: Highly susceptible to noise (even using filters which have their own downfalls). Does not often have a single cause which makes troubleshooting difficult. Also often uses best-fits just like feature best-fits such as least squares, to represent lots of data with simplistic models. 14
Profile and Statistical Process Control Calculating a Cpk value using a single measured value for profile will not produce a result that represents the true Cpk. A single measured value for profile does not reflect part to part variation, it simply represents the amount of tolerance that was used. Using a single numerical value approach (representing a single region or feature(s) of interest) is more effective in determining part to part variation. 15
Data Reduction: Filters The decision for each of these three pictures is the same. Reject. However the data volume is over 100 times larger in the raw format. We use specialized hardware and software filters (there are many types, it s a topic of it s own). This is just one step in many to reduce the data to an understandable form. Voxel Grid (6.3 GB) Point Cloud (413 MB) Mesh From Points (46 MB) 16
Data Reduction: Region of Interest and Feature Extraction Surfaces toleranced with Profile require surface data Features of Size toleranced with Position typically require the resolved geometric features to determine size and location. With Position manufacturing needs measured values from the resolved geometry (axis interpretation) for both size and location to monitor/adjust processes Actionable Information. 17
Cause and Effect: Stent Measurement We might want to know what the measurements are on a cross section on a certain height. Why? How the thicknesses of the struts vary as you move axially down the stent matters. Profile of the strut all at once does not. Combining the evaluation of profile of adjacent sections would be incorrect 18
Blade Measurement: 3D-2D-1D Leading Edge Thickness Cpk = 1.6 19
3D Blade: Example This data is useful for computational models (design) and compliance (quality). But how would you use this as is for process control? Break the data down. 20
2D Blade Reduction Why are sections so important? For every reason we ve discussed Data reduction Dimension Reduction Region of interest Causation 21
1D Blade Reduction How would profile answer the thickness question for a blade? Benefit: Statistics can be calculated from these parameters. Benefit: Trends for regions, parts, and processes can be discovered For instance, what s happening from root to tip on this blade? How would you get the same information from a large 3D data set? 22
Now decisions can be made! 23
Metrology Insights: How can we better use this information? Why take all this data if it s just going to be reduced? We can still extract all the cause and effect relationships we do today. It s FAST! It is easier to reduce data than to try and get it back (especially after the part has shipped!) Virtual reality: Map the entire part in 3D, and archive the data. Change an analysis, rerun compliance tests against the virtual parts Add a new analysis, see if a hypothesis about part failure in the field matches the reality. 24
Conclusion The goal is actionable information In order to provide actionable information, we must understand decision making. Data requirements are very different based on your goal: Design, Compliance, Process Control. Timeliness, relating cause and effect, and providing understandable results all drive data output requirements. Decisions require preprocessing of data into cause and effect relationships for reasons of efficiency and timeliness The technology exists to gather and analyze large amounts of data. We must learn to manage and reduce data sets to actionable data that can be used for effective decision making. 25
Thanks for your attention! Questions? 26