PerceptEngine Features - Part III
In the previous blog post, I discussed how the steep 3D-Printing learning curve slows the adoption of Additive Manufacturing by increasing training & talent acquisition requirements. But there's another leg to this problem that's even more subtle and is interconnected with Additive component quality.
This is going to be a longer blog than usual but bare with me...
I've written at length about how quality and traceability are primary keys to Additive adoption. Quality meaning the ability for a given printed part to meet customer requirements & traceability meaning there is engineering & manufacturing data backing up that part to verify customer requirements are met.
Traditional manufacturing is held to an extremely high standard when it comes to traceability and quality control. To be a Gold tier supplier with BAE Systems, a front runner in the traditional manufacturing world, 100% of the parts you deliver must meet all quality and traceability specifications.
The reason for that is people's lives depend on parts not failing unexpectedly, that's why in engineering, factors of safety are so critical. Factors of safety exist to ensure that a component doesn't fail even when subjected to multiple times the originally intended loading conditions. In traditional engineering design, known values for material mechanical properties, via various standardized manufacturing methods and testing regimes, are used to calculate component reaction(s) to loading. This process leads to qualifiable parts people can count on.
This process is almost entirely non-existent in Additive. 3D-Printing processes significantly suffer from unreliability, geometric inconsistency and unrepeatable mechanical property results. Even components on the same build plate, made of the same material, with the same machine parameters, can produce entirely different mechanical properties.
The level of mechanical and geometric variation 3D-Printing processes produce make traditional manufacturers and quality departments reel with distrust. Traditional manufacturing requires proved out production processes with extensive testing to ensure engineering factors of safety values are accurate. Without accurate factors of safety, customer assurance and modern manufacturing supply-chain adoption is hopeless.
If process reliability & repeatability is the foundation for mass Additive adoption, why hasn't a solution been created?
The problem with 3D-Printing processes is that 3D-Printers are basically robots tasked with carrying out hundreds of thousand, if not millions of mechanical operations, in order to produce just one part.
The problem with this is that it only takes one incorrectly executed mechanical operation to forever alter a components ability to meet customer specification(s). An incorrectly executed mechanical operation can cause mechanical failure, leaks, geometric distortion, and even build failure mid-process.
So reliability & repeatability is a lost cause for Additive Manufacturing?
Not quite, it just creates new challenges manufacturing hasn't dealt with before. One of those challenges is determining how well an Additive build is progressing during the fabrication process. Currently, highly-trained operators "Baby-Sit" 3D-Printers and verify things are going as planned based on their training and individual personal experiences. This method of quality control during the fabrication process reminds me of a famous quote:
"All is opinion." - Marcus Aurelius
Every operator has their own experiences, those experiences influence and often dictate what an operator perceives as sufficient quality during the 3D-Printing process. If you have an operator with little training, they don't know what errors to look for. If you have an operator with a lot of training, they are better at identifying errors during the fabrication process. But are operators really watching 3D-Printers non-stop as millions of mechanical operations are performed to produce a final component? How precise is the human eye? What if they missed one error? Do you really want to base process reliability and people's lives on an opinion?
Alright smart guy, what do you propose that will improve confidence in Additive process reliability?
We should take a big data approach to this problem. If there are millions of operations happening in real-time, why not try and capture the end result of each of those operations and compare them against the expected end result. At the very least we should start recording optical images of each completed layer's geometry during the Additive fabrication process and compare that collected data against what was suppose to happen geometrically for that layer.
Makes sense, but how?
To accomplish this, we devised in-situ monitoring solutions that enabled autonomous optical data collection every layer during the fabrication process and then we devised a method to compare that information to expected results for a given layer in real-time. We coined the real-time output of this analysis the SolidScore metric.
The SolidScore metric is derived from a variance image that shows the differences between the digital model for a given layer and what was actually printed that layer. This variance image for a given layer shows operators during the fabrication process where there is too much material, where there isn't enough material, as well where material has been placed correctly via the 3D-Printing process.
The SolidScore metric value is a measurement of variance significance, ranging from 0% to 100%, for a given layer. A 0% SolidScore means the build is a failure, a 100% SolidScore means the geometry is being printed correctly without any significant variance, and a SolidScore value between 0% and 100% indicates to an operator the level of variance significance from the digital model for a specified layer. This is how PerceptEngine autonomously perceives Additive build quality in real-time, eliminating the need for an operator to "Baby-Sit" a 3D-Printer.
. How is the 'SolidScore' information conveyed to an operator?
PerceptEngine stores and displays the SolidScore values for each layer as the Additive build progresses. Users are able to view, in real-time via PerceptEngine's interface, how the SolidScore values are changing layer by layer and are able to view the variance images for each layer while the Additive build proceeds.
The data generated by the 3D-Printing process is autonomously collected, stored, and analyzed by the SolidScore methodology. Using this information, users discern discrepancies between what was suppose to be fabricated and what exists geometrically as the final printed component. The autonomous data collection, storage, and analysis provides data driven answers for reliability & repeatability concerns within Additive processes. This allows users, operators, and managers to hunt down sources of quality issues, failures, efficiency concerns, etc. in a fact driven manner, rather than relying solely on opinion.
Automated geometry verification is just the start for the SolidScore metric. Our ultimate objective is to roll in every aspect of Additive component quality, that includes mechanical property analyses in real-time during the fabrication process. In the next few blogs, I'll be discussing the thresholding feature and how we format the collected data during the 3D-Printing process to provide a permanent quality & traceability report for every Additive build.
- Joseph M. Sinclair
Interested in seeing if PerceptEngine is right for you? Sign up for our free Beta Program and join the revolution today!