Competition for Sigma Labs is NOT Likely
The surrounding environment and machine behavior is ever existing and changing force. Arcam describes this as "quantifiable" and "non-quantifiable" conditions in their quality assurance program, LogStudio, page 189
http://utwired.engr.utexas.edu/lff/symposium/...ngblad.pdf
Quote:
“Special cause variations are such that are not previously observed and non-quantifiable. It can for instance be due to malfunction or degradation of a system functionality , changed skill levels when changing system operators , raw materials of lower quality or changes in peripheral conditions such as electrical power stability .”
These "non-quantifiable" conditions are very difficult to control as they are not very well understood. Sigma Lab's (SGLB) in-process monitor, PrintRite3D, offers stability by ensuring that the weld pool completely changes in phase and becomes homogenous in property. This is important in limiting porosity:
http://www.industrial-lasers.com/articles/pri...ation.html
Different materials have different properties. To achieve homogenicity in weld pool, the parameters must be known for an optimal condition to be targeted for that particular metal. By examining the weld pool, the need to constantly examine the environment is not needed as the state of the weld pool is the result of the constant effect of the environment or process conditions .
And PrintRite3D works automatically to achieve the target value:
http://appft.uspto.gov/netacgi/nph-Parser?Sec...D+vivek%29
Quote:
When the weld pool volume varies from its desired setpoint , this can be detected by the thermal sensors, the weld pool frequency sensors, or combinations thereof. The thermal sensors can detect the volume change because as the melt isotherm location moves, all other isotherms move accordingly as viewed from a quasi-static reference frame attached to the heat source. The frequency sensor can directly detect the variation in weld pool volume on account of a frequency shift. [0055] i. Once a shift is detected, the real-time controller 14 can have an internal logic table that analyzes the nature, sign and magnitude of the change. Based on pre-determined rules that can be determined by a distillation of multiple modelling runs or an expert system or otherwise generated based on experimental data, the controller 14 can decide which machine variable will be the best choice for adjustment so that the weld pool volume will return to its target value . These so-called "response vectors" can define a desired control strategy for a given change in weld pool characteristics. [0056] j. The controller 14 can then make changes to the process variables welding machine tool 15. [0057] k. New measurements can be made with the sensing system and steps above can be used to recalculate the weld pool volume, and the process continues indefinitely as long as the welding process 10 is in progress .
These "predetermined rules" to define target values have been studied at Los Alamos National Laboratory (LANL) with the use of acoustic energy and published for the scientific community.
The features chosen for acoustic monitoring were shown to be reliable indicators of bond quality in this study, which was submitted to the National Institute of Standards and Technology:
http://www.boulder.nist.gov/div853/Events%20-...-paper.pdf
Quote:
Perfect classification accuracy was found to occur only during the acoustic energy power spectrum feature descriptor (test 20). Upon further investigation, the authors found a unique characteristic ringing for both an acceptable and conditional part . This ringing occurs during the second burst of acoustic energy (i.e., when speed reaches 0 rpm). The existence of this phenomenon is readily discernible in the power spectrum. Consequently, bond integrity classification of acceptable versus unacceptable and conditional versus unacceptable bonds becomes trivial – which is the ultimate goal of a feature descriptor.
Acoustic Energy studies by Mark Cola & Vivek Dave
http://www.google.com/url?sa=t&rct=j&...mp;cad=rja
PURPOSE:
"The goal of this work was to investigate the feasibility of acoustic monitoring for inertia friction welding of three austenitic stainless steels. Process results and a discussion of microstructural evolution for the same sets of welds are outlined in a companion presentation."
RESULT:
The three alloys exhibit distinct acoustic signatures versus time that varied slightly with changes in welding parameters. The bursts appear to correlate temporally with features of the torque curve. The differences in acoustic signatures for welds made on the different alloys using identical welding conditions likely stem from differences in the acoustic, thermo-physical and thermo-mechanical properties between the alloys. This technique may prove useful for determining the phase balance of the weld region after welding and predicting weld quality in-situ.
REFERENCES:
2. D.A. Hartman, M.J. Cola, V.R. Dave, N.G. Dozhier and R.W. Carpenter: Nondestructive, in-process inspection of inertia friction welding: an investigation into a new sensing technique . 6th International Conference: Trends in Welding Research; Pine Mountain, GA; USA; 15-19 Apr. 2002. pp. 948-954. 2003.
Here is REFERENCE 2, an earlier study in using acoustic energy for sensing:
http://www.google.com/url?sa=t&rct=j&...mp;cad=rja
The invaluable quality of the scientists of Sigma Labs is the depth of their knowledge in metal characteristics and behavior. The lack of material knowledge is a very big problem with AM. Prabhjot Singh, Manager of GE Global Research, stated:
http://m.technologyreview.com/featuredstory/4...-by-layer/
Quote:
So far, manufacturers don't have enough data to predict exactly how a part will turn out and how it will hold up, or how production variables—including temperature, choice of material, part shape, and cooling time—affect the results. That can be frustrating, says Singh: "3-D printing often ends up being a black art. A part is made out of thousands of layers, and each layer is a potential failure mode. We still don't understand why a part comes out slightly differently on one machine than it does on another, or even on the same machine on a different day."
It is unpredictability from the lack of material knowledge that is the problem. Vander Wel of Boeing :
Quote:
"We know how to make the metals strong enough," says Boeing's Vander Wel. "But we worry about the unpredictability . Can we repeat a result to get 100 parts that are exactly the same? We're not sure yet." "
3D printer makers have yet to understand material properties. Joris Peels, formerly of Shapeways and Materialise, founder of Voxelfab:
http://voxelfab.com/blog/2013/12/richemont-an...tch-cases/
Quote:
“There is no closed loop control on the machines and no automated processes for checking part accuracy after production. Often parts have to be hardened, smoothed and post processed sometimes with as many as five different steps on five or more machines. This entire process will have to be automated in order for 3D printing to really work as a metal manufacturing technology.”
Ask Dr. Singh what he wishes to have and he responds:
http://3dprintingindustry.com/2013/07/30/a-pe...ris-peels/
Quote:
“Singh wishes to see the following advancements in GE’s AM Lab: closed loop control, published databases of materials and material characteristics , systems to inspect parts, significant improvements to post processing, an expansion and improvement to the design tools used for 3D printing and cobalt chrome, iconel and other materials optimized for aerospace applications.”
So, no, printer makers do not have an understanding of material properties. If they did, a closed-loop control would have already been developed and the problems with inefficiency would not exist.
The invaluable quality of Sigma Lab's scientist is their reliability as consultants. Not only have they created technology to solve printer inefficiency but they serve as experts in the area of material knowledge. It takes material knowledge to create IPQA parameters to classify welds as acceptable, conditional, and unacceptable. The large amount of data Greg Morris, GE Additive Business Strategist, talks about in the concern for data mining for the AM process, the scientists are able to reduce the data a million fold and still maintain physiologic data to allow classification of welds in-process. http://www.youtube.com/watch?v=W_Rw63GIxnM
Quote:
But what do you do with these gigabytes of data, how do you leverage and use that so that when you produce a component anywhere on that platform you know exactly what you have coming out the other end? (And by the way I would say there are a number of service providers and people who have been involved in the last 10 years that are actively working with universities, or just doing the work themselves. So I think we’re going to come to a point soon where we have some very interesting solutions.) GE for instance recently signed a JTDA with Sigma Labs.
This is why IPQA cannot function by itself, the range needs to be determined. Setting the wrong parameters and ALL the parts produced will be incorrect, every single one. The knowledge of material properties is key, just as Prabhjot Singh expressed.
IPQA cannot define quality by itself
No manufacturer can simply develop an IPQA system because the system needs programming of parameters. IPQA cannot in itself define quality. This is the limitation of IPQA:
http://www.b6sigma.com/uploads/media/aerospac...qa_by6.pdf
Quote:
IPQA does have some imitations however, and the main limitation is that it cannot define quality a priori . IPQA requires expert input from the manufacturing engineer or the quality engineer to ensure that the normal baseline signature established does indeed correspond to process conditions that are capable of making a good part, as independently verified by off-line, post-process inspection.
This is where the knowledge of Sigma Lab's scientists is important and invaluable. Without their knowledge of material properties, the amount of data is overwhelming:
http://www.b6sigma.com/uploads/media/aerospac...qa_by6.pdf
Quote:
• Feature extraction.
Raw data by itself is ill-defined, large volume, and nearly impossible to interpret . As a result, the feature extraction step of IPQA is perhaps the most important in that it achieves a million-fold reduction in data density while preserving key process physics that allow discrimination between nominal, off-nominal, and anomalous events.
This is why a manufacturer cannot simply build an IPQA system. The system engineer needs to know what data to extract and still maintain physiologic parameters to produce a good part. Without proper understanding, the parameters would be incorrect and all the parts produced will be of sub-standard quality. This is the advantage and invaluable quality of Sigma Lab's scientists who are able to understand the exhaustive information from in-process data and implement control for quality assurance.
