ELDEC Corporation (Lynnwood, WA) designs and manufactures electronic sensors and systems for the demanding environment found in aerospace and defense applications. Its products are in use on all major U.S. and European aircraft.
One of ELDEC’s standard product lines is the magnetically operated proximity sensor. This device uses magnetic fields to detect the presence of conducting or magnetic materials. These sensors are typically used to detect positions of such items as aircraft doors being closed or locked, if landing gear is up or down or locked, whether thrust reversers are deployed, and for the detection of flap and slat operation.
Twelve years ago, ELDEC engineers researched the use of finite element analysis (FEA) in the process of proximity sensor design. In doing so, they investigated several commercial FEA software providers. Ultimately, they selected Maxwell software from Ansoft Corporation (Pittsburgh, PA) because the technology is highly accurate, easy to use, and capable of some interesting extended operations. Challenges in Designing Aerospace Magnetic Sensors
The aerospace environment is harsh and thus demands rugged sensors. The sensor must operate over wide temperature extremes, function normally in very high electromagnetic interference (EMI) fields, endure high vibration and shock loads, and work reliably in ice, sand, dust, oils, caustic fluids, and other elements. In addition to these considerations, the product must be built at an affordable cost to the customer, be profitable for ELDEC, be designed within a preset budgeted time and expense, and have a low scrap rate.
ELDEC, in order to meet these requirements efficiently, had to reduce its design cycle time and add design abilities that would provide more rugged products. ELDEC discovered that Maxwell provided the abilities in design that were needed, and some that were never possible before.
Part of sensor design requires optimizing the shape of the sensor, selection of materials, evaluation of temperature characteristics, and evaluation of production variances. Without Maxwell, this task was one of trial and error. Material selections were best guesses. Temperature evaluation was a matter of building and testing. Production variances were handled by which product was experiencing the highest scrap rate, while sensor shape was the old tried and true shape from the past.
One of ELDEC’s first priorities using Maxwell was to utilize the ability it had to allow an external program to vary the magnetic model design. ELDEC utilized a second computer with an optimization program to evaluate the output from Maxwell, then manipulated the design, looking for optimum performance. Initial results were beyond expectation, with desired parameters such as sensitivity increasing by as much as 100%.
The time to finish the first stage of the design was reduced, thus allowing ELDEC to take advantage of other features of the program. One of those extra features allowed ELDEC to vary conductivity and permeability of the materials which make up the sensor, thus verifying that the unit would work over temperature as well as with the variances of the received material itself. Another variance that ELDEC could now deal with and predict results was in the mechanical variance of the parts. By making models over the mechanical variances, ELDEC could predict the production expectations, and define what tolerances needed to be with a sound reason for them.
ELDEC calls these their “three-sigma” studies. In running three-sigma studies, some interesting side benefits occur. In addition to knowing what parameter dimensions must be held, they also discovered which parameter dimensions do not need to be held tightly. This phenomenon can help save money. In this process, ELDEC engineers will typically model a +3 signal design and a -3 signal design. Those two models will show the two end points of this product and how, over the lifecycle of the product, 99.7% of the designs should fall within those two points.
Ansoft is now including optimization into its 3D software package, but like ELDEC, users can also use Ansoft’s axial symmetric 2D package with a second computer and optimize designs in that environment.
Because a software program is only as good as the parameters that the user provides, ELDEC also needed to perform some material studies. That is to say, that there were many materials that they were using that had no published conductivity or permeability characteristics versus temperature. Using Maxwell, they found that they could create a mapping function that allowed them to make eddy-current readings on a material. From that data, they determine the material conductivity and permeability. Using this method, they also discovered advantageous annealing profiles to produce parts with the least variance or some other optimized parameter.
By using the process of magnetic FEA and Maxwell, “we were able to improve an ELDEC product in every area of benefit.” ELDEC states they now have an all metal sensor that is extremely reliable. Signal to Noise
As an example of improvements, signal to noise is one of the main parameters to optimize. The environment to which the product is exposed sets noise levels, such as temperature and EMI. The customer usually specifies these. Sometimes the device can be designed to reduce the effect of the noise, such as an all metal housing that will reduce the effect of EMI. Once this task is completed, the only thing left to do is increase the sensitivity to the target. The target is a piece of metal of a particular material and annealing to which the ELDEC sensor is sensitive. For instance, a specific sensor indicates when landing gear drops down for landing and folds up during take off by recognizing the presence of the target attached to a member of the landing gear. As the landing gear drops down, the target nears the sensor.
ELDEC uses Maxwell to help increase their products’ target sensitivity as high as possible. For a variable reluctance sensor, sensor target sensitivity is measured as the percent inductance change from a target at a defined actuation point to the target not present. After implementing Maxwell, along with optimization, the sensitivity ELDEC had previously achieved improved dramatically. This was a relatively shocking improvement, and a real blow to the ego of old time “seat of pants” designers. At first, they thought they did something wrong with the software or improperly conducted the study. So, they built a physical prototype and tested it only to learn that, indeed, the device reached the sensitivity Maxwell stated. This early effort brought ELDEC into a whole new signal to noise ratio arena. Benefits of Maxwell
Several years ago, ELDEC challenged its employees by encouraging them to participate in a company-wide contest to help improve product quality. Called the ELDEC Quality Improvement Program, it solicited the best and the brightest to demonstrate how new technologies and methodologies helped improve product quality. At the time of this contest, ELDEC had just started using Maxwell and tying it in with the optimization process. The sensor designers using Maxwell entered the contest and won first place.
Recently, ELDEC decided to attempt to design a sensor that could be assembled without any need for calibration. This would lead to reduced production costs, and potentially to an even higher reliability unit. By using Maxwell, new methods were discovered and the product is selling very well today as an inexpensive, high signal to noise ratio product. Maxwell and the optimization package were used for this and many other ELDEC projects. Learning Curve
ELDEC sensor designer Kevin Woolsey is a key Maxwell user. He says that Maxwell is not hard to learn. “It’s pretty straight forward. Before we selected Maxwell, we evaluated all the products on the market in the late 1980s. We selected Maxwell because of its front-end functionality. Today, I believe the product is as good or better than other products out there because of its ease in modeling. The ACIS front end, automatic mesh generation, and the ability to add, subtract, and intersect objects makes the initial modeling a snap. The parametric interface also allows you to tie in optimization. That’s really nice.” Trends for the Future
Woolsey recommends computer-aided technology for those companies that have not yet taken the plunge. “As far as cost savings and time to market, companies have no choice but to develop strategy such as what we have put in place. In terms of remaining competitive, manufacturers must perform the design iterations on the computer. They will gain a higher quality product and understand the true three-sigma variance in a shorter period of time. The bottom line is the manufacturer gains a higher profit margin while the customer gains a high quality product and a reasonable time to market. I don’t see any other way that is as cost effective or more accurate than this method.”