What do you do when your highly accurate sensor is in a suboptimum location?
It’s nice to be able to place your calibrated sensor right up and close to the physical variable being measured, such as temperature, pressure, speed, or force. In some cases, that’s easy to do; in other cases, sensing at a distance is not a problem and can be fairly accurate (think IR temperature sensing).
But what do you do when you have “interference” between the sensing system and what is being sensed? That was the problem facing the Spanish instrumentation team Centro de Astrobiología (CAB), a NASA-affiliated joint center of Consejo Superior de Investigaciones Cientificas and the Instituto Nacional de Tecnica Aeroespacíal (CSIC-INTA). Their project was one of the scientific instrumentation packages for the extremely successful Mars Curiosity rover mission (Figure 1) that landed on the planet’s surface in August 2012 (Ref. 1). The team was responsible for the suite of sensors on the Rover Environmental Sensing Station (REMS), a boom-mounted assembly designed to acquire data on the Martian weather, including wind speed and direction, air and ground temperature, atmospheric humidity, and pressure, among other factors.
Figure 1: The Mars Curiosity Rover was “dropped” on the planet’s surface by an innovative and risky hovering platform that lowered it from a pre-set height. (Source: NASA/JPL-Caltech)
The problem occurred because of various constraints in the Curiosity design; the REMS sensing boom couldn’t extend out very far from the Rover. Thus, its sensors would not be out of the Rover’s fluid volume and “influence” zone. The wind stream would be perturbed by the Rover, which would likely distort wind-flow and temperature measurements.
Therefore, the team had to determine, in advance, the effects of the Rover on the wind stream flowing past it. Of course, in a mission such as this, it’s not possible to go back and adjust the location or other factors once the unit is tested and qualified. The correction factors, if any, need to be determined as part of the pre-launch analysis and test.
For the REMS team, the most obvious solution was to use a wind tunnel, but time constraints and limited access to facilities didn’t allow for that approach. That’s not surprising, being that facilities such as wind tunnels, optical and radio telescopes, and supercolliders are scheduled to the minute.
The team went to a different approach, which once might have been called “the next best thing” but is no longer a second-tier solution. They used computational fluid dynamics
(CFD) and did extensive simulations based on detailed models (Ref. 2). As a result, they found the interference based on the Rover’s shape, position, and thermal condition for the various orientations of their boom with respect to wind-stream readings (Figure 2).
Figure 2: The REMS package was charged with taking a full suite of weather-related data from the Mars surface, but faced some serious and potentially accuracy-affecting positioning considerations. (Source: NASA/JPL-Caltech/INTA)
CFD is a powerful tool and getting even more so as number-crunching power increases. Instead of trying to build an accurate “black box” macromodel of the situation, CFD uses a highly detailed micromodel with tiny cells forming a mesh. It then calculates the state and dynamics of each cell along with the effect on neighboring cells. It’s computationally intensive, but that’s much less of an issue with our present computational capabilities, unlike the situation of even a decade or two ago.
For the REMS project, the team included both thermal conduction in solids and free-forced convection in their model, which had a total of 3.2 million cells — 2.3 million fluid cells, 590,000 solid cells, and 300,000 partial cells. Computation time was between 10 and 15 hours with their 24-CPU arrangement.
While that seems like a lot of “waiting” time and effort, compare it to the time and headache of hooking up sensors to a full-sized or even a scale model. Furthermore, unless the sensors on a model have small size and thermal mass, they affect the accuracy of the wind-tunnel assessment, which is a large-scale extrapolation on Heisenberg’s uncertainty principle
, which many test setups must consider. Of course, the CFD approach is also much easier for evaluating the many “what if?” options that may need to be explored as well.
This is just one example of where the sensor’s positioning was not ideal, but it’s a situation that occurs often in test and data acquisition arrangements. Even if your sensor is calibrated and sufficiently accurate, it is affected by its placement, the impact it has on the system under test, and other factors. The challenge is that using a “perfect” sensor may lull you into thinking that the acquired data is also accurate when that may not be the case at all.
Have you ever had a situation in which the sensor was more than good enough, but the data was not, or one where you only realized that discrepancy late in the project (perhaps even after some “conclusions” were made and even acted on)?
1. “The Right Kind of Crazy: A True Story of Teamwork, Leadership, and High-Stakes Innovation
,” by Adam Steltzner (who led the Entry, Descent, and Landing team) and William Patrick.
2. Aerospace & Defense Technology, October 2017, “Using Thermal Simulation to Model the Effects of Wind on the Mars Curiosity Rover