Peter Nachtwey
Member
Last year at about this time I was invited to go to the National Airport Pavement Test Facility to help out with their test system. Usually our controllers are used to test landing gear, blades, flaps etc but in this case the purpose was to test the runway. The big A380s are going to put a lot of stress on the runways so the FAA wants to simulate runway wear to determine the rate of wear and how to make better runways.
The previous system consisted of a Mitsubishi Q PLC that talked to many smaller Mitsubishi PLCs that controlled each of the 10 actuators. There are 5 per side and two wheel per actuator. The first system did not control very well for a number of reasons. RMC75s were chosen to replace the smaller Mitsubishi PLCs. The motion controller has a position feed back for controlling the up and down motion of the wheel and a pressure sensor on each side of the piston to calculate a differential force across the piston and the offset can be changed so a net force takes into account the weight of the wheels.
The first day or two was spent by Ryan Rutter and me training each other. I trained Ryan on the motion controller and Ryan trained me on what the machine had to do and its basic operation.
The first challenge was easy. We got the Profibus working between the PLCs and motion controllers quickly. We also got the tuning for the force to work accurately under static conditions.
Problems started to occur when we tried to control pressure under dynamic conditions.
Someone had put a 3Hz filter in the feedback path. The feedback was an analog voltage for pressure. I said that will not do because you really end up controlling the filter and not the actual force. When the 3Hz filter was removed the feedback was extremely noisy. I left and came back a few months later after all the wiring was redone to reduce the noise.
The next hurdle was controlling the force as wheels went up and down over the ruts in the runway. No matter how high we turned up the gains the wheels would not track the target force and if we turned up the gains too high the actuator would oscillate. Not good. The FFA PLC engineer was concerned but I told him that it would be 'fun' and by the time we got done with this it would be a project to talk about. I was confident.
During the static force tests I noticed that the force increased by about 18000 lbf for every 0.1 inches actuator was pushed down. This relation ship wasn't that accurate but there is not point in sweating details if the general approach doesn't work. The ratio of force to position depends on temperature, the tire pressure and how much surface area the tire has in contact with the runway. My plan was to map the elevation of the runway. If I know the elevation of the runway at each point along it I can move the tire up and down in position control so the wheels are always compressed the same amount and therefore maintain a steady force on the runway. The runway test second is up to 300 ft long. I built an array in the motion controller for 601 readings so there was one every 6 inches. The controller can do a cubic interpolation between the points to compute a position and velocity. In the article there is a formula that shows the the flow entering or leaving the cylinder must be equal to the velocity times the area of the piston if the pressure is to remain constant.
We then rolled the tires over the runway slowly enough so the force PID would control the force accurately enough. The actuator positions were recorded every 6 inches and put into the array. The array was used to make a big cam table or spline that the actuators could follow as a function of the tests machine's position along the runway. This improved the response a lot. It was kind of like a feed forward or disturbance rejection because now the controller can predict the position of the actuator to achieve a force rather than to wait for a very compliant tire to push on the actuator.
As noted above, the ratio of 18000 lbf to 0.1 inches was a good estimate but not always accurate. There was still need for the force control. The force control PID was modified. Instead of outputting a control signal directly to the hydraulic valve it generated a position offset for the position PID that was following the elevation spline. This helped to correct for errors in our assumption that the force was proportional to how far the actuator was pushed down.
The idea worked great BUT there was yet another problem. As each wheel came down to hit the runway it lifted the WHOLE machine up a little and this meant that the wheels that are already in contact had to push down a little farther to maintain the same force. This meant that our tables needed yet another offset but we didn't know how much it should be.
The answer was to get laser to measure the distance from the frame of the machine to the runway. The laser can detect how far the machine is being lifted up as the next set of wheel come down. It could also map the runway elevation much more accurately than rolling the wheels over it. This was a big improvement. Ryan Rutter, control engineer, implemented this on his own. By this time he knew how to program the motion controllers.
So, this project uses cascaded control with an inner position loop and outer force loop. It uses splines to predict how far the actuator should be extend down to generate a force and all of this is geared to the motion of the machine as it travels up and down the test runway. Each of the wheels had to have a separate offset into the spline table. The synchronization was done over Profibus.
I was at the NAPTF for two one week stretches. The first week was have training and just getting the basics set up. It was the second week where the cascaded loops and splines were implemented.
http://www.airporttech.tc.faa.gov/naptf/
http://www.controleng.com/single-ar...c-controls-for-runway-testing/e17b8b3882.html
The Nation Airport Pavement Test Machine is a one of a kind. The US tax payer funds the testing and shares the test data with the whole world so that safer runways can be made. Think of the mess if Chicago, Heath Row or the Atlanta airports had to shut down a runway for maintainance for months at a time. The runways have sensors buried 20 inches below the surface to measure stress and strain. What is interesting is that the weight that pushes down on the center of the runway actually deforms the runway. The edges of the runway are actually rise up as the middle is pushed down. This happens even though the runways at 20+ inches thick. One more thing. The actual landing is not that bad on the runway since most of the weight is still supported by the wings. The runways are stressed the most when they are warm or hot and the plane is rolling slowly.
The article got delayed because unnecessary people in the FAA got furloughed last summer for a couple of weeks. That included Ryan Rutter, the engineer that is doing most of the control work. I know we get our money's worth from his efforts. The FAA has their own version of this article for their PR purposes.
The previous system consisted of a Mitsubishi Q PLC that talked to many smaller Mitsubishi PLCs that controlled each of the 10 actuators. There are 5 per side and two wheel per actuator. The first system did not control very well for a number of reasons. RMC75s were chosen to replace the smaller Mitsubishi PLCs. The motion controller has a position feed back for controlling the up and down motion of the wheel and a pressure sensor on each side of the piston to calculate a differential force across the piston and the offset can be changed so a net force takes into account the weight of the wheels.
The first day or two was spent by Ryan Rutter and me training each other. I trained Ryan on the motion controller and Ryan trained me on what the machine had to do and its basic operation.
The first challenge was easy. We got the Profibus working between the PLCs and motion controllers quickly. We also got the tuning for the force to work accurately under static conditions.
Problems started to occur when we tried to control pressure under dynamic conditions.
Someone had put a 3Hz filter in the feedback path. The feedback was an analog voltage for pressure. I said that will not do because you really end up controlling the filter and not the actual force. When the 3Hz filter was removed the feedback was extremely noisy. I left and came back a few months later after all the wiring was redone to reduce the noise.
The next hurdle was controlling the force as wheels went up and down over the ruts in the runway. No matter how high we turned up the gains the wheels would not track the target force and if we turned up the gains too high the actuator would oscillate. Not good. The FFA PLC engineer was concerned but I told him that it would be 'fun' and by the time we got done with this it would be a project to talk about. I was confident.
During the static force tests I noticed that the force increased by about 18000 lbf for every 0.1 inches actuator was pushed down. This relation ship wasn't that accurate but there is not point in sweating details if the general approach doesn't work. The ratio of force to position depends on temperature, the tire pressure and how much surface area the tire has in contact with the runway. My plan was to map the elevation of the runway. If I know the elevation of the runway at each point along it I can move the tire up and down in position control so the wheels are always compressed the same amount and therefore maintain a steady force on the runway. The runway test second is up to 300 ft long. I built an array in the motion controller for 601 readings so there was one every 6 inches. The controller can do a cubic interpolation between the points to compute a position and velocity. In the article there is a formula that shows the the flow entering or leaving the cylinder must be equal to the velocity times the area of the piston if the pressure is to remain constant.
We then rolled the tires over the runway slowly enough so the force PID would control the force accurately enough. The actuator positions were recorded every 6 inches and put into the array. The array was used to make a big cam table or spline that the actuators could follow as a function of the tests machine's position along the runway. This improved the response a lot. It was kind of like a feed forward or disturbance rejection because now the controller can predict the position of the actuator to achieve a force rather than to wait for a very compliant tire to push on the actuator.
As noted above, the ratio of 18000 lbf to 0.1 inches was a good estimate but not always accurate. There was still need for the force control. The force control PID was modified. Instead of outputting a control signal directly to the hydraulic valve it generated a position offset for the position PID that was following the elevation spline. This helped to correct for errors in our assumption that the force was proportional to how far the actuator was pushed down.
The idea worked great BUT there was yet another problem. As each wheel came down to hit the runway it lifted the WHOLE machine up a little and this meant that the wheels that are already in contact had to push down a little farther to maintain the same force. This meant that our tables needed yet another offset but we didn't know how much it should be.
The answer was to get laser to measure the distance from the frame of the machine to the runway. The laser can detect how far the machine is being lifted up as the next set of wheel come down. It could also map the runway elevation much more accurately than rolling the wheels over it. This was a big improvement. Ryan Rutter, control engineer, implemented this on his own. By this time he knew how to program the motion controllers.
So, this project uses cascaded control with an inner position loop and outer force loop. It uses splines to predict how far the actuator should be extend down to generate a force and all of this is geared to the motion of the machine as it travels up and down the test runway. Each of the wheels had to have a separate offset into the spline table. The synchronization was done over Profibus.
I was at the NAPTF for two one week stretches. The first week was have training and just getting the basics set up. It was the second week where the cascaded loops and splines were implemented.
http://www.airporttech.tc.faa.gov/naptf/
http://www.controleng.com/single-ar...c-controls-for-runway-testing/e17b8b3882.html
The Nation Airport Pavement Test Machine is a one of a kind. The US tax payer funds the testing and shares the test data with the whole world so that safer runways can be made. Think of the mess if Chicago, Heath Row or the Atlanta airports had to shut down a runway for maintainance for months at a time. The runways have sensors buried 20 inches below the surface to measure stress and strain. What is interesting is that the weight that pushes down on the center of the runway actually deforms the runway. The edges of the runway are actually rise up as the middle is pushed down. This happens even though the runways at 20+ inches thick. One more thing. The actual landing is not that bad on the runway since most of the weight is still supported by the wings. The runways are stressed the most when they are warm or hot and the plane is rolling slowly.
The article got delayed because unnecessary people in the FAA got furloughed last summer for a couple of weeks. That included Ryan Rutter, the engineer that is doing most of the control work. I know we get our money's worth from his efforts. The FAA has their own version of this article for their PR purposes.