Silvana Capsize Report
Joshua Leighton
Thursday 10/10/2013
Setup and Configuration
Silvana was equipped with a depth sounder mounted off the starboard stern while the modem towfish was being towed from the left side of the vehicle. The bridle was not used to increase separation, and hopefully limit interface, between the towfish and depth sounder. PID gains were as follows:
Kp = 0.4
Kd = 0.6
Ki = 0.015
Integral limit = 750
Silvana was running a typical lawnmower mission at 60% thrust. These are fairly typical settings. William and Josh were running the kayaks on this day.
Events
About halfway through the second mission of the day we spotted Silvana turning in a tight clockwise circle on pMarineViewer and then suddenly lost WiFi comms. After spotting the vehicle capsized with the binoculars we set out to retrieve it. The vehicle was almost completely turtled, with the thruster angled near 90 degrees. The thruster was off, but it sounded as though a gear was slipping as the steering servo tried to turn the thruster. The e-stop did not turn off the steering servo, so the battery was disconnected and the vehicle towed back to the dock.
In total the vehicle was capsized for about 15 minutes, and the return to the dock took another 15 minutes or so. There was approximately 6” of water in the vehicle after righting it – enough to completely cover the lower rows of waterproof connectors.
Damages
The cpu, radio, and battery boxes remained completely dry and no repairs of them were required. Some water had entered the motor driver box, perhaps through the unused second battery connector. The only failure seems to be the power switch for the motor driver box, whose contacts were fused together after shorting out. The motor driver itself worked fine after drying the box and bypassing the switch.
The large “mega-servo” used for steering was removed from the kayak and its case opened. Inside, several loose screws had caused its gears to not fully mesh. After tightening everything up, the servo worked correctly and was reinstalled in the kayak.
Lastly, a 5200 sealant “plug” had come loose from a 0.25” hole in the bottom-center of the kayak, allowing water to quickly drain out when the kayak was recovered. The 5200 does not adhere well to the plastic of the kayak, so the plug was replaced with a short ¼-20 bolt that was more easily sealed.
All plugs were dried with compressed air, though most had only gotten wet after the kayak was on shore and boxes were removed. The kayak was fully reassembled and everything tested ok on the dock. More complete testing will be done in the water on Friday the 11th.
Cause
Initially we suspected the towfish had badly snagged on some underwater obstacle and, mounted to the port side, pulled the kayak over. However, the logs show the kayak deviated from its trackline slightly and traveled some distance before entering a tight clockwise circle. The kayak capsized after some five circles. The compass showed a maximum roll of nearly 45 degrees before the logs ended. If the modem were snagged the kayak should not have traveled so far from the first deviation, and the circles should have been in the other direction with the kayak mounted on the left side.
We believe the steering servo failed first, causing the kayak to first err from its trackline. Eventually the thruster got stuck at an extreme angle, causing the clockwise spin and roll to port. Because the low-level controller has no feedback of the true thruster angle, the thrust limiting that normally occurs at large angles did not engage. Coincidentally, the modem was hanging from the port side of the boat and contributed to pulling the boat over.
It is unknown if the hole in the kayak was leaking before the capsize or not, but winter modifications may include some ability to detect excessive (inch or more) amounts of water in the kayak.
Planned Changes
Operationally, the modem should always be hung from its bridle. Not only does the bridle keep the modem centered, but it also makes it much harder to roll the kayak. Additionally, any slop in the steering servo that cannot be attributed to a visible loose component may be an indication that something inside the servo needs tightening. Removing, maintaining, and reinstalling the servo is only an hour-long job for one person.
The e-stop is connected directly to the motor driver, which then communicates with the low-level Arduino controller over serial. Without the motor driver functioning, the e-stop cannot be used to turn off the steering servo. The Arduino code will be revised to disable the steering servo if it cannot communicate with the motor driver.
Lastly, the MOOS kayak driver will be modified to monitor the vehicle’s roll reported by the compass and disable the thruster if the roll exceeds some critical value. The need for a compact steering assembly makes proper feedback of the steering angle impractical.