Ada’s Post-Return Analysis: What happened?

Feb 19

This first blog post will serve as context and a backbone behind some of the decisions and design choices which have been made for Raye. As additional blog posts are written, links will be added to point to further details as the content is written. The focus in this update is to provide insights and lessons learned that we have used to make Raye, our latest fully autonomous sailboat to be more robust and reliable.

Raye will be following the Vic-Maui racecourse in the summer of 2021. As we get into the final stages of constructing and testing, our team will release a series of updates/blog posts to show the in-depth technical decisions behind Raye and her underlying systems. We begin by addressing what we learned from the retrieval of our first ocean going vessel Ada. If you are interested in the story of Ada and her voyage, check out our introduction for more details!

Learn more about Ada's Story

On December 1st, 2017, our team got a call from a crew member of the Research Vessel Neil Armstrong that informed us that Ada had been spotted off the coast of Florida. They asked us if we would like them to retrieve the boat for us.

To paint a picture, when this call was made, UBC Sailbot had already spent a year in the design phase of our next boat (yet to be named Raye). Based on our assumptions of Ada’s failure and the limited information we had from Ada’s journey at the time, preliminary design ideas for Raye were being finalized. The retrieval of Ada presented us with an opportunity few engineers get the chance to experience - the ability to dissect what went wrong.

Not Designed for Hurricanes!

As the story goes, Ada didn’t exactly have a good time after she suffered a mechanical failure a few days into her journey. On August 27th, 2016, Ada’s telemetry system reported back to our server that she was traveling at 12.41 knots, this is over twice her designed hull speed! We believe that Ada was able to achieve these speeds by running down enormous swells in hurricane-force winds.

The year 2016 marked the first above-normal hurricane season in the Atlantic since 2012. “There were 15 named storms, including seven hurricanes, three of which were major hurricanes”(2017). Between August 2016 and December 2017, Ada would have been exposed to multiple harsh weather conditions on her journey. The fact that she was found floating after a year and a half is nothing short of a miracle.

The ‘Catastrophic’ Rudder Failure

A core failure of Ada’s system was the rudder, which was essential to directional control. We break down our analysis into three suspected failure modes:

1. When Ada was retrieved, we found that the nut that secured the tiller arm pivot to the rudder was missing. We believe that the rudder experienced large impact loads while riding down the enormous swells caused by high hurricane winds. We also hypothesize that an overactive software controller trying to maintain heading in rough conditions could have also contributed to the added stress on the pivot nut.

These repeated stress cycles, on top of potentially insufficient amounts of thread locking compounds, are the suspected cause for the nut to loosen and come undone. Without this nut, the pivot pin would have fallen out and would have left Ada with no directional control.

Missing pin and bolt that attach the rudder to the tiller arm

2. One of the rudder hard stops had been broken/sheared off. Without a hard stop, the rudder would be able to move beyond the intended limit causing excessive stress on the rudder motor itself.

Assembly which includes rudder hard stops and support for the tiller arm

3. Water had managed to enter the rudder motor box which may have caused rudder motor failure. This box was intended to act as an additional layer to keep water from pooling and potentially entering the rudder motor itself. But as the image shows, this secondary layer of protection failed.

The rubber bellow attached to the box, which allowed the tiller push rod to penetrate the rudder motor box, had cracked. This may have been where water ingress happened, but it may have also happened along the box lid itself.

Rudder box, designed to add a secondary layer of protection from the rudder motor

There are two possible explanations behind why the rubber bellow may have failed. One reason was the shearing of the rudder hard stop, which would have overstretched the bellows. Another possible reason was the composition of the rubber, which was not intended for a year of UV exposure and temperature fluctuations.

Rudder box gasket and bellows boot to allow tiller arm to penetrate

Rudder box bellows boot worn, likely due to extensive UV exposure

As a result of Ada’s failure analysis, Raye’s rudder design has been revised to reflect the lessons learned. We are avoiding rubber bellows because UV tolerant rubber is hard to find. Furthermore, our current design uses space more efficiently by integrating rotary shaft seals that can handle salt water exposure; these seals are protected from direct UV exposure. In a future blog post, we will detail the journey taken to improve Raye’s rudder design to make her journey successful.

Where Did Ada’s Rigging Go?

When the R/V Neil Armstrong spotted Ada, her rigging tripod was notably missing. Ada was designed with a windsurfing rig to primarily sail downwind - going from Newfoundland to Ireland is primarily downwind sailing. The windsurfing rig was supported by a tripod structure as shown here pre-launch.

Ada being lowered into the water for testing - tripod rig present to support windsurfing sail

To make sense of our overall investigation of the missing rigging, we broke down our analysis into four parts:

1. Rig Mounting Hardware

Our inspection found that the two threaded studs securing each of the chainplates, which held the rigging stays (ropes) to the hull, had sheared off. We believe that high wind conditions caused these fasteners to be repeatedly loaded with forces well above their design limits.

We found what appeared to be fatigue striations on the remaining parts of the studs. This made us believe that the stud sheared due to fatigue failure, instead of a single catastrophic failure.

Ada in Rusty Hut - notably no tripod rig present!

Bolt heads missing that were responsible for securing the chainplates

2. Rig Impact Damage

Deck partial puncture damage - does not fully penetrate hull, likely caused when the rig detached

There exists blunt impact damage on the amidships port side of the deck. This could have been caused by the tip of the tripod rig crashing down. Under this assumption, we believe that each of the tripod legs failed in succession, specifically the port tripod leg failed first.

Remarkably this impact did not puncture through the honeycomb Nomex structure that was used to build Ada’s deck. Underneath the deck in the impact zone, we can feel the indent and can confirm that it did not penetrate. The bottom face skin of the honeycomb structure held up well!

3. Rig Abrasive Wear

There is extensive abrasive wear on the gunwales and transom as shown in the picture. We hypothesize that the wear on the gunwale and transom are a result of the rig being caught, dragging overtime before becoming free and released into the ocean. We also hypothesize that the rig was responsible for damaging the wind sensors and GPS units on the deck. The rig in its partially attached condition may have swept across the deck, scything down spindly components and causing the visible wear seen along the transom.

Wear on the transom of Ada, likely due to the rig still partially attached

A support for the clothesline rigging broken, the hollow tube this support held worked to prevent entanglement

4. Rigging Ropes

The support that held a hollow tube to mitigate rig entanglement was ripped off the deck surface as shown on the right. Some of the rig lines were frayed and worn out. However, given that these lines were exposed to continuous sun, saltwater, storms for a year and a half they have held up remarkably well.

The lines used on Ada are as follows:

“Clothesline” - New England Ropes T-100 Technora Double Braid, ⅛” diameter
- Technora-Spectra Braid core, braided polyester sheath
Main Sheet - SAMSON ROPE–WarpSpeed Dyneema Double Braid (Blue), ¼” diameter
- Dyneema core, braided polyester sheath

Worn ropes, frayed but still in one piece!

Another example of the durable rig ropes that were used

As a result of Ada’s rig analysis, the rig design for Raye is completely different from Ada’s. For one, Raye uses a traditional jib and main sail configuration to allow for more diverse points of sails. Furthermore, the weakness in the chainplate design holding Ada’s tripod has been thoroughly analyzed and improved upon for Raye. We’ll explore more of Raye’s rig design in a future post!

Water and Electricity Don’t Mix!

Ada suffered extensive damage to its electronics, namely those that were outside and attached to the deck. However, there are also lessons about how said devices connected to internally-stored enclosures are meant to keep water out.

Ada before the voyage with all her sensors intact (wind sensor, light, AIS whip antenna, A25 GPS antenna)

There are three parts to our analysis of sensor/electronics damage:

1. Deck Sensor Damages

Almost all of our sensors on the deck sustained some form of damage. Of prominence is the missing starboard side wind sensor which is broken at the base. The other wind sensor also has sustained damage from the year-and-a-half-long exposure to ocean conditions.

Ada with missing sensors (wind sensor, light, AIS whip antenna, A25 GPS antenna), one of the wind sensor rods broken clean off. Likely culprit the rig detaching

GPS V104 molded into the hull, minor damage but mainly intact

Diagram depicting a potential failure mode of the sensors due to waves crashing and washing over the deck repeatedly over a long period

Ada was also equipped with a whip antenna for AIS receiving, two Hemisphere A25 antennas for redundant GPS, and a white light for detection. These devices were missing when we found Ada. Interestingly, our main GPS, the Hemisphere V104, was molded into the deck and sustained some damage but otherwise remained on the deck intact.

As mentioned before, the suspected cause of sensor damage was the loose rig sweeping across the deck and smashing against these fragile components. However, we also believe that water washing over the deck repeatedly caused an upwards force to be present, which may have contributed to fatigue and/or detachment of these sensors.

Solar panel with a permanent layer of salt decreasing its effectiveness

2. Solar Panel Performance Impacts

The performance of Ada’s solar panels suffered from dry saltwater obfuscating the cells themselves. After weathering a year and a half at sea, a seemingly permanent thin layer formed on all panels which have reduced their ability to generate a voltage. However, these Solbian Panels are remarkably durable as they are still able to produce power.

However, stormy weather had a bigger impact on our ability to regenerate power via solar panels. Ada’s power consumption was likely underestimated or its power regeneration was overestimated. Either way, we did not have enough power, which meant that after 8 or so days at sea (after our rudder failure), Ada went silent for a 5 day period.

3. Cable Water Ingress

Cables that connected external devices to electronic enclosures in Ada’s internals were subjected to capillary action. Some of the solar panel connectors had completely detached from the unit itself. On the panels where this occurred, we also noticed water inside the cable itself reaching the connection point of our battery boxes inside Ada.

This was a surprise. It meant that our waterproof enclosures below the deck had to actively keep water out. However, water had seeped through the cables instead of through the lids as there was evidence of internal condensation at the boxes where the solar panel connector was detached due to corrosion.

Corrosion present internal to the waterproof solar panel connector

Other electronic enclosures suffered the same problem where cables that had been terminated externally to the boat allowed the same kind of water to enter via the cables. This was clear evidence of capillary action, where tight seepage paths such as cables allow the eventual ingress of water.

Surprisingly, the box which housed our satellite modem and flush deck mounted satellite antenna successfully kept water out. It is this enclosure that allowed us to remain in contact with Ada up to 3 months after launch.

pasted image 0 (3).png — Corrosion detected within one of the electronic enclosures - likely due to capillary action moving water down improperly terminated or damaged cables

Condensation on the inside of one of our electronics enclosures

Satellite modem electronics box kept dry

Deck mounted satellite antenna - very robust design

When we first lost contact with Ada, we believed at the time that she was lost at sea forever. However, 5 days later, Ada came back to life and sent a message over Iridium’s satellite network. The message itself was not useful, but it was proof that Ada was still alive and gave us the ability to roughly triangulate Ada’s location by analyzing the position of the satellites involved.

Taking lessons from Ada, Raye’s deck-mounted sensors will be faired into the hull to reduce the risk of detachment. On top of that, Raye’s solar panel capacity and positioning have been optimized to compensate for dry spots and partial shading. Above all, the cabling system has been overhauled to address the waterproofing issues found in Ada.

A blog post on the redesign of the cabling, waterproofing, and electronics will be released in a future update/blog.

Hatches, Seals, and Enclosures

Waterproofing access points to Ada’s internals turned out to be a big lesson learned, stressing the importance of getting it right the first time. A big lesson learned in working with Ada pre-launch was how important it is to have quick access to the boat’s internal hardware.

Ada’s main hatch - many bolts to undo to gain access with potted glands for cable penetration

Our small circular hatches offered a good tradeoff of keeping water out while maintaining easy access. The enclosures below deck had success with keeping water out where cables did not leak water in as was described above.

However, the main hatch was a custom-designed hatch with lots of bolts and screws. This meant that anytime the software or electrical team needed to fix something it was an involved process. The hatch worked reasonably well, though some evidence of minor leakage was present.

Looking into the main hatch’s design, sandwiching a thin piece of hard rubber between two relatively thin (0.25") pieces of plywood does not give much compression to create a seal to fill in gaps.

Example of uncovered circular hatch - fairly dry in this compartment

Gasket seal failed to keep water out in some areas

Rudder box seal failure - some water pooling within the rudder box

We believe that the hatch itself may have deformed more than the sealing surface. Efforts were made to make the sealing surface super smooth and uniform, but this only goes so far. This was a similar situation for the rudder box lid, where water was able to get into the enclosure. However, it’s unclear if the rudder box lid seal or the bellow failed first.

The battery boxes which had properly terminated connectors kept water out as intended, despite them sitting in water that had gotten in through some leakage of the circular hatches.

Our study on Ada’s sealing led us to use highly deformed gaskets for Raye. The idea is that gaskets that are softer and deform more should better seal imperfections on the sealing surface. Furthermore, these kinds of gaskets allow for higher contact force, making them more effective at resisting pressure from waves falling on, or washing over, the deck.

An in-depth post on the analysis of waterproofing and enclosures is a topic for another update/blog post.

Dry battery box demonstrating that the waterproof enclosure worked

Conclusion

Our team got valuable feedback from being able to carefully inspect what failed and what worked with Ada. The list of lessons learned grew tremendously from the failure analysis we were able to conduct as a result of retrieving Ada. The investigation has had design impacts for Raye, most notably, the rudder design was extensively re-thought. In the coming blog posts we will refer back to Ada’s design and some of the points outlined in this failure analysis to describe how we have made modifications to improve and mitigate the identified weak points for Raye’s design/construction.

UBC Sailbot