Triton 2025 - Magazine - Page 41
PAGE HEADER
PERSPECTIVES
at patterns over time and identifying
different instances of failure, such as a
ship drifting off course due to a power
blackout. Intriguingly, the AIS system
often continues to function during a
blackout, allowing for post-failure analysis
even if the ship has temporarily lost
manoeuvrability.
By inputting this data into a high-speed
ship simulator, Axel can simulate millions
of voyages in a matter of minutes. These
simulations introduce variables - like
mechanical failures or human errors - and
predict how likely they are to result in
an allision with a bridge or alternatively
running aground. This creates a datadriven foundation for understanding and
mitigating risk in a way that far surpasses
traditional engineering assumptions.
Bene昀椀ts and Limitations:
The Double-Edged Sword
of AI
Despite its promise, AI modelling is not
without its challenges. For one, actual
allision incidents are rare in speci昀椀c
locations - occurring perhaps once every
昀椀ve years. This rarity makes it dif昀椀cult
to build a robust training dataset for
AI. What’s more, not all anomalies are
failures. A ship slowing down might simply
be “slow steaming” rather than drifting
due to mechanical issues. Discerning
the difference requires a nuanced
understanding of maritime behaviour something AI doesn’t yet fully possess.
unrecorded and unreported. Insurers, such
as The Swedish Club, may be unaware
of near misses that never escalate to
claimable events. However, AI does offer
the potential to uncover these hidden
incidents in the future, improving the
industry’s overall understanding of risk and
informing better insurance and operational
policies.
Introducing STAPS: A
New Era in Ship Tra昀케c
Simulation
To aid in this mission, Axel has developed
his own ship traf昀椀c simulation model
known as STAPS (Ship Traf昀椀c Allision
Probability Simulator). STAPS represents
a signi昀椀cant leap forward in modelling
allision risk, building on groundwork that
hasn’t seen major innovation since the
late 1990s.
“In layman’s terms, STAPS can simulate
the behaviour of vessels over the course of
a year in just a few minutes”, he says. “By
running the simulation millions of times
and tweaking various inputs - like fatiguerelated human errors or ship engine
failures - I can derive a distribution of likely
outcomes, identifying scenarios
with the highest probability of
allision.”
As Axel explains, “While AI can recognise
unusual behaviours, it lacks the
contextual expertise and judgement that
experienced sailors bring. This highlights
the importance of integrating human
know-how into AI frameworks. As AI
evolves, it will require input from maritime
professionals to re昀椀ne its interpretation
of what constitutes a true navigational
failure.”
Additionally, a vast amount of failure
data - such as blackouts that occur at sea
but are resolved without incident - goes
Triton
41
41
Triton 2025
2025
