The article provides proof that the diagnostics of marine internal combustion engines and other ship power plant machines should take into account the randomness and unpredictability of certain events, such as wear, damage, the variations of mechanical and thermal loads, etc., which take place during machine operation. In the article, the energy E, like the other forms (methods) that it can be converted into (heat and work), is considered the random variable Et; at time t, this variable has the mean value
E
¯
t
{\bar E_t}
, which is the observed value of the statistic
E
¯
s
t
{\bar E_{st}}
with an asymptotically normal distribution
N
(
E
(
E
t
)
,
σ
t
n
)
N\left( {E\left( {{E_t}} \right),{{{\sigma _t}} \over {\sqrt n }}} \right)
, irrespective of the functional form of the random variable Et. A proof is given that shows that the expected value estimated in the above way, considering the time t of the performance of task Z by a marine internal combustion engine or other ship power plant machine, can be used to determine the machine’s possible action (DM). When compared to the required action (DW) needed for task Z to be performed, this possible action makes it possible to formulate an operating diagnosis concerning whether the engine or machine of concern is able to perform task Z. It is assumed that an energy device of this type is able to perform a given task when the inequality DM≥DW holds. Otherwise, when DM < DW, the device cannot perform the task for which it was adopted in the design and manufacturing phase, which means that it is in the incapability state, although it still can be started and convert energy into the form of heat or work..