An engine failure simulation test was carried out on a Z6170 marine medium speed diesel engine. The multi-dimensional characteristic parameters of instantaneous speed signal monitored under different operating conditions were normalized by the MIN-MAX method

and the t-distributed stochastic neighbor embedding(t-SNE) algorithm was used to reduce the dimensionality of the characteristic parameters. The random forest(RF) algorithm was used to establish an engine cylinder fault diagnosis model. Results show that based on the monitoring of instantaneous speed

which contains rich information of engine operating condition

the evaluation of diesel engine operating conditions and the cylinder fault diagnosis model established by the t-SNE and RF algorithm can effectively evaluate the health condition of an diesel engine

so as to quarantee it running safely and reliably.