In this paper, an electronically controlled diesel engine
fueled
with Fischer–Tropsch fuel was selected to optimize soot and
NO
x
emissions. First, the effects of injection
parameters on exhaust performance and combustion properties were studied
on an engine test bench and then a prediction model based on a support
vector machine (SVM) was established according to the test results.
On this basis, a decision analysis of soot and NO
x
solutions assigned with different weights was performed based
on the TOPSIS analysis method. It turned out that the “trade-off”
relation between soot and NO
x
emission
was improved effectively. As a matter of fact, the Pareto front selected
by this method showed a significant decline compared with the original
operating points, in which soot declined by 3.7–7.1% and NO
x
declined by 1.2–2.6%. Finally, the
experiments were used to confirm the validity of the results, which
indicated that the Pareto front corresponded well with the test value.
The maximum relative error between the soot Pareto front and the measured
value is 8% while it is 5% for NO
x
emission,
and the R
2 values of soot and NO
x
under various conditions are more than 0.9. This
instance proved that research on diesel engine emission optimization
based on the SVM and NSGA-II is feasible and valid.