Tuning of box jenkins model parameters to improve accuracy

i am using bj function to identify my system using

sys = bj(data, nb, nc, nd, nf, nk);

after getting reduced order model using Akaike information criteria AIC, i am unable to achieve accurate results.

Is there any method to "tune the model parameters of bj model to improve the accuracy of the model fit, as we manaually tune the parameters in OE and ARX model.