You indicate that your integral becomes infinite when b is non-zero. You try to compensate for that by using (b==0) * integral() thinking that it will "select" the integral() calculation when b is 0 and thinking that otherwise it will skip it. But that is not what happens. When you calculate (b==0)*integral() then both sides are calculated no matter what the value of b is. When b is 0 then that is fine, as you get (0==0)*finite_value which is the finite_value. But when b is non-zero then you get (nonzero==0)*infinite_value which is 0*infinite_value which is NaN.
You cannot use the logical_condition*expression calculation form when the expression can be infinite or nan in any situation where the logical_condition is false.
To deal with this, you will either need to use the symbolic toolbox piecewise() function, or you will need to write a small function that uses if or logical indexing so that you do not calculate the integral in the b ~= 0 case.
