Improved particle swarm algorithm is adopted in order to evaluate the flatness measurement error accurately. Firstly, mathematical model of flatness error evaluation is established. Secondly, particle swarm algorithm is improved, which is included inertia weight control based on membership function of double sigmoid type, Triangle function adjustment process, and selecting fitness function. Finally, the algorithm termination condition and flow are given. Experimental simulation results show that the convergence of improved particle swarm optimization algorithm is fast, the flatness error is 9.496 μm at an average of 30 experiments, which is smaller than that of other optimization; the standard deviation of the experiment is 0.048 2 μm, which is smaller than that of other algorithms as well, so that the evaluation precision is improved effectively.