The aero turbine is spun by high-temperature and high-pressure burning gases. The practice has proven that the directional solidification (DS) turbine blade with perfect column grains has still excellent high-temperature performance in this kind of working environment. This means that the size and orientation of column grains have great influence on the high-temperature property and performance of turbine blades. On the other hand, the high-quality blade is not easy to be produced in DS process due to the difficulty of obtaining the desired temperature field needed to produce the grains with ideal morphology. In addition, the growth of columnar grains in the wide-chord hollow guide blade is obstructed by the complex camber and the platform. How to produce turbine blades with desired microstructures is the key problem in the DS process. Numerical simulation of the DS process is an effective way to investigate the growth and the morphology of the grains and hence to optimize the process. In this work, a mathematical-physical model for simulating the DS process of wide-chord blade is established in which nucleation and grain growth in the blade in the DS process are modeled by the cellular automation (CA) method with multi-scale dynamic bidirectional coupling technology. Some general analytic indicators are proposed to assess the morphology of mushy zone and grains in a blade quantitatively. Based on the simulated results by using the usual starter blocks 1, 2 and 3, a new starter block is designed considering numerically controlled cutting. Temperature fields and grains in DS processes and corresponding indicators at different withdrawal rates for above 4 starter blocks are numerically predicted to investigate the influences of varying these technological parameters, and hence to determine the influence mechanism to the DS process. For comparison, the DS validation experiments by using starter blocks 1, 2 and 3 have been carried out. The numerical and experimental results agree well, their morphologies including those faulty grains are similar. It is found that higher withdrawal rate leads to larger concavation of mushy zone, but the effect of chill is stronger than that of withdrawal rate if the contact area between casting and chill plate is large enough. Better grain structure in a blade is achieved by starter block 3 than by starter blocks 1 and 2. By starter block 4, the amount of column grains is larger and the amount of lateral grain boundaries is smaller, compared with that of starter blocks 1, 2 and 3. Therefore higher withdrawal rate could be adoptable without excessive concavation of mushy zone, resulting in parallel column grains, finer dendrites in the blade, and much higher blade productivity. Optimum withdrawal rates are also determined for starter blocks 3 and 4.