Abstract Double-electrode submerged arc welding (DE-SAW) is an innovative process to achieve the same metal deposition rate at reduced heat input. To maximally take advantage of this new capability, the heat input needed to produce acceptable welds needs to be reduced. A joint gap is successfully introduced to reduce the required heat input but it introduces variations affecting currents/heat input. To minimize the resultant variations in currents and resultant welds, the process is modeled and a multivariable predictive control algorithm is developed to control the currents at desired levels. Experiments verified the effectiveness of the resultant feedback controlled DE-SAW process.