The restoration of one- to three-limb failures in traditional parallel six-dimensional acceleration sensing mechanisms has been addressed, while the restoration rate for four-limb failures remains generally low. To overcome this limitation, a reconfigurable parallel six-dimensional acceleration sensing mechanism is designed, and a fault-tolerant processing algorithm matched to four-limb failures is proposed. First, a metamorphic mass block is designed; by rotating the spherical joints on the mass block, the orientation layout of the limbs can be reconfigured. Second, based on the inherent constraint relationships among the limbs of the reconfigurable sensing mechanism, six charge coordination equations are derived, and a coordinated closed loop in the shape of a regular hexagon is constructed. By introducing the Cartesian product and based on the coordinated closed loop, a self-restoration algorithm aimed at four-limb failures is developed. The study reveals that 471 types of four-limb failures can be restored using this algorithm, achieving a self-restoration rate as high as 95.2%. Virtual experimental verification shows that after integrating the fault self-restoration algorithm, the decoupling error of the six-dimensional acceleration sensor is reduced to as low as 0.002%, fully meeting the high standards of precision, stability, and reliability required in industrial applications.