The traditional Practical Byzantine Fault Tolerance (PBFT) approach suffers from three critical deficiencies: arbitrary primary node election, excessive network transmission overhead, coupled with the absence of node incentive mechanisms. To address these issues, this study proposes a refined PBFT strategy utilizing dual scoring (Double Scoring Practical Byzantine Fault Tolerance, DS-PBFT). The algorithm innovatively combines hardware performance evaluation with a dual-dimensional node scoring system. The algorithm first employs bucket sorting technology to quantitatively evaluate node hardware resources, followed by constructing a comprehensive scoring model through credit values and recommendation values. According to the scoring outcomes, the framework hierarchically divides nodes into primary node, follower node and backup nodes groups in a 1:4:5 ratio, substantially decreasing the quantity of nodes involved in consensus. Additionally, this approach streamlines the Commit-Reply stages within the consistency protocol, substantially reducing communication overhead. Experimental validation demonstrates that DS-PBFT maintains security while achieving notable improvements in consensus efficiency, significant reductions in communication costs, and enhanced defense capabilities against malicious nodes.