Transfer Learning-based Dynamic Balancing For Human-Robot Collaborative Disassembly Line

End-of-life products have the dual attributes of environmental pollution and resource regeneration,and proper disposal can bring great economic benefits.Disassembly,as a key step in the remanufacturing process,is the basis for achieving the remanufacturing goal,and thus has received extensive attention from academia.However,for a long time,most researches are mainly oriented to the deterministic disassembly environment,which cannot cope with the various uncertainties existing in the actual industry and greatly limit their applications.In recent years,with the development of information technology,disassembly systems that can effectively cope with environmental uncertainties such as product quality have attracted attention and made progress.In addition,the construction of automatic disassembly systems represented by robotic disassembly lines is a long-term goal of disassembly research.However,the deficiencies in flexibility and cognitive ability of robots lead to some complex tasks that still need to be performed by humans.Therefore,the construction of human-robot collaborative disassembly lines that can cope with various environmental uncertainties should be a feasible direction for the current and future disassembly field.Aiming at the above issues,this paper constructs a model of dynamic line balancing problem of human-robot collaborative disassembly line under product quality uncertainty and propose an dynamic line balancing algorithm based on feature transfer learning and sample transfer learning,respectively.The effectiveness of the proposed model and algorithm is verified by designing and implementing a dynamic line balancing system of a human-robot collaborative disassembly line with uncertain product quality.The main research work of this paper is described as follows.(1)Research on the model of the dynamic line balancing problem of human-robot collaborative disassembly line under uncertain product quality.The uncertainty of humans and the inherent uncertainty of the End-of-life(EOL)products are the main uncertainties in the human-robot collaborative disassembly line,which determines that the disassembly line balancing problem is a dynamic optimization problem.Therefore,this paper constructs a human-robot collaborative disassembly line that can track uncertainties and calculate the optimal solution in the new environment.A task-based dynamic disassembly process model is developed.A model of the dynamic line balancing problem of human-robtot collaborative disassembly line with uncertain product quality is constructed to minimize the cycle time,number of humans,and the number of operators.The proposed problem model is validated by the exact solution method.(2)Research on the dynamic line balancing algorithm of the dynamic line balancing problem of human-robot collaborative disassembly line under uncertain product quality.Aiming at the human-robot collaborative disassembly line model under uncertain product quality,a dynamic line balancing algorithm is proposed based on feature transfer learning.First,the balanced distribution adaptation method finds the optimal feature transformations between the historical environment and the current environment,thus generating a latent space.Then,the closest solution to the historical Pareto optimal solution set is found in the latent space and used as the initial population in the new environment,which eventually accelerates the evolutionary process.In addition,an improved dynamic line balancing algorithm based on sample transfer learning is proposed.The algorithm searches for the most similar environment to the target environment from multiple historical environments to avoid negative transfer to improve the solution quality and convergence speed.The proposed two algorithms are applied to the small and large-scale dynamic line balancing problems of human-robot collaborative disassembly line and the effectiveness and superiority of the proposed algorithms are verified through comparative experiments.(3)Design of dynamic line balancing system of human-robot collaborative disassembly line.To better solve the dynamic line balancing problem of human-robot collaborative disassembly line in practical applications and verify the effectiveness of the proposed model and algorithms,a dynamic line balancing system of human-robot collaborative disassembly line based on the model of the dynamic line balancing problem of human-robot collaborative disassembly line is built,which contains multiple dynamic line balancing algorithm based on transfer learning.