Product–service systems (PSS) accelerate the transition of value creation patterns for manufacturing industries, from product design and production to the delivery of overall solution integrating products and services. Existing PSS configuration solutions provide customers with preferable product modules and service modules characterized by the module granularity. Every service module is essentially a whole service flow. However, the performance of the PSS configuration solution is greatly influenced by service details. In summary, this paper studied the configuration optimization of product-oriented PSS using a fine-grained perspective. A multilayer network composed of (i) a product layer, (ii) a service layer, and (iii) a resource layer was constructed to represent the elements (product parts, service activities, resources) and relationships in PSS. Service activities selection and resource allocation were considered jointly to construct the mathematical model of PSS configuration optimization, thus enabling the calculation of optimizing objectives (time, cost, and reliability) under constraints closer to the actual implementation. The importance degree of service activity was considered to improve the performance of service activities with higher importance. Corresponding algorithms were improved and applied for obtaining the optimal solutions. The case study in the automotive industry shows the various advantages of the proposed method.

Assembly sequence planning is a critical step of assembly planning in product digital manufacturing. It is a combinational optimization problem with strong constraints. Many studies devoted to propose intelligent algorithms for efficiently finding a good assembly sequence to reduce the manufacturing time and cost. Considering the unfavorable effects of penalty function in the traditional algorithms, a new discrete firefly algorithm is proposed based on a double-population search mechanism for the assembly sequence planning problem. The mechanism can guarantee the population diversity and enhance the local and global search capabilities by using the parallel evolution of feasible and infeasible solutions. All parts composed of the assembly are assigned as the firefly positions, and the corresponding movement direction and distance of each firefly are defined using vector operations. Three common objectives, including assembly stability, assembly polymerization and change number of assembly direction, are taken into account in the fitness function. The proposed approach is successfully applied in a real-world assembly sequence planning case. The sizes of feasible and infeasible populations are adequately discussed and compared, of which the optimal size combination is used for initializing the firefly algorithm. The application results validate the feasibility and effectiveness of the discrete double-population firefly algorithm for solving assembly sequence planning problem.

Welding quality inspection is critical to the quality control of the welding structure. Traditional manual detection requires experienced workers and the method is time-consuming. Currently, deep learning has made great progress in the field of image recognition. However, in terms of industrial defect detection, the contradiction between huge computational parameters and limited imbalanced samples still exists, which makes the deep learning method unable to play its role to the maximum extent. In this case, a combination of deep learning methods and transfer learning methods is considered to improve the performance of the model on limited and imbalanced data sets. The VGG16 model, which is pre-trained on a large number of source data sets, is fine-tuned on our target data sets. In this paper, a new DTN-VGG16 structure is proposed, in which a Global Average Pooling (GAP) layer, Batch Normalization (BN) layer, and Soft-max classifier are added on top of the frozen pre-training model, which can effectively reduce the number of parameters and accelerate the convergence speed of the model. In addition, focal loss function is used to replace the commonly used multi-class cross entropy loss function. By adjusting the parameters of the loss function, the training process has better convergence. Experimental results show that DTN-VGG16 has better performance than other traditional machine learning methods and deep learning models. And the proposed model has good robustness and generalization performance when valid on the real data of aerospace welding seams which is difficult to learn.