In digital manufacturing, data-driven methods are promising to revolutionize various decision-making processes. However, the relationships between variables in high-dimensional data of data-driven decision-making methods are only correlations. Important causal relationships and knowledge between process variables are not considered. Therefore, existing data-driven systems are unstable, which could result in unreliable and dangerous decisions. To establish a stable decision-making model for complex processes with high-dimensional data, a causal-based decision-making framework that combined causal relationships and knowledge between key manufacturing variables was proposed. The causal relationships between state, decision, and objective data were established in the form of a direct acyclic graph formed by breaking an unexcepted loop between variables using a shadow objective variable. Then, causal knowledge of high-dimensional states was introduced to the neural network, forming a stable decision-making model. Compared with data-driven methods used in robotics and manufacturing scenarios, the proposed framework provided better and more stable decisions, particularly in noised environments.

Mechanism models often have limited performance due to incompleteness, while neural networks are acknowledged to lack interpretability and extrapolation stability. Augmenting incomplete mechanism models with neural networks can achieve accurate prediction and naturally inherit mechanism models' interpretability and extrapolation stability. According to the original purpose of establishing mechanism models augmented with neural networks (MMANNs), their effectiveness depends on the accuracy of mechanism models. However, since the directly labeled data of the mechanism model or residuals are inaccessible, this effectiveness criterion is difficult to meet in practice. To overcome this challenge, we introduce two heuristic constraints for MMANNs, and propose a mutual information (MI) guided multiobjective learning framework. In this framework, except for minimizing the end-to-end error and the norm of residuals, the MI of the mechanism model and labeled data is maximized, while the MI of the residuals and labeled data is minimized. Compared with the other seven methods, the verification results of the three cases show that all four MMANNs trained under the proposed framework can learn more accurate mechanism parameters and have better extrapolation stability.

The state prediction of key components in manufacturing systems tends to be risk-sensitive tasks, where prediction accuracy and stability are the two key indicators. The physics-informed neural networks (PINNs), which integrate the advantages of both data-driven models and physics models, are deemed as an effective approach and research trends for stable prediction; however, the potential advantages of PINN are limited for the situations with inaccurate physics models or noisy data, where the balancing of the weights of the data-driven model and physics model is very important for improving the performance of PINN, and it is also a challenge urgently to be addressed. This article proposed a kind of PINN with weighted losses (PNNN-WLs) by uncertainty evaluation for accurate and stable prediction of manufacturing systems, where a novel weight allocation strategy based on uncertainty evaluation by quantifying the variance of prediction errors is proposed, and an improved PINN framework is established for accurate and stable prediction. The proposed approach is verified with open datasets on tool wear prediction, and experimental results show that the prediction accuracy and stability could be obviously improved over existing methods.

Transfer learning has received much attention recently and has been proven to be effective in a wide range of applications, whereas studies on regression problems are still scarce. In this article, we focus on the transfer learning problem for regression under the situations of conditional shift where the source and target domains share the same marginal distribution while having different conditional probability distributions. We propose a new framework called transfer learning based on fuzzy residual (ResTL) which learns the target model by preserving the distribution properties of the source data in a model-agnostic way. First, we formulate the target model by adding fuzzy residual to a model-agnostic source model and reuse the antecedent parameters of the source fuzzy system. Then two methods for bias computation are provided for different considerations, which refer to two ResTL methods called ResTLLSand ResTLRD. Finally, we conduct a series of experiments both on a toy example and several real-world datasets to verify the effectiveness of the proposed method.

Feature-based technologies are widely researched for manufacturing automation. However, in current feature models, features once defined remain constant throughout the whole manufacturing lifecycle. This static feature model is inflexible to support adaptive machining when facing frequent changes to manufacturing resources. This paper presents a new machining feature concept that facilitates responsive changes to the dynamics of machining features in 2.5/3D machining. Basic geometry information for feature construction of complex parts with various intersecting features is represented as a set of meta machining features (MMF). Optimum feature definition is generated adaptively by choosing optimum merging strategies of MMFs according to the capabilities of the selected machine tool, cutter, and cutting parameters. A composite function block for dynamic machining feature modelling is designed with Basic Machining Feature Function Block, Meta Machining Feature Extraction Function Block and Feature Interpreter Function Block. Once changes of the selected machining resources occur, they are informed as input events and machining features are then updated automatically and adaptively based on the event-driven model of function blocks. An example is provided to demonstrate the feasibility and benefits of the developed methodology.