Abstract:Yield and quality improvement is of paramount importance to any manufacturing company. One of the ways of improving yield is through discovery of the root causal factors affecting yield. We propose the use of data-driven interpretable causal models to identify key factors affecting yield. We focus on factors that are measured in different stages of production and testing in the manufacturing cycle of a product. We apply causal structure learning techniques on real data collected from this line. Specifically, the goal of this work is to learn interpretable causal models from observational data produced by manufacturing lines. Emphasis has been given to the interpretability of the models to make them actionable in the field of manufacturing. We highlight the challenges presented by assembly line data and propose ways to alleviate <a class="link-external link-http" href="http://them.We" rel="external noopener nofollow">this http URL</a> also identify unique characteristics of data originating from assembly lines and how to leverage them in order to improve causal discovery. Standard evaluation techniques for causal structure learning shows that the learned causal models seem to closely represent the underlying latent causal relationship between different factors in the production process. These results were also validated by manufacturing domain experts who found them promising. This work demonstrates how data mining and knowledge discovery can be used for root cause analysis in the domain of manufacturing and connected industry.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to solve the problems of increasing production and improving quality in the manufacturing industry. Specifically, the author achieves this goal by discovering the causal factors that affect these key indicators. The following are the core problems of the paper: 1. **Identifying the joint causal structure**: - One of the goals of the paper is to identify the joint causal structure in a specific manufacturing area. This includes determining the causal relationships between various variables in the production process. 2. **Determining the key causal factors**: - Another goal is to focus on those key causal factors that can increase production. Specifically, given various measurement data recorded during the production and testing of products, the author hopes to identify the important factors that affect production and their complex interrelationships. ### Background and challenges Traditionally, the methods for solving these problems are through Design of Experiments (DoE). However, the DoE method is both expensive and time - consuming to conduct experiments in the actual production environment, so usually only a small number of factors can be examined. As the manufacturing process becomes more and more complex, it becomes very difficult to select appropriate factors for further investigation. Currently, most manufacturing experts rely on domain knowledge, intuition, and basic statistical methods to guide their efforts to increase production and improve quality. ### Solutions To solve the above problems, the author proposes a method based on data mining and knowledge discovery, especially applying causal structure learning techniques to identify the key causal relationships in the manufacturing production line. The advantages of this method are: - **Interpretability**: The learned causal model has high interpretability and can be used by practitioners to take meaningful actions. - **Meeting challenges**: The author identifies the challenges brought by assembly line data and proposes corresponding solutions, such as using the time - sequence information of data, adjusting the conditional independence test parameters, and clustering highly correlated features. - **Verifying results**: Through standard evaluation techniques and verification by domain experts, it is proved that the learned causal model is highly consistent with the actual causal relationships in the production process. ### Conclusions This work shows how to use data mining and knowledge discovery techniques for root cause analysis in the manufacturing industry. Through the application of a large amount of production data, the author successfully identifies the key factors that affect production and provides actionable insights to help manufacturing enterprises improve efficiency and product quality.

Causal Discovery for Manufacturing Domains

Understanding unforeseen production downtimes in manufacturing processes using log data-driven causal reasoning

$\texttt{causalAssembly}$: Generating Realistic Production Data for Benchmarking Causal Discovery

Spatial-Temporal Causality Modeling for Industrial Processes with a Knowledge-Data Guided Reinforcement Learning

Two-stage Approach to Causality Analysis-Based Quality Problem Solving for Discrete Manufacturing Systems

A hierarchical ensemble causal structure learning approach for wafer manufacturing

Interpretability of Causal Discovery in Tracking Deterioration in a Highly Dynamic Process

Data-driven dynamic causality analysis of industrial systems using interpretable machine learning and process mining

Increasing the Accessibility of Causal Domain Knowledge via Causal Information Extraction Methods: A Case Study in the Semiconductor Manufacturing Industry

Modeling and Discovering Direct Causes for Predictive Models

Neural Network Weight Comparison for Industrial Causality Discovering and Its Soft Sensing Application

COKE: Causal Discovery with Chronological Order and Expert Knowledge in High Proportion of Missing Manufacturing Data

Anwendung von Causal-Discovery-Algorithmen zur Root-Cause-Analyse in der Fahrzeugmontage

The Landscape of Causal Discovery Data: Grounding Causal Discovery in Real-World Applications

Data-driven Dynamic Inferential Sensors Based on Causality Analysis

Causality, Causal Discovery, and Causal Inference in Structural Engineering

Learning domain-specific causal discovery from time series

LLM-Enhanced Causal Discovery in Temporal Domain from Interventional Data

Graph-enabled cognitive digital twins for causal inference in maintenance processes

Causal discovery in a complex industrial system: A time series benchmark