A Framework for Exploratory Analysis of Multivariate Time-Series Data Emanating from Batch Processes

This thesis presents a framework for managing, structuring, and analyzing multivariate time-series (MVTS) data derived from batch processes in industrial settings. The framework addresses key challenges associated with handling this data type, including high dimensionality, large volume, inconsistent metadata, and the need for adaptable analysis. By connecting theoretical data science methods with real-world industrial needs, it provides a practical solution to these challenges. The thesis focuses on three main areas: data structuring and metadata management, computational analysis, and visualization. The framework introduces standardized techniques for data ingestion, structuring, and metadata integration, ensuring consistent handling of diverse MVTS datasets. This approach enables reliable analysis of each dataset and facilitates comparisons across multiple datasets. A modular computational framework supports automated event detection, segmentation, and flexible KPI calculations tailored to specific industrial processes. Within this framework, mechanisms for anomaly detection, data integrity validation, and error logging ensure reliability and maintain a continuous workflow. Additionally, it offers interactive visualization methods for detailed exploration of MVTS data patterns. These include customizable stacked time-series plots, heatmaps, and anomaly annotations. These tools enable detailed insights into process quality and efficiency, supporting both exploratory and operational analysis. The framework¿s adaptability and effectiveness are validated through a case study, the KaPPI system, demonstrating its application in the construction sector for ground stabilization projects. Selected framework components have also been tested in other projects, demonstrating its flexibility and potential for broader use across batch process industries. These contributions are documented in peer-reviewed publications, highlighting the framework¿s relevance to diverse applications of MVTS data analysis.