Modeling and predicting thermodynamic properties of binary mixtures is crucial in chemical engineering. Understanding how the mixture behavior, represented as a scalar matrix, depends on properties of pure substances offers valuable insights into substance interactions. While there is robust support for pattern-based sorting of matrices in general, limited support exists for evaluating patterns against external attributes available in many fields. In this paper, we introduce an interactive software to detect and analyze block patterns in scalar matrices using annotated domain knowledge. Therefore, we revisit canonical matrix patterns, explore their translation to this application, and describe a workflow to fit the matrix ordering. Our interactive software allows users to explore hierarchical aggregation levels, rating them based on additional domain-specific data properties of various type. Using our tool, chemical engineers are able to identify and interpret cluster structures in their mixture data. These insights contribute to the development of improved prediction methods for thermodynamic properties, forming the foundation for modeling and simulation in chemical engineering.