This diploma thesis investigates data mining and data analysis methods which are applicable to heavy plants and machinery. The aim of these techniques is to facilitate the extraction of information, knowledge and understanding from machine data combined with meta-data. The results can be used to support: Commissioning processes, engineering feedback, structural health monitoring, predictive maintenance and efficiency measurements. Several approaches and concepts for time series data mining are reviewed and an introduction to the engineering characteristics of heavy machinery is given, whereby commonly used machine components, such as hydraulics, are presented. The advent of Cyber-Physical Systems (CPS) enables the collection, management and analysis of sensor data from physical systems distributed on a global scale. The data analytics proposed here enables the solution of the inverse problem associated with the observation of the physical system. In this manner the solution abides by the laws of physics governing the system. This enables the use of causality as a measure of significance and not mere correlation as used in classical data mining. Within this thesis the approach of lexical analysis is presented as a solution for sensor data analysis. The lexical analysis combines the abilities of dealing with inverse problems in real time, clustering machine sensor data and lossless compression of sensor data in a complete generic and continuously variable manner. To verify the functionality of the lexical analysis, sensor data for six years of three structurally identical reclaimers - each reclaimer had several gigabytes of sensor data - with each twelve sensors at a sampling time of one second was analysed. The lexical analysis delivers an approach for the analysis of machine sensor data based on a physical model.