Prognostics and health management (PHM) has emerged in the last few years as one of the most efficient approaches in failure prevention, predicting reliability and remaining useful life of various engineering systems and components. The diagnostics of unusual performance trends typically is associated with the requirement for a continuous monitoring of system's behaviour using data from sensors. Often it is necessary to monitor the system in real-time, especially in the case of safety critical applications, to predict when a fault will occur and/or to assess the remaining useful life. In this paper we present an investigation on the suitability of a number of data analysis algorithms to realise real-time diagnostics, prognostics and health monitoring of engineering systems. The focus is on the following two techniques for data-driven PHM: (1) Euclidean Distance (ED) and (2) Mahalanobis Distance (MD). These techniques are implemented into a real-time operating platform for health management and tested using data from high power light emitting diodes (LEDs). These LEDs are tested for the PHM of semiconductor devices/solid state lighting systems and include monitoring of current, temperature and light intensity. We also demonstrate the real-time PHM capability of these algorithms by programming the associated sensor data manipulation and numerical computations. Although for the application reported in this work the real-time aspect is not essential, the actual prognostics techniques can be applied in other applications where this capability might be necessary.