Rotating blades are considered to be the one of the most common cause of failures in rotating machinery. Blade failure modes normally occur as a result of cracks due to unexpected operating conditions, which are normally caused by accidents of foreign objects damage, high cycle fatigue, blade rubbing, blade root looseness, and degradation from erosion and corrosion. Thus, detection of blade faults has an important role in reducing blade related failures and allowing repairs to be scheduled for the machinery. This in turn will lead to reduction in maintenance costs and thus raise productivity and safety aspects of operation. To maintain vital components of rotating machines, such as blades, shafts, bearings and gear boxes, at optimal levels, detection of failures in such components is important, because this will prevent any serious damage that could affect performance. This research study involves laboratory tests on a small rig with a bladed disc rotor that applied vibration measurements and analysis for blade fault detection. Three measurements: shaft torsional vibration, on-bearing vibration (OBV) and on-casing vibration (OCV), are used. A small test rig of a single stage bladed disc holding 8-blades was designed and manufactured, to carry out this research study to assess the usefulness and capability of each vibration technique in detection of incipient defects within machine blades. A series of tests was conducted on a test rig for three different cases of blade health conditions: (a) healthy blade(s) with mistuned effects, (b) blade root looseness and (c) cracks in a blade on two different blade sizes (long and short blades) in order to discover changes in blades' dynamic behaviour during the machine running-up operation. The data were collected using the three measurements during machine run-up and then recorded. The measured vibration data were analysed by computing the blades' resonance at different engine orders (EOs) related to the blade(s) resonance frequencies and their higher harmonics, to understand the blade(s) dynamics behaviour for the cases of healthy and faulty blade(s). Data have been further processed using a polar plot presentation method which provides clear results that can be used for monitoring blade integrity. To validate the obtained experimental results, a simplified mathematical model was also developed. Finally, a comparative study between three methods was undertaken to understand the relative advantages and limitations in the blade heath monitoring.