Signals, Symptoms, Faults; Condition oriented monitoring of systems in operation
Poznań University of Technology, POLAND
Poznan, September 2000
Every system in operation is a vast the resource of information. When treated holistically, we can receive information on its production process or service (mission), its structure, design, load, and operating condition. Holistic observation of system means also, that we are analysing momentary responses of the system, as well as the evolution, where system condition will change during its life or mission. Short time response is usually used for identification purposes (model and design improvement), and for creation of symptoms of system condition. Usually for condition monitoring we create symptoms using technology of signal processing, sometimes of different physical origin. In this way we can observe condition of system in operation multidimensionally, but trouble is that current condition monitoring technology is in essence one dimensional. This means we do observe many symptoms, but inference on system condition concerns every symptom separately. Typical example comes from vibration condition monitoring of turbosets, where vibration limits of every bearing in every direction is separately limited, and there is no combined measure of turboset condition.
The paper in hand try to present some proposal of synergetic treatment of multidimensional condition observation. In such observation of system condition, there is much redundancy. But when we apply Singular Value Decomposition (SVD) to symptom observation matrix (non quadratic matrix with columns as symptom readings), we can reduce significantly this redundancy, and pass to life dependent information, carrying mainly system condition. In this way we can define system generalised faults, its measures and indices, and combined measures and indices of system wear or modification. This can help us much in reducing risk of system operation, and undertake right decision concerning future system operation. This entire line of reasoning is better understood with the help of enclosed Figure 1, (p.t.o.).