Wyniki wyszukiwania - BazTech

1

Metody doboru modeli prognostycznych w automatycznych systemach diagnostycznych

Tabaszewski M.

Diagnostyka

|

2004

|

Vol. 30, T. 2

125--128

PL

Praca zawiera próbę rozwiązania problemu wyboru optymalnego modelu prognostycznego w automatycznych (skomputeryzowanych) systemach nadzoru diagnostycznego. Wraz z napływem nowych danych pomiarowych założony wcześniej model prognostyczny może się dezaktualizować co wymaga uruchomienia mechanizmu generacji (wyboru) nowego modelu prognostycznego bez wymaganej wiedzy eksperta w oparciu tylko o zgromadzone wcześniej dane. W pracy przedstawiono wyniki różnych proponowanych metod zastosowanych w celu przeciwdziałania dezaktualizacji modelu prognostycznego.

EN

The paper contains an attempt of solution of optimal choice of forecasting model in automatic condition monitoring systems. As the new measured data are gathered an assumed earlier forecasting model can became inadequate. It requires to start a mechanism of generating or choosing a new forecasting model without expert knowledge, based only on some gathered earlier data. The paper presents results of the various methods used for avoidance of loss of timelines of forecasting models.

2

Reduction of Redundancy of Symptom Observation Space in Systems Condition Monitoring

Cempel C.

Bulletin of the Polish Academy of Sciences. Technical Sciences

|

2001

|

vol. 49, nr 2

267-284

EN

It was shown in this paper that classical approach to the assessment of systems condition evolution can be much improved by special processing of observed symptoms of condition. When we have a large symptom data base, we can apply singular value decomposition (SVD) as the newest data mining procedure, to obtain a symptom and condition evolution model. By using SVD it is possible to have two additional independent fault discriminants: named here SD and SG, with high dynamics of evolution during system life. Moreover, we can add an additional column of system life count, as the first approximation of a logistic vector describing the unit life history. It is also possible to use the value of a pseudo - determinant of a symptom observation matrix, and correlation between this new discriminant and the symptom observation matrix to minimize the redundancy of symptom measuring space, and choose the best symptom for condition observation.

3

Modelling and data mining in vibration condition monitoring

Cempel C.

Prace Naukowe Instytutu Górnictwa Politechniki Wrocławskiej. Konferencje

|

1999

|

Vol. 86, nr 24

33-46

EN

It was shown in this paper that classical approach to systems condition evolution assessment can be much improved by special processing of observed symptoms of condition. When we have a large symptom data base, we can apply singular value decomposition (SVD), as the newest data mining procedure to obtain a symptom and condition evolution model. By using SVD it is possible to have two additional independent fault discriminants: named CD and SG, with high dynamics of evolution. Moreover we can an additional column of system life count, as the first approximation of a logistic vector describing the unit life history. It is also possible to use the value of a pseudo - determinant of a symptom observation matrix, and correlation between this new discriminant and the symptom observation matrix to minimize the redundancy measuring space, and chose the best symptom for condition observation.