Breakdown voltage modelling for leatherite paper dielectrics using fuzzy logic technique & estimating the lifetime using step-stress test

Abstract

OBJECTIVE
The real insulation systems are often heterogeneous and some times nonlinear. Quality of insulation is accessed in terms of break down strengths. Partial discharge caused in insulation system by local defects and the resultant overstressing caused by them ultimately lead to breakdown. So the estimation has to be done properly to save insulation from failure. The use of modern computers in bdv analysis has lead to the estimation based on fuzzy logic modeling. The mamdani fuzzy logic using triangular and trapezoidal mf used for the modeling. The bdv got from the modeling section is used to get the weibull parameters using MLE. The shape parameters are used for the life estimation of the dielectric.
DESCRIPTION
Fuzzy logic modeling is widely used in those fields where the boundary between having a property and not having it is not sharp. The construction of this model can be viewed as a process in which a collection of objects called variables and parameters of the model are related by some other objects called the operators of the model. In the present case it is tried to estimate the bdv of dielectrics depending upon various input conditions. The most important source of partial discharge and breakdown in dielectrics is the voids. Voids are produced due to process control errors at the time of production of most of the solid dielectrics. This is a gas discharged event. The test dielectric is taken as leatherite paper and the estimation is based on data experimentally generated in the laboratory using a CIGRE-2 electrode. The choice of test procedure to know the breakdown voltage of a typical insulation material on insulation system is determined by the test objective. Constant voltage tests provide reliable comprehensive data for the distribution function of the breakdown time but is very time consuming. An accelerated test with increase in voltage stress in discrete steps is quite often used for an electrical insulation study and is widely accepted by the insulation designers. With this method the stress at which the insulation breaks down and time to failure is taken as 6 observed variable The effect of void dimensions on the output is studied and implemented in MATLAB environment. The various steps in modeling include study of the range variation, grouping, rule list generation and simulation. Present system is a MISO system having three inputs (thickness of dielectric, depth and diameter of void) and one output (bdv). The min max algorithm is used as t-norm and s-norm operator. Coa is used for difuzzification. Programming approach is adopted for estimation. The surface plot is plotted to study the variation. Weibull probability has gained wide acceptance in the statistical treatment of time to electrical breakdown of solid dielectrics. It seems to fit experimental data well. MLE is used for parameter estimation. Confidence interval is chosen to get lower and upper limits of the parameters within which the estimation lie for a surety. Throughout the experiment the step stress test is considered. The inverse power law is applied to life estimation. From the slope of the graph the slope is to be found out and used for estimating the life.
RESULTS
In the mamdani fuzzy logic modeling using the triangular and trapezoidal mf the Mae is found out to be 1.4% and 1.324% respectively. The weibull parameters and life estimation values have close resemblance with the experimentally generated value.
CONCLUSION
Fuzzy logic provides an easier and a better computation technique based on the fuzzyness of rules. By accurately choosing the parameters and deciding the rule bases the error can be significantly reduced. The weibull parameter calculation using MLE and lifetime also found to be in good agreement. Thus the results indicates that the modeling can be well implemented for such kind of estimation.