Formation of an explainable intelligent model based on a hierarchical fuzzy inference system for analyzing the condition of runway sections

Abstract

As part of the research, an explainable intelligent model is proposed that generates assessments of the condition of runway sections and classifies them by condition. The provisions of the fuzzy set theory were chosen as a basis for the model, which allow the implementation of fuzzy inference systems (FIS) to generate assessments of objects of analysis under conditions when the parameters characterizing the state of an object have different units of measurement and the level of influence on the final result. The capabilities of the FIS allow distributing elements of the analyzed set of objects into condition classes. An important property of the FIS is the ability to explain the reasons for forming a certain assessment. The suitability characteristics stipulated by the Federal Aviation Rules (FAR-262) were used as a set of parameters used to assess the runway section. The proposed model aggregates eleven parameters. Given the number of parameters, the model itself is implemented as a cascade hierarchical FIS. As a result of the experiments, it was shown that the model has a high sensitivity to changes in the values of the estimated parameters in the range of «rather bad», «bad» or «emergency» values. The formation of a «low» final assessment and the correlation of the object of analysis with «bad» classes of state occurs, among other things, when a «bad» value is manifested in at least one of the parameters taken into account. This property is valuable because FAR-262 provides for making a decision on the unsuitability of the runway even if at least one of the parameters has a value beyond the limits of suitability. The model also coped with the distribution of objects of analysis by classes of state in cases where tuples with different combinations of values of the parameters taken into account were fed to the inputs. Arrays of intermediate values made it possible to consistently identify the variables that influenced the formation of a low output assessment. The ability to adjust various values of input parameters made it possible to determine possible options for control actions to transfer the studied object of analysis to a «higher» class of state.