Measurement error compensation using metacognitive elm based artificial neural network

Measurement of a physical parameter can adversely be affected by several factors like manufacturing process variations, environmental conditions etc. The measured output may non-linearly vary with many of these factors and may cause error in the parameter being measured. A precise modeling of the input-output relationship of measurement systems is indeed a necessity in order to have an accurate measurement. This paper proposes an innovative software error correction method, based on McELM learned Artificial Neural Network (ANN). The network is trained using an advanced fast learning algorithm called Extreme leaning Machine (ELM). The performance of the network on a typical dataset indicate encouraging results with a mean square error better than 1e
^-04
. Further improvement of the algorithm in terms of its capability to efficiently handle online data has also been demonstrated. This has been achieved by building intelligence into ELM learning, using a Metacognitive approach.