ABSTRACT
In this chapter a “non-optimal” method of block pulse function coefficient computation has been presented which employs trapezoidal integration instead of exact integration. The properties of the “non-optimal” block pulse function set have been discussed briefly. This apparently “non-optimal” procedure has been applied for control system analysis and identification, mainly because the trapezoidal integration uses only samples of the function to be expanded via BPF and thus reduces computational burden drastically. The results of such analysis, when compared to the results obtained via traditional BPF approach, are found to contain less error and also, results of identification are found to be much more superior to conventional BPF based analysis. It is also found that the results of identification via BPF are highly erroneous and contain undesired oscillations in many cases. Three systems of two different orders have been studied with two standard inputs each to provide strong support for the “non-optimal” approach. Also, an error estimate has been presented for function approximation via this new approach and the results are compared with estimated error via conventional BPF analysis. At the end of this chapter, seven study problems are included for the benefit of the readers.
