A decomposition approach for parameter identification in large scale networks

An efficient parameter identification method for large-scale networks based on the circuit decomposition technique is presented. The parameter identification technique has wide applications in circuit modeling, fault diagnosis, testing, and calibration. Its implementation (based on the sensitivity approach) is very useful in practice. However, it cannot handle large-scale circuits because the sensitivity matrix is dense, requiring an enormous amount of memory space to store and taking much time to compute. A method based on circuit decomposition is presented as a means of overcoming these deficiencies. The organization of this method, its basic features, and its algorithm are presented. Computer results for comparison of this method with a conventional, sensitivity-based technique are given. Advantages of the new method are summarized