This paper presents a novel method of cavity filter tuning with the usage of an artificial neural network (ANN). The proposed method does not require information on the filter topology, and the filter is treated as a black box. In order to illustrate the concept, a feed-forward, multi-layer, non-linear artificial neural network with back propagation is applied. The method for preparing, learning and testing vectors consisting of sampled detuned scattering characteristics and corresponding tuning screw deviations is proposed. To collect the training vectors, the machine, an intelligent automatic filter tuning tool integrated with a vector network analyzer, has been built. The ANN was trained on the basis of samples obtained from a properly tuned filter. It has been proved that the usage of multidimensional approximation ability of an ANN makes it possible to map the characteristic of a detuned filter reflection in individual screw errors. Finally, after the ANN learning process, the tuning experiment on 6 and 11-cavity filters has been preformed, proving a very high efficiency of the presented method.
Abstract-This paper presents a new method of sequential microwave filter tuning.For filters with R tuning elements (including cavities, couplings and cross-couplings), based on physically measured scattering characteristics in the frequency domain, the Artificial Neural Network (ANN) is used to build inverse models of R sub-filters. Each sub-filter is associated to one tuning element. The sub-filters are obtained by successive opening or shorting of resonators and by removing coupling screws. For each sub-filter, the ANN training vectors are defined as physical reflection characteristics (input vectors) and the corresponding positions of the tuning element, which is detuned, in both directions, from its proper setting (output vectors). In the tuning process, such inverse models are used for calculating the tuning element increments needed for setting the tuning element in the proper position. The tuning experiment, conducted on 8-and 11-cavity filters, has shown the performance of the presented method.
Abstract-This work presents a novel approach in building a multidimensional approximator which is used as a linear operator for mapping the vector of detuned filter characteristic to the vector of deviations of tuning elements. This has been done for the purpose of using it in postproduction filter tuning algorithm. With the use of collected sets of deviations of tuning elements and filter characteristics corresponding to them, the least squares method (LSM) is applied to determine the matrix which realizes the linear mapping between these vectors. The matrix found in this method approximates the vectors of both spaces (filter characteristics and corresponding deviations of tuning elements). In tuning process this matrix is used to determine the vector of tuning element deviations for a given detuned filter characteristic read from Vector Network Analyzer. To increase the "quality" of linear operator filter characteristics are transformed with the use of Karhunen-Loeve transform (Principal Component Analysis). In contrast to non-linear artificial intelligence approximators used in filter tuning and published to-date, this method does not require a time-consuming training process. Filter tuning experiments have been performed and proved the correctness of the presented approach.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.