Recipient size estimation for induction heating home appliances based on artificial neural networks

Abstract

In induction home appliances, recipient size estimation is very important for the adjustment of the power that the hob must supply to the cooking recipient. Conventional techniques first calculate a simple R-L equivalent circuit from voltage and current waveforms in the heating inductor and then estimate physical parameters (such as recipient size) by regression. In this paper, a new technique is proposed for recipient size estimation, based on spectral analysis and artificial neural networks (ANN), which, for two reasons, is more accurate than current procedures: (i) The new technique performs a direct estimation of recipient size from voltage and current, without needing to compute an intermediate electrical equivalent circuit (which in fact only represents a rough approximation), and (ii) due to their nonlinear modeling capabilities, ANNs are more appropriate than regression for this problem. By using a database of cooking recipients, our procedure provides an accuracy of 85-90 %, outperforming the 58 % of conventional techniques (70 % including an additional sensor). The new technique has been implemented and verified by using a commercial embedded processor, similar to those included in current domestic induction home appliances. It could be built in by manufacturers at no extra cost as it requires no additional sensors and makes use of computational resources integrated into the microcontroller or the digital signal processor that controls the home appliance.

Load Estimation for Induction Heating Cookers Based on Series RLC Natural Resonant Current

“…Another of these studies [21] used the voltage, current waveform, and related period under different switching intervals to calculate equivalent parameters. [22] detected equivalent parameters by using the power factor and the absolute amplitude of the impedance |Z|. As IHCs based on quasi-resonant converters have the shortage of high-voltage stress on the power transistors [23], many studies [24][25][26][27][28][29][30] have focused on estimation methods for HBSRCs; specifically, these studies have proposed estimation methods based on single values or the average value of feedback electrical signals within a bus voltage cycle [24][25][26][27].…”

“…Several studies have focused on developing online estimation methods for reducing the effects of variations in R eq and L eq on power control due to cookware movement or removal from the heating coil during heating. For example, studies [20][21][22] have proposed online estimation methods for quasi-resonant converters. Specifically, one of these studies [20] proposed four estimation methods for measuring the V CE of an IGBT under constant current control, measuring the turn-on current I C of the IGBT under constant voltage control, comparing times at which V CE is greater than the DC link voltage, and measuring the diode turn-on time.…”

Convergence in Orlicz spaces by means of the multivariate max-product neural network operators of the Kantorovich type and applications

“…The study of neural network type operators has been recently investigated by many authors from the theoretical point of view, see e.g., [14,[16][17][18]46].…”

Approximation Results in Orlicz Spaces for Sequences of Kantorovich Max-Product Neural Network Operators

“…The theory of the NN operators has been widely studied in last decade in view of its relation with the well-known theory of the feed-forward artificial neural networks, see [1,[12][13][14]. Neural networks have been widely used in many applications fields: such as in artificial intelligence ( [11,[31][32][33]41]), neuro sciences and in approximation theory ( [3,35,40,42]).…”