Hidenori Maruta

A new digital control DC-DC converter with neural network predictor

The purpose of this paper is to present a new digital control method of a forward type multiple-output dc-dc converter with both a P-I-D feedback and a new feed forward control. In this converter, a new method with a neural network predictor is presented. The dynamic characteristics of digital control dc-dc converter are improved as compared with the conventional one. Especially, the digital control dc-dc converter with the neural network predictor can realize excellent dynamic characteristics. As a result, the undershoot voltage, the overshoot voltage and the convergence time of the output voltage are improved to 43%, 65% and 63% respectively.

Smoke detection in open areas using its texture features and time series properties

In extensive facilities such as port facilities, chemical plants, and power stations, it is important to detect a fire early and certainly. The purpose of this paper is to present a new smoke detection method in open areas, as smoke is considered as a significant signal of the fire. It is assumed that the camera monitoring the scene of the open area is stationary. Since smoke does not keep stationary shape or image features like edges, it is difficult apply ordinal image processing techniques such as the edge or contour detection directly. In this paper, we propose a novel method of the smoke detection in an image sequence, in which we combines the several images techniques to detect smoke. We apply it to images of open areas under general environmental conditions. First, moving objects are detected from gray.scale image sequences, and then the noise is removed with the image binarization and the morphological operation. Furthermore, since the smoke pattern must be examined, the smoke feature is extracted with the texture analysis. Then, to obtain the final result of the proposed method, we discussed the properties of the proposed features as the time series data.

A new approach for smoke detection with texture analysis and support vector machine

Early and certain fire detection is one of the important issues to keep safe infrastructures. Especially, it becomes an urgent problem in large facilities like port facilities, large factories and power plants, due to its large harmful effect to surrounding areas. In these places, to detect the fire or flame directly is have some difficulties because they are open and hence have problem to set sensor devices. So smoke is an important and useful sign to detect the fire robustly even in such open areas. In this study, we propose a novel and robust smoke detection method based on image information. Smoke is consider a moving objects, so firstly we extract moving objects of input images as smoke candidate regions. Because smoke has a characteristic pattern as image information, we treat smoke patterns as textures. Here we use texture analysis to extract feature vectors of images. To classify extracted moving objects are smoke or non-smoke, we use support vector machine(SVM) with texture features as input features. Extraction of moving objects are sometimes easily affected by environmental conditions. So we accumulate the result of the classification with SVM about time to obtain accurate extraction results of smoke regions.

Image based smoke detection with local Hurst exponent

Smoke is an important sign for early fire detection. Image based detection methods are more useful than other methods which use some special sensor devices. When treating image information of smoke, it is important to consider characteristics of smoke. In this study, we consider that the image information of smoke is a self-affine fractal. We focus on the nature of smoke and present a new smoke detection method based on the fractal property of smoke. We use the Hurst exponent H, which is one of the widely known exponent of fractals. We calculate H of smoke from a relation between H and the wavelet transform of the image. So we detect smoke areas in images with H through the wavelet transform. Moreover, to obtain the accurate detection result, we use the time-accumulation technique to smoke detection results of each image. In experiments, we show the effectiveness of our method with the fractal property of smoke.

Reference modification control DC-DC converter with neural network predictor

The purpose of this paper is to present a new digital control method for dc-dc converters by reference modification with the neural network predictor. In the proposed method, the reference in the proportional control term of the conventional PID control is modified using the neural network predictor during the transient interval. The neural network is repeatedly trained to predict the output voltage using former predicted data for the modification of the reference. After the training, the reference in the P control is modified by the predictor to improve the transient response. By using the proposed method, the undershoot of output voltage is suppressed to 41% compared with the conventional methods one. The convergence time is also improved to 48% compared with the conventional methods one. Therefore, it is confirmed that the proposed method has the superior performance to control dc-dc converters.

A New Control Method for dc-dc Converter by Neural Network Predictor with Repetitive Training

This paper proposes a novel prediction based digital control dc-dc converter. In this method, a neural network control is adopted to improve the transient response in coordination with a conventional P-I-D control. The prediction based control term is consists of predicted data which are obtained from repetitive training of the neural network. This works to improve the transient response very effectively when the load is changed quickly. As a result, the undershoot of the output voltage and the overshoot of the reactor current are suppressed effectively as compared with the conventional one in the step change of load resistance. The proposed method is based on the neural network learning, it is expected that the proposed approach has high availability in providing the easy way for the design of circuit system since there is no need to change the algorithm. The adequate availability of the proposed method is also confirmed by the experiment in which P-I-D control parameters of the circuit are set to non-optimal ones and the proposed method is used in the same manner.

Digital peak current mode DC-DC converter for data center in HVDC system

The purpose of this paper is to present the digital peak current mode phase-shifted full-bridge zero volt switching dc-dc converter for the data center in the higher voltage direct current (HVDC) system, and show its transient response. The proposed peak current mode control circuit is composed of the RC integrator and comparator only and can detect the peak current in real time. At first, the basic operation of proposed method is confirmed by using the buck type dc-dc converter. Furthermore, the proposed method is applied to the prototype of 380V phase-shifted full-bridge zero volt switching dc-dc converter. The system has to be always stable in the transient response even if the input voltage is changed. As a result, the convergence time of output voltage is improved by about 50% compared with the conventional digital voltage mode control when the input voltage is 380V and 300V.

Anisotropic LBP descriptors for robust smoke detection

Image based smoke detection is a difficult problem especially in open areas since it is heavily affected from its environmental objects. This comes from the transparent property of smoke itself. Therefore, to realize robust smoke detection in such situation, it needs to take into account the effect of the degree of transparency, the change of background objects and so forth. To describe smoke information by image features, they are affected from degree of transparency of smoke, background objects, and other environmental conditions such as the direction and the speed of wind. To address such problems, we apply a novel image feature named anisotropic LBP descriptors, which is considered as a extended variants of LBP. The anisotropic LBP descriptors are simply extended from LBP, which are defined as texture operator using anisotropic neighborhood pixel values. Therefore, they can describe anisotropic deformed image information, which are caused from environmental conditions. To obtain more accurate detection results, we also adopt AdaBoost which uses anisotropic LBP descriptors as input vectors. In this study, each AdaBoost classifier is trained for every anisotropic LBP descriptor and the detection result is obtained from the combined result of those AdaBoost classifiers. We evaluate our presented method and confirm that the our approach works well to obtain robust results.

A new fast digital PID control dc-dc converter

This paper presents a new fast response digital PID control circuit for dc-dc converter, its dynamic characteristics and the design criterion. In the conventional digital PID control method, the sampling rate is single and the calculation process of P-I-D control is series. On the other hand, in the proposed method, the sampling rate is divided each sampling period. The calculation process of P and I-D controls is parallel. The sampling interval and points for I-D control are same to the conventional method. However, the sampling point of P control is oversampling during the switching period of the converter. The simulation and experiment results of the output voltage against the step change of the load are discussed. As a result, it is revealed experimentally that the fast transient response is obtained. The convergence time of output voltage of proposed method is less than 1/3 of that of conventional one when the output smoothing capacitance becomes small.

A novel timing control method for neural network based digitally controlled DC-DC converter

Generally, the training based control method for dc-dc converters has a problem due to the fact that there is a difference of behavior between a controlled system before the training and one after the training. Therefore, when it is adopted to improve the transient response of dc-dc converter, it is needed to consider the bad effect of the training since the training term does not take its own effect to the transient response into account. Especially, the suitable timing and duration of the training based control term is affected since the behavior of the system is changed by its own effect. In this paper, we study a timing control method for a neural network based digital control method to improve the transient response of dc-dc converters. The neural network control is a suitable training based method since it can be a time series predictor to compensate the transient response. In our presented method, the standard three-layer neural network is adopted to improve the transient response converters by reference modification of a conventional PID control. To address the problem about the timing control, we present a method to obtain the suitable timing and duration effect of the neural network control term with simple criteria. This timing control works to avoid the bad effect of the neural network control term and obtain improved results. Experimental results show that our method can contribute to the improvement of transient response effectively.