Abstract
〈Vol.10 No.4(2017.7)〉

Titles
[CONTRIBUTED PAPERS]


■ Analysis of Active Sounds for Fire Localization

Hirosaki University・Yasushi IWATANI,
Shoji KASAI, and Hiroyuki TORIKAI

Localization of fire in smoke environments enables fire-fighters to perform fire-fighting operations safely in a short time, and fire localization is also required to develop autonomous fire-fighting robots. This paper provides a sound property available for fire localization by investigating sound transmission through flame-air interfaces, sound transmission across a gas flow whose gas density is different from the air, and sound transmission in smoke. Sound analysis is performed for several sound frequencies, heat release rates and concentrations of smoke. It is shown that sound through flame fluctuates, and the fluctuation range does not depend on the sound frequency, the heat release rate, or the combustion process. Sound through flame can be distinguished from sound through gas-flow or smoke. The sound property shown in this paper informs the existence of flame in the sound path.


■ Analysis of Heat Transfer in Thermocouples Immersed in High Temperature Gas with Short Insertion Length

Keio University・Satoshi HONDA and
Kawasaki Steel Corp.,・Tadaaki IWAMURA

Thermocouples are used to measure high temperature in process and steel industries. To measure high tem- perature gas, thermocouples should be inserted deeply enough;commercial specification documents recommend at least 5 times the diameter of sheath or more. This paper analyzes the effect of insertion length on thermocouple outputs in high temperature gas measurement. The theoretical evaluation is revised to include the radiation effect in the authors’previous work presented in SICE AC 2015 and compared with experimental data in a heating furnace more precisely than the work presented in SICE AC 2016.


■ A System Identification Based Approach for Design of IIR Digital Filters with Robust Stability

Hiroshima University・Masayoshi NAKAMOTO,
Tomohiro HIRAKAWA, and Toru YAMAMOTO

Design problem of infinite impulse response (IIR) filters is generally a non-linear optimization problem due to the presence of denominator polynomial. Additionally, the stability condition (position of poles) must be considered when optimizing the filter coefficients. Hence, an iterative optimization is usually required to solve the design problem for stable IIR filter. In this paper, we present a new method for the design of IIR filters without iterative optimization. We employ a system identification method for time series signal where the input signal and its ideal output signal are generated by a Gaussian stochastic process with a prescribed frequency characteristic. Then, based on the Parseval's theorem, we can obtain the IIR filter in the frequency domain. The advantage of the proposed method is to compute the IIR stable digital filters as a closed-form solution. That is, we can approximate the given frequency response and the constant group delay without using any iterative optimization. Also, we present a design method with specified maximum pole radius to achieve robust stability. Finally, design examples are presented to illustrate the effectiveness of the proposed method by designing a high-pass and low-pass IIR digital filter.


■ Estimation of Marbling Score in Live Beef Cattle Using Bayesian Network

Saga Univesity・Osamu FUKUDA,Iqbal AHMED,and Nagasaki Agricultural
and Forestry Technical Development Center・Daisuke HASHIMOTO

To estimate more accurately the Beef Marbling Score (BMS) of live beef cattle, the Bayesian Network Model (BNM) could be used in parallel with other developed methods, such as Ultra sound (US) image analysis with Neural Network (NN), Biological Impedance Analysis (BIA) and visual inspections of an experienced inspector. Additionally, most of these methods individually represents positive trends of estimating subjective BMS in Japan. This research reveals that the approach of using BNM to include body condition parameters, exhibit more accurate estimation of BMS with other methods. The measurement was conducted with 28 Japanese Black Beef cattle before one-month slaughter. The weight, chest, abdominal circumference, and longissimus muscle area have been taken into consideration of body measurement parameters for evaluating BMS. The estimation of BMS with BNM, combined with other approaches displayed the higher accuracy rate of almost 90%. Moreover, this research compared the findings with other individual method and combined methods. The estimation of BMS using US image analysis using NN represents 28% accuracy, then BIA provides only 40%, and combing both US and BIA method illustrates 50% of accurate BMS estimation. However, Body condition indices, US and BIA together outreaches all estimation methods and the BNM provided more accurate estimation of BMS with high confidence.


■ A Randomized Algorithm for Chance Constrained Optimal Power Flow with Renewables

Osaka University・Takayuki WADA, Gifu University・Ryosuke MORITA,
Nagoya University・Toru ASAI,Kobe University・Izumi MASUBUCHI,
and Osaka University・Yasumasa FUJISAKI

A chance constrained AC optimal power flow is to find the optimal economic operation plan whose probability satisfying AC power flow equations and various inequality constraints on operating limits of the power system is greater than a specified probability level. Even if a constraint condition for each uncertain power supply value is convex with respect to decision variables, the chance constrained problem is not convex in general.  Thus, it is difficult to solve the problem within reasonable computational time.  Employment of randomization techniques for this issue is proposed in this paper. A main advantage of the framework leads to a solution with a theoretical guarantee. Its sample complexities are of polynomial order for parameters of a given accuracy. The efficiency of this algorithm is demonstrated by applying it to the Japanese power system models.


■ Attitude Control of Two-Wheel Spacecraft Based on Dynamics Model via Hierarchical Linearization

Panasonic Factory Solutions Co.,Ltd.・Tomohiro FUKAISHI,
Tokyo City University・Kazuma SEKIGUCHI, and Kenichiro NONAKA

In this paper, we propose an attitude control law for underactuated two-wheel spacecraft under non-zero total angular momentum. Attitude control with non-zero total angular momentum is complicated in the case that the number of reaction wheels equipped on a spacecraft is two. For a spacecraft in this situation, an attitude control law has been proposed based on a kinematics model in [7]. However, a dynamics controller is more desirable for a practical system. Thus, in this paper, we expand the controller to a dynamics model. Nevertheless, in the case of dynamics model, the expansion is not straightforward because of the singularity of input transformation. Therefore, we propose to apply the hierarchical linearization technique which separates a system into several subsystems and linearizes the subsystems step by step. Using this method, the input transformation becomes well-defined, and the system is linearized partially. Additionally, the dimension of linearizable state increases compared to the ordinary input-output linearization. Numerical simulation is conducted to show the validity of the proposed controller.


■ Cutting a Parameter Space for a Multi-Legged Robot Based on Model Checking

Nagoya University・Keisuke NOMURA,Shinkichi INAGAKI

This paper proposes a technique to cut a parameter space in dynamic simulations based on model checking. The design object addressed in this paper is a multi-legged robot which is controlled by a Follow-the-Contact-Point (FCP) gait control. The design parameters are about both the walk controller and the robot construction, but the parameter space is too wide to evaluate by the dynamic simulation. The robot behavior is modelled by timed-automata, and specifications for the robot to successfully walk are described in a computational tree logic (CTL). The proposed method is based on repetition of applying model checking, UPPAAL, to each set of parameters. Its availability is shown by adopting the proposed technique to designing a multi-legged robot.