This study aims to construct a deterioration diagnosis support system for coating of steel structures. This system is comprised of two main image-processing components (Phase I and II). Phase I is a component to detect deterioration spots of coating, based on the core-extraction clustering method with SOM. Phase II is a component to evaluate deterioration level of coating, whose input date is generated by Phase I, i.e., Phase I and II are sequentially connected. Estimation provided by Phase II is represented by fuzzy sets such as (Level A, B, C) = (0.0, 0.6, 0.3)
Free-hand drawing of graphical elements using the mouse is one of most user-friendly methods. In non-visual communication, graphical elements can be easily expressed using tactile communication tools; however, these methods are not competent in expressing properties and relations of graphical contents in elementary geometry. Graphical elements are most optimally recognized using inputted hand-written curves. Corner detection is a complex technique, which serves an important role in these recognitions. We describe recognition aspects in inputted graphical contents of elementary geometry, and improve the break-bend judgment using Choquet's integral like evaluation functions and generic algorithms. Our learning method and the corresponding normalization method are made universal, and can be applied to other problems.
In manufacturing, both productivity and energy-saving have been required for sustainable economic progress. This article proposes an operational condition estimation method for improving energy efficiency by monitoring energy consumption in production machines. Our proposed estimation method aims to analyze causalities among the productivity performance indices of QCDES and the factors of 4M + 2E for improving manufacturing process. They stand for Quality, Cost, Delivery, Ecology, Safety, Man, Machine, Material, Method, Environment, and Energy. In our proposed method, operational condition such as stop, idling, and tact delay, is estimated from the machine behavior that is extracted by the energy consumption, and the energy loss is quantified by the operational condition. Machine behavior can be extracted by energy consumption if the energy consumption is measured with short sampling time less than 1.0 second. According to the estimation results of the operational condition on our factory line, it was successfully to improve an operating rate and energy efficiency by recognizing energy loss, which means energy consumption of non-contributive to the productivity. The production volume was increased and energy consumption was reduced concurrently because the operators could easy to select improvement plans, because the energy loss is sorted to a cause.
Recently, more people try to perform social survey or marketing by analyzing data on social media. Such attempts are called “Social Media Mining”. Attributes of social media users are not specified. Some attributes of social media users are easy to be estimated. Others are not easy to be estimated. In previous researches, they can succeed in estimating age, gender and habitants of social media users. While, it is difficult to estimate occupation among user attributes of social media. The proposed method use support vector machine to create classifier for occupation estimation. It succeed in estimating occupation with practical accuracy of 0.85 and practical recall of 0.77.
This study aims to develop a quality evaluation system by analyzing the skills of experienced puffer cooks from well-established intermediary wholesalers in order to create a model for quality evaluation. A total of 560 fish dressed by the cooks (Migaki) were graded into five classes to determine what aspects of the surface color and meat freshness of the fish the cooks use for evaluation. Analysis of the appearance evaluation by cooks, combined with information on fish coloration and meat freshness, indicated that the cooks focus on four colors on the surface of the fish. Fuzzy inference models were then constructed and evaluated using the colors as four antecedent-part variables, and results corresponded with the assessments by cooks with more than 80% certainty. The results confirm the validity of the proposed method, which incorporates the knowledge of skilled cooks in the design of the Migaki quality evaluation system.
By analyzing participants' interactive behavior, we demonstrate how everyday table talk conversations are structured. Unlike previous works, which have focused on the utterances and eating behavior of speakers, this study analyzes the behaviors of the hearers. Three groups of triadic female conversations during a meal were videotaped and transcribed, and the relationship between hearers' utterances and eating actions were analyzed. An analysis of the hearers' 869 responsive utterances and 343 eating actions reveals that when hearers are highly involved (e.g., being directly addressed) in conversations, their eating actions generally occur just after their responses. When their involvement is lower, their eating actions often occur after the others' responses and sometimes they eat without their own or others' responses. When speakers are silent, hearers tend to eat their food and may wait until the current speaker's next utterance. We discuss that table talk has a structure in which hearers adjust their responses and eating actions to the level that is necessary to maintain cooperative conversations.
This material aims to discuss the educational characteristics of fuzzy (Sugeno) integral which facilitates as a method dividing a classroom into groups. The model of fuzzy integral here is a criterion for assessing “the goodness of the group” based on the evaluated result of the relevant tasks offered to students. We here show that the criterion is an appropriate model in terms of taking into consideration both to the collective value as a student's group benefit and the individual value as a student's benefit.