In assembly tasks, contact relationships between parts play an important role in cancellation of uncertainties. This paper describes a method of robot-teaching for assembly, especially mating of parts, by using object-oriented analysis and design. It explicitly uses an assembly strategy which is based on the state transition of contact relationships. The first part of this paper models the assembly strategy, where the model consists of 4 components; state, transitional-task, verifier and supervisor. The second part introduces object oriented analysis and a design technique for easy implementation of the model. The last part explains how to program and teach mating tasks of polyhedral blocks in a real system. Our method has several advantages : (1) Uncertainties on positions and orientations can be canceled when a robot plays back to instructed motions. (2) Modifications of source code in robot software can be localized in case of the addition of new mating functions because of object-oriented analysis and design of the software.
Data Envelopment Analysis (DEA) is a nonparametric technique for measuring and evaluating the relative efficiency of a set of decision making units using observational input and output data. A fundamental DEA model is formulated as linear programming and the input and output data are assumed to be quantitative positive values. However, in the practical application, there are some uncertainty in the input and output data such as an observational disturbance and subjective data. In this paper, we consider ambiguous data expressed as a fuzzy categorical variable and propose a DEA model for a noncontrollable fuzzy categorical input variable. The proposed model gives a reasonable efficiency score, which compensates for the lack of information caused from the discrete categorization, and has robustness against the change of the boundaries and the fuzziness of categorization.
This paper proposes an idea for determining the inspection order when a new sample is classified by a neural-network-based classification system. In real world classification problems such as medical diagnoses, inspection costs for measuring many inspection items can not be negligible.Therefore, it is useful to classify a new sample by measuring a small number of inspection items. In this paper, first we propose a method for classifying a new sample by partial information on its attribute values in a neural-network-based classification system. The proposed method is based on the interval representation of incomplete (i.e., partial) input values. Next we propose an idea for determining the inspection order of input values for a new sample. Last, we illustrate the proposed approach by computer simulations on a numerical example and the iris data.
Operability Study is a systematic technique for identifying hazards or operability problems throughout an entire facility. In this paper, we have proposed an approach using knowledge engineering techniques to the automated Operability Study. The computer-aided Operability Study System consists of the plant specific knowledge base, the generic data base and the inference engine. Causal relationships between input variable deviations and output variable deviations for components are modeled using decision tables. Decision tables for components are developed by the user and stored in generic data base in computers. The plant structure (Piping and Instrumentation diagram) and reaction types are inputted to the plant specific knowledge base in computers. Each process variable of equipment is examined in sequence by searching the generic data base, and Operability Study are generated resulting from the search. We demonstrate via solvey process how computer-aided Operability Study can identify hazards, and substantiate usefulness of the method.
This paper describes a practical method of a new type of an acoustic receiver using the co-sensor with no vibration sensor in order to decrease the influence of the equi-status noise (NMN : normal mode noise). The co-sensor is physically arranged close to the acoustic sensor as a counter measure for the noise circumstances. The acoustic signal is only received by the acoustic sensor and NMN is separately detected by both the acoustic sensor and the co-sensor. Then the recieved noise can be eliminated or canceled without the phase processing. From the experimental results using this method, SNR (signal noise ratio) increases around 35dB in the air 13dB under the water respectively.