We present a data-driven anomaly detection system based on representation learning for normal and faulty classification. The proposed system extracts bottle-neck features, which contribute to discriminating the healthy and faulty states of machinery, through a deep neural network, and then detects an anomaly by a verifier constructed on the bottle-neck feature space. Experimental comparisons conducted using vibration data observed at wind turbines demonstrated that the bottle-neck features yielded improvements in anomaly detection over the conventional feature.

Diagnosing the complex system accurately based on stochastic method requires an enormous amount of data, both with and without faults. Acquiring operation data with all kinds of faults for each components is very hard and costly. To generate data for rotating machinery diagnosis, we have developed rotating machinery library using Modelica. It provides the basic components such as rotor, shaft, bearing, coupling, housing and support. Its component models are implemented on the basis of the transfer matrix method in rotor dynamics. To validate our models, we compared both Modelica simulation and experiment with a rotor kit as a test case.

In this paper, we discussed the thinning pipe wall detection using coupled oscillation between cylindrical shell and attached sub-strcture. At first, the resonance frequencies of the in-plane bending deformation mode are compared between numerical and analytical model. Thereby, it is confirmed that the resonance frequency is changed at the pipe wall thinning by both models. The dependency of sensitivity of the resonance frequency to the mass of the attached sub-structure is evaluated.

In this research, we propose a method based on "frequency down-conversion" phenomenon in which two simultaneously applied ultrasonic waves with different frequency are mixed at the contact-type damage and the component with the differential frequency emerges. In the proposed method, it has been confirmed that the magnitude of the difference frequency component excited changes depending on the instantaneous frequency, the difference frequency and the size of the damaged part by changing the input wave as two chirp waves whose frequency varies with time. In this paper, in order to clarify how the magnitude of the difference frequency component is determined, a model explaining the emergence of the difference frequency component was formulated assuming a quadratic nonlinearity at the contacting surfaces in the damage part. The derived solution using a perturbation method suggested a proportional dependence of the magnitude of the difference frequency component to the product of the amplitudes of the two input waves, which was then confirmed by an experiment in the case the input frequencies were time-invariant.

In this paper, we discussed the vibration analysis of cylindrical shell with sub-structure using semi-analytical receptance method. At first, semi-analytical receptance method was introduced the generally vibratory system. In addition, the eigenvalue problem of cylindrical shell with sub-structure was solved by the proposed method. Moreover, the investigation was conducted the comparison between the result of proposed method and theoretical value of ring model.

In-plane bending vibration mode is a good candidate for pipe inspection. First, using finite element method the characteristics of in-plane bending mode are investigated. It is confirmed that in-plane bending mode has small dependence on the pipe length and the boundary conditions at the pipe ends, it appears in the audio frequency, and it has linear dependence on the pipe thickness. In addition, formula with two dimensional ring approximation treating as in-plane bending of curved beam is derived for parameter identification. Moreover, in-plane bending mode is observed experimentally on the actual pipe system and its resonant frequency shows good match with the ring approximation. For application to actual pipe inspection, the ring approximation is compared with finite element model which has distributed deteriorated regions.

Due to the rapid development of industrial wireless communication technology with the spread of IoT technology, it is becoming possible to detect abnormality prediction by collecting the state of facility equipment in factories etc in real time. In this paper, we verified the 920MHz band wireless sensing system, which has superiority in the factory where there are many structures that can hinder the radio wave propagation in our factory. As a result, the sensing system was possible to remotely monitor the state of many manufacturing equipment, and it was confirmed that abnormality prediction can be detected.

Fault tolerance control is required on plants for productivity and economy. The use of a threshold is proposed as a fault detection method. In the conventional research, a variable threshold is utilized. Noise countermeasures were taken with a variable threshold. Early fault detection is a trade-off with noise countermeasures. Therefore, early fault detection was difficult. In this research, using Volterra kernel and SVM, a method is proposed to improve the above problem. The effectiveness of the proposed design scheme is verified by simulations.

Using an engine oil adding soft matter and a rice bran oil refined from rice bran, high pressure density and high pressure viscosity tests have been carried out, and high pressure physical properties of the test oils have been obtained. As a result, it was found that the engine oil M16A adding soft matter with molecular aggregation property shows the almost same pressure viscosity coefficient as PAO40 with high viscosity. The same tendency between the rice bran oil and a comparative mineral oil with close viscosity grade to the rice bran oil was confirmed. In addition, entrapped EHL oil film formation and its variations over time under impact loading condition have been examined. In both of M16A and PAO40, decreasing process of the entrapped EHL oil films caused by liquid behavior of the oil film was observed, and the oil film of M16A disappeared quickly compared with PAO40 having large lubrication parameter αη₀. In the case of the rice bran oil, the entrapped oil film with thickness of about 70 nm remained till 90 second from the impact loading, and so the influence of unsaturated fatty acid constituting the rice bran oil was suggested.

Crack growth analysis of ferrite by AE (Acoustic Emission) technique was analyzed. In the analysis, fractal dimensions were calculated using image processing and AE energy calculation techniques. According to the analysis of AE energy, the micro crack propagation strength generated in the material was analyzed. In the experiment, ferrite material was used and a fatigue experiment was conducted. As a result, crack damage occurred when a proportional relationship was observed in the AE parameter crack volume. By the presented study, it was found that AE nondestructive inspection which evaluated the micro crack propagation strength of material damage by AE energy analysis is possible to evaluate from the proposed image processing technique as well.

In preventive maintenance to carry out check and maintenance before outbreak of a trouble, needs of condition based maintenance (CBM) increases. CBM is a method to carry out check and maintenance when it always observe an apparatus and grasp a sign of the deterioration. This report the result that examined a switch by an electromagnetic actuator. A characteristic of this method is to be able to grasp deterioration from change of the current and voltage of the electromagnet coil. Therefore, to watch a state without adding a sensor. As a result, confirmed that a change the operation waveform of the electromagnetic actuator on the condition that simulated a change of state. Therefore, state monitoring of the switch is enabled by operation waveform change.

In this study, the applicability of the nonlinear piezoelectric impedance modulation technique is preliminarily discussed as a methodology of in-service monitoring of the meshing condition of mating spur gears. In the conventional approaches of the gear condition monitoring, vibration-based methodologies have been intensively studied and already commercialized because of the advantages of their simplicity. It is, however, hard to apply these approaches to the condition monitoring of low-speed gears because they are based on the detection of the impulsive response of the gear induced by a damaged tooth. The goal of this study is to develop more direct and active approach to the condition monitoring of gear teeth meshing. To this end, an ultrasonic elastic wave-based techniques and the concept of smart structures are introduced. In this paper, a piezoelectric wafer actuator is attached to the end face of the target gear, and the electro-mechanical coupling admittance of the piezoelectric actuator is measured to evaluate the sensitivity of the natural frequencies of the gear in the high-frequency range to the rotation angle and the change of the transfer torque under a static meshing condition. Discussions are made by comparing the experimental results with a finite element analysis.

This paper proposes a bridge inspection robot capable of traveling on inner right-angle paths. Newly developed robot can move with two motor driven wheels that consist of iron and neodymium magnets and form a magnetic circuit during adsorbing to a steel plate and generates strong magnetic force. With this strong magnetic force, it becomes possible to mount heavy object such as an ultrasonic sensor. However, this robot need a pushing mechanism to cancel a strong magnetic force of the running steel surface. Using the pushing mechanism, the robot achieves traveling on any inner right-angle paths. In this study, experiments have been conducted to check the performance of the robot. It have been found that the robot has sufficient running and carrying ability to carry various sensors and achieve traveling the inner right-angle path by the pushing mechanism.

In recent years, developed infrastructures after the period of high economic growth have been aging, which has become a problem in Japan. Periodic inspections have been conducted to detect and repair their damaged areas early. However, enormous costs and a long time are required to do them. In addition, there are insufficient inspectors for the number of structures. Therefore, there is a strong demand for bridge inspection robots to aid in this task. Practical bridge inspection robots should be able to move on the inspection route with simple maneuver. For that purpose, an appropriate control system needs to run autonomous on the bridge. Therefore, we have developed a simulation model of a bridge inspection robot BIREM-IV which has 10 degrees of freedom. And, it is possible to reproduce the behavior of BIREM-IV running on a wall surface. Using this model, we also developed control system to stabilize the robot motion at a right angle part like the wall-to-ceiling part.

In recent years, from the purpose of environmental impact reduction and earthquake-proof improvement, even if it carries out an industrial apparatus pair not only in a building, the earthquake-proof performance attracts attention.

So the electrical machinery apparatus maker is tackling the earthquake-proof improvement in industrial apparatus.

In this research, the FEM analysis model of cast resin transformer was built, and comparison verification of the analysis accuracy was carried out with the experiment. Since it united and the technical point required for earthquake-proof improvement was considered, the result is shown.