Flexible Strain-Sensitive Silicone-CNT Sensor for Human Motion Detection

This article describes the manufacturing technology of biocompatible flexible strain-sensitive sensor based on Ecoflex silicone and multi-walled carbon nanotubes (MWCNT). The sensor demonstrates resistive behavior. Structural, electrical, and mechanical characteristics are compared. It is shown that laser radiation significantly reduces the resistance of the material. Through laser radiation, electrically conductive networks of MWCNT are formed in a silicone matrix. The developed sensor demonstrates highly sensitive characteristics: gauge factor at 100% elongation −4.9, gauge factor at 90° bending −0.9%/deg, stretchability up to 725%, tensile strength 0.7 MPa, modulus of elasticity at 100% 46 kPa, and the temperature coefficient of resistance in the range of 30–40 °С is −2 × 10−3. There is a linear sensor response (with 1 ms response time) with a low hysteresis of ≤3%. An electronic unit for reading and processing sensor signals based on the ATXMEGA8E5-AU microcontroller has been developed. The unit was set to operate the sensor in the range of electrical resistance 5–150 kOhm. The Bluetooth module made it possible to transfer the received data to a personal computer. Currently, in the field of wearable technologies and health monitoring, a vital need is the development of flexible sensors attached to the human body to track various indicators. By integrating the sensor with the joints of the human hand, effective movement sensing has been demonstrated.

Drift Compensation of the Electronic Nose in the Development of Instruments for Out-of-Laboratory Analysis

A technique was developed to evaluate and compensate for the drift of eight mass-sensitive sensors in an open detection cell in order to estimate the influence of external factors (temperature, changes in the chemical composition of the background) on the out-of-laboratory analysis of biosamples. The daily internal standardization of the system is an effective way to compensate for the sensor signal drift when the sorption properties of sensitive coatings change during their long-term, intensive operation. In this study, distilled water was proposed as a standard for water matrix-based biosamples (blood, exhaled breath condensate, urine, etc.). Further, internal standardization was based on daily calculation of the specific sensor signals by dividing the sensor signals for the biosample according to the corresponding averaged values obtained from three to five standard measurements. The stability of the sensor array operation was estimated using the theory of statistical process control (exponentially weighted moving average control charts) based on the specific signal of the sensor array. The control limits for the statistical quantity of the central tendency for each sensor and the whole array, as well as the variations of the sensor signals, were determined. The average times required for signal and run lengths, for the purpose of statistically substantiated monitoring of the electronic nose’s stability, were calculated. Based on an analysis of the tendency and variations in sensor signals during 3 months of operation, a technique was formulated to control the stability of the sensor array for the out-of-laboratory analysis of the biosamples. This approach was successfully verified by classifying the results of the analysis of the blood and water samples obtained for this period. The proposed technique can be introduced into the software algorithm of the electronic nose, which will improve decision-making during the long-term monitoring of health conditions in humans and animals.

Evaluation of Machine Learning Methods for Monitoring the Health of Guyed Towers

This paper presents the development of a methodology to detect and evaluate faults in cable-stayed towers, which are part of the infrastructure of Brazil’s interconnected electrical system. The proposed method increases system reliability and minimizes the risk of service failure and tower collapse through the introduction of predictive maintenance methods based on artificial intelligence, which will ultimately benefit the end consumer. The proposed signal processing and interpretation methods are based on a machine learning approach, where the tower vibration is acquired from accelerometers that measure the dynamic response caused by the effects of the environment on the towers through wind and weather conditions. Data-based models were developed to obtain a representation of health degradation, which is primarily based on the finite element model of the tower, subjected to wind excitation. This representation is also based on measurements using a mockup tower with different types of provoked degradation that was subjected to ambient changes in the laboratory. The sensor signals are preprocessed and submitted to an autoencoder neural network to minimize the dimensionality of the resources involved, being analyzed by a classifier, based on a Softmax configuration. The results of the proposed configuration indicate the possibility of early failure detection and evolution evaluation, providing an effective failure detection and monitoring system.

A model for estimating the value of the applied pressure based on the analysis of tactile sensor signals using machine learning methods

В рамках настоящего исследования был проведен анализ существующих работ, посвященных интерпретации показаний тактильных сенсорных устройств, по результатам которого была предложена модель машинного обучения, позволяющая осуществлять оценку величины приложенного давления к поверхности тактильного сенсора давления емкостного типа. В качестве опорных моделей обработки и интерпретации сигналов данного устройства в работе рассматривались несколько методов машинного обучения: линейная регрессия, полиномиальная регрессия, регрессия дерева решений, частичная регрессия наименьших квадратов и полносвязная нейронная сеть прямого распространения. Обучение опорных моделей и апробация конечного решения проводилась на авторском наборе данных, включающем в себя более 3000 экземпляров данных. Согласно полученным результатам, наилучшее качество определения величины приложенного давления продемонстрирован решением на основе полносвязной нейронной сети прямого распространения. Коэффициент детерминации и средний модуль отклонения для данного решения на тестовой выборке составили 0,93 и 13,14 кПа соответственно. Currently, in the field of developing sensing systems for robotic means, one of the urgent tasks is the problem of interpreting the data of tactile pressure and proximity sensors. As a rule, the solution to this problem is complicated both by the dependence of the indicators of tactile sensors on the type of object’s material and by the design features of each individual device. In this study, an analysis of existing works devoted to the interpretation of the readings of tactile sensor devices was carried out. According to the analysis results a machine learning model was proposed that allows estimating the amount of pressure applied to the surface of a tactile pressure sensor of a capacitive type. The architecture of the proposed model includes two key blocks of data analysis, the first one is aimed at recognizing the type of interaction object’s material and the second is devoted to the direct assessment of the magnitude of the pressure applied to the sensor. Several machine learning methods were considered as supporting models for processing and interpreting the signals of this device: linear regression, polynomial regression, decision tree regression, partial least squares regression and a fully connected feedforward neural network.

Verification of the method of finding a faulty sensor in reduced meters

The reliability of the method for determining a failed sensor in a redundant angular velocity meter (AVM) by means of its experimental verification is considered. The mutual non-orthogonal arrangement of six axes of sensor sensitivity has been optimized to reduce the instrumental errors of each sensitive element and ensure the equality of their contribution. Provides approximately the same sensitivity to the level of error in case of failure. One of the six experimentally obtained sensor signals contains an error that exceeds the specified permissible limit. The algorithm for searching for a sensor is checked, the error of which exceeds the specified one, and which, for this reason, is considered faulty.

Fault Detection by Signal Reconstruction in Nuclear Power Plants

In this work, the recently developed auto associative bilateral kernel regression (AABKR) method for on-line condition monitoring of systems, structures, and components (SSCs) during transient process operation of a nuclear power plant (NPP) is improved. The advancement enhances the capability of reconstructing abnormal signals to the values expected in normal conditions during both transient and steady-state process operations. The modification introduced to the method is based on the adoption of two new approaches using dynamic time warping (DTW) for the identification of the time position index (the position of the nearest vector within the historical data vectors to the current on-line query measurement) used by the weighted-distance algorithm that captures temporal dependences in the data. Applications are provided to a steady-state numerical process and a case study concerning sensor signals collected from a reactor coolant system (RCS) during start-up operation of a NPP. The results demonstrate the effectiveness of the proposed method for fault detection during steady-state and transient operations.

Predicting Human Motion Signals Using Modern Deep Learning Techniques and Smartphone Sensors

The global adoption of smartphone technology affords many conveniences, and not surprisingly, healthcare applications using wearable sensors like smartphones have received much attention. Among the various potential applications and research related to healthcare, recent studies have been conducted on recognizing human activities and characterizing human motions, often with wearable sensors, and with sensor signals that generally operate in the form of time series. In most studies, these sensor signals are used after pre-processing, e.g., by converting them into an image format rather than directly using the sensor signals themselves. Several methods have been used for converting time series data to image formats, such as spectrograms, raw plots, and recurrence plots. In this paper, we deal with the health care task of predicting human motion signals obtained from sensors attached to persons. We convert the motion signals into image formats with the recurrence plot method, and use it as an input into a deep learning model. For predicting subsequent motion signals, we utilize a recently introduced deep learning model combining neural networks and the Fourier transform, the Fourier neural operator. The model can be viewed as a Fourier-transform-based extension of a convolution neural network, and in these experiments, we compare the results of the model to the convolution neural network (CNN) model. The results of the proposed method in this paper show better performance than the results of the CNN model and, furthermore, we confirm that it can be utilized for detecting potential accidental falls more quickly via predicted motion signals.

Pocketable Labs for Everyone: Synchronized Multi-Sensor Data Streaming and Recording on Smartphones with the Lab Streaming Layer

The streaming and recording of smartphone sensor signals is desirable for mHealth, telemedicine, environmental monitoring and other applications. Time series data gathered in these fields typically benefit from the time-synchronized integration of different sensor signals. However, solutions required for this synchronization are mostly available for stationary setups. We hope to contribute to the important emerging field of portable data acquisition by presenting open-source Android applications both for the synchronized streaming (Send-a) and recording (Record-a) of multiple sensor data streams. We validate the applications in terms of functionality, flexibility and precision in fully mobile setups and in hybrid setups combining mobile and desktop hardware. Our results show that the fully mobile solution is equivalent to well-established desktop versions. With the streaming application Send-a and the recording application Record-a, purely smartphone-based setups for mobile research and personal health settings can be realized on off-the-shelf Android devices.