MEMS Inertial Sensor Fault Diagnosis Using a CNN-Based Data-Driven Method

Micro electro mechanical systems (MEMS) technology has developed considerably in recent years and many sensors utilizing this technology are available in the market. MEMS technology enables miniaturization, mass production, and cost reduction of many sensors. In particular, MEMS inertial sensors that include an acceleration sensor and an angular velocity sensor (gyroscope, or simply ldquogyrordquo) are the most popular devices. Applications of inertial sensors have now extended into the field of networked sensing systems. In this presentation, we will describe current MEMS inertial sensors and their applications.

The use of global navigation satellite system receivers for navigation still presents many challenges in urban canyon and indoor environments, where satellite availability is typically reduced and received signals are attenuated. To improve the navigation performance in such environments, several enhancement methods can be implemented. For instance, external aid provided through coupling with other sensors has proven to contribute substantially to enhancing navigation performance and robustness. Within this context, coupling a very simple GPS receiver with an Inertial Navigation System (INS) based on low-cost micro-electro-mechanical systems (MEMS) inertial sensors is considered in this paper. In particular, we propose a GPS/INS Tightly Coupled Assisted PLL (TCAPLL) architecture, and present most of the associated challenges that need to be addressed when dealing with very-low-performance MEMS inertial sensors. In addition, we propose a data monitoring system in charge of checking the quality of the measurement flow in the architecture. The implementation of the TCAPLL is discussed in detail, and its performance under different scenarios is assessed. Finally, the architecture is evaluated through a test campaign using a vehicle that is driven in urban environments, with the purpose of highlighting the pros and cons of combining MEMS inertial sensors with GPS over GPS alone.

Small, lightweight, power-efficient and low-cost microelectromechanical system (MEMS) inertial sensors and microcontrollers available in the market today help reduce the instability of Multibeam Sonars. Current MEMS inertial measurement units (IMUs) come in many shapes, sizes, and costs - depending on the application and performance required. Although MEMS inertial sensors offer affordable and appropriately scaled units, they are not currently capable of meeting all requirements for accurate and precise attitudes, due to their inherent measurement noise. The article presents the comparison of different MEMS technologies and their parameters regarding to the main application, namely Multibeam Echo Sounders (MBES). The quality of MEMS parameters is crucial for further MBES record-processing. The article presents the results of undertaken researches in that area, and these results are relatively positive for low-cost MEMS. The paper undertakes some vital aspect of using MEMS in the attitude and heading reference system (AHRS) context. The article presents a few aspects of MEMS gyro errors and their estimation process in the context of INS processing flow, as well as points out the main difficulties behind the INS when using a few top MEMS technologies.

The procedure of determining the initial values of the attitude angles (pitch, roll, and heading) is known as the alignment. Also, it is essential to align an inertial system before the start of navigation. Unless the inertial system is not aligned with the vehicle, the information provided by MEMS (microelectromechanical system) sensors is not useful for navigating the vehicle. At the moment MEMS gyroscopes have poor characteristics and it’s necessary to develop specific algorithms in order to obtain the attitude information of the object. Most of the standard algorithms for the attitude estimation are not suitable when using MEMS inertial sensors. The wavelet technique, the Kalman filter, and the quaternion are not new in navigation data processing. But the joint use of those techniques for MEMS sensor data processing can give some new results. In this paper the performance of a developed algorithm for the attitude estimation using MEMS IMU (inertial measurement unit) is tested. The obtained results are compared with the attitude output of another commercial GPS/IMU device by Xsens. The impact of MEMS sensor measurement noises on an alignment process is analysed. Some recommendations for the Kalman filter algorithm tuning to decrease standard deviation of the attitude estimation are given.

Two adaptively combined gyro and triaxial accelerometer arrays of commercial-grade Micro Electro-Mechanical Systems (MEMS) inertial sensors and a real-time embedded algorithm suite significantly enhance accuracy over that of single commercial-grade devices. Denoised sensor data is combined with a Kalman-filter bias compensation algorithm to achieve better than 0.1 degree/hour in-run gyro bias stability (standard deviation). One denoise-filtered triaxial accelerometer sensor array provides better than 30 μg accelerometer bias stability. The hardware is embedded in an exoskeleton that maintains accurate fine alignment adjustments and provides protection when in rugged environments. Analysis of results measured from Allan variance computations of real-time data from several different IMU hardware units demonstrates that each of three orthogonally positioned precision gyro arrays, and their associated processing, achieve less than 0.03 deg integrated angle drift over 5 hours. New IMU hardware is now in fabrication for delivery in early 2010, when received, we expect to overcome the limits of the present hardware design to produce additional significant improvements. The size, weight, and power of these new IMU's are suitable for man-portable, hand-held systems, UAVs, and munitions systems. Early work for this technology was first published at ION/PLANS 2008.

Recently, navigation and localization for ground vehicles have become a hot research topic. The use of low-cost MEMS (micro-electromechanical system) inertial sensors and GNSS (global navigation satellite system) receivers are widespread in this industry. To perform ZUPT (zero velocity update) measurement update for the navigation filter to correct PVA (position, velocity, attitude) errors of a typical inertial navigation system (INS) using inertial sensors, in this paper, a method of detecting ZUPT phases of a typical ground vehicle is proposed based on machine learning. The algorithm detects stationary and dynamic movements of vehicles through SVM (support vector machine), which is well-known for solving binary classification problems. The proposed SVM-based ZUPT detection approach can improve the accuracy of a GNSS/INS integrated system by replacing the traditional threshold-based detector. The test and results indicate that the proposed method could recognize the ZUPT phases of a vehicle with high efficiency and accuracy.

In biology, mechanosensors, equipped with differing hair-like structures for signal pick-up, are sensitive to a variety of physical quantities like acceleration, flow, rotational rate, balancing and IR-light. As an example, crickets use filiform hairs for sensing of low-frequency flows to obtain information about the environment and avoid e.g. predator attacks crickets. Their filiform hairs are able to sense airflows with velocity amplitudes down to 30 m s -1 and operate around the energy levels of thermal noise. Taking these hair-sensors as a source of inspiration, hair-sensor inspired flow-sensors for measurement of (tiny) ac-airflows using capacitive readout have been designed and fabricated using technology generally denoted as MEMS (microelectromechanical systems). To determine the performance of these flow sensitive hair-sensors, three different setups for oscillatory airflow are used for thorough characterization. Each of these flow sources has specific properties regarding frequency range, pressure dependence and bandwidth. By combining information from the used flow setups important insights in the sensor operation are gained and discussed.

Over the few years civilian and military applications are maneuvering with multi-sensor Unmanned Aerial Vehicles (UAV). The aim of this paper is to improve autonomous UAV positioning by using Inertial measurement unit (IMU) and Global Navigation Satellite System (GNSS) is considered. The positional accuracy of UAV is limited to GNSS signal blockage. The advancements in micro electro mechanical systems (MEMS) helps to design low-cost IMU with meter-level accuracy in positioning. The estimation of attitude, position and velocity measurements are determined from a nine degree of freedom Arduino embedded IMU sensor. The Low-cost IMU sensor has been tested under static condition and then a dynamic movement of sensor is observed under moving base-line conditions on February 2 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">nd</sup> 2020 at Koneru Lakshmaiah Education Foundation (KLEF). In this paper, covariance filters are implemented to eliminate the gyro bias and accelerometer bias for improving attitude determination of IMU navigation system. The positioning from IMU sensor is derived using standard strap-down inertial navigation system (SINS) formulation. The position error from SINS indicates that IMU sensor is able to perform well with maximum position error of 5 meters and 2.5 meters for static and dynamic environments respectively. The outcome of this work would be useful for integration of navigation sensors using kalman filter based advanced techniques for UAV applications.

Efficient fault diagnosis approach for solar photovoltaic array using a convolutional neural network in combination of generative adversarial network under small dataset

Rolling bearings are essential for ensuring smooth operation of rotating machinery, when it suffer from faults, the extraction of fault features and fault diagnosis of rolling bearings become difficult due to the interference of noise and other vibration components. A fault diagnosis model based on symplectic geometric mode decomposition (SGMD), convolutional neural networks (CNN) and optimized long short-term memory networks (LSTM) is proposed to facilitate feature extraction and improve robustness of fault diagnosis. Firstly, bearing signals from different faults are decomposed using the SGMD, selecting the symplectic seometric components (SGCs) based on their correlation and constructing feature vectors by fuzzy dispersion entropy (FDE). Then, a CNN-LSTM model is built by combining the CNN and the LSTM. Inputting training data into the model to adjust its hyperparameters, and the softmax layer is replaced with a least squares support vector machine (LSSVM). The parameters of the LSSVM are optimized with the electric eel foraging optimization (EEFO) algorithm and the optimized model is used for fault identification and classification. Finally, experiments are conducted on two public datasets, achieving accuracy rates of 98.57% and 97.22% respectively, which validates the feasibility of this method and demonstrates its superiority over traditional CNN models.

Autonomous landing system is an important part of unmanned aerial vehicle (UAV) autonomous flight. Large fixed‐wing UAVs often consume more onboard space and computing resources. With the improvement of system integration and the reduction of high‐performance global navigation satellite system (GNSS) costs, real‐time kinematic (RTK) GNSS has been widely used in small‐scale and low‐cost fixed‐wing UAVs. This article uses less onboard hardware and ground facility requirements to design an autonomous taking off and landing system suitable for general fixed‐wing UAVs. A modified gradient descent attitude estimator, based on microelectromechanical system (MEMS) inertial measurement unit (IMU) and RTK GNSS, and a guidance method based on a tracking differentiator for using in a high dynamic environment are proposed, and circular path following and vertical transient simulations are performed. The simulation results show that this method can improve attitude estimation accuracy, motion control accuracy, and altitudinal transient quality. Finally, the actual flight test verifies its feasibility in the actual environment. The flight test results prove that it is feasible to implement the autonomous landing process of fixed‐wing UAV with the proposed system and method.

In the process of Unmanned Aerial Vehicle (UAV) flight testing, plenty of compound faults exist, which could be composed of concurrent single faults or over-limit states alarmed by Built-In-Test (BIT) equipment. At present, there still lacks a suitable automatic labeling approach for UAV flight data, effectively utilizing the information of the BIT record. The performance of the originally employed flight data-driven fault diagnosis models based on machine learning needs to be improved as well. A compound fault labeling and diagnosis method based on actual flight data and the BIT record of the UAV during flight test phase is proposed, through labeling the flight data with compound fault modes corresponding to concurrent single faults recorded by the BIT system, and upgrading the original diagnosis model based on Gradient Boosting Decision Tree (GBDT) and Fully Convolutional Network (FCNN), to eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM) and modified Convolutional Neural Network (CNN). The experimental results based on actual test flight data show that the proposed method could effectively label the flight data and obtain a significant improvement in diagnostic performance, appearing to be practical in the UAV test flight process.

This paper exploits the idea of integrating the outputs of several optical flow sensors (OFSs) and a micro-electro-mechanical systems (MEMS) inertial measurement unit (IMU). Several optical flow sensors are fixed on an unmanned aerial vehicle (UAV). Their locations and orientations are different so as to detect the optical flow from different directions. By fusing the optical flow information and inertial navigation system (INS) measurements using an extended Kalman filter (EKF), the INS errors can be estimated and compensated. The integration of optical flow sensors and MEMS-based IMU provides a small size, low cost and self-contained scheme for UAVs navigation at low height above the ground in GPS denied environments. The preliminary simulation results presented in this article illustrate the effectiveness of the proposed integrated navigation scheme.

Natural disasters are recurrent weather phenomena whose occurrence has increased worldwide in the past few decades. These disasters cause devastating effects on transportation routes by causing significant damage and obstruction on frequently traveled roads. This research focuses on developing an autonomous network of unmanned aerial vehicles (UAVs) for transportation disaster management using convolutional neural networks (CNNs). The autonomous network of UAVs will allow first responders to optimize their rescue plans by providing relevant information on inaccessible roads. The autonomous UAV system development will increase the affected regions' recovery rate by identifying blocked transportation routes and associating them with their corresponding locations to update the virtual map in real-time. Live footage from the unmanned aerial vehicles is fed to ground control, where the CNN classifies the type of damage encountered and then updates a virtual map through the ArcGIs software. Preliminary results of the classification models such as AlexNet show average accuracy of 74.07%. Furthermore, transfer learning and cross-validation techniques were applied to the CNN models to obtain high confidence levels due to the small dataset size used to train and test the CNNs. To choose the best CNN model, a quantitative analysis was performed to measure the statistical precision, statistical recall, and F1 score on each model to optimize the classification.

&amp;lt;p&amp;gt;With consumer-grade unmanned aerial vehicles (UAVs) on the rise, which enable easy, time-flexible, and cost-effective acquisition of very high-resolution RGB data, the mapping of forest tree species using solely RGB imagery is of high interest, as it does not rely on sophisticated sensors, does not require extensive calibration and preprocessing and, therefore, enables the application by a wide audience. In combination with convolutional neural networks (CNNs), which particularly exploit spatial patterns and, therefore, highly benfit from very high-resolution remote sensing, this offers great potential for accurately mapping forest tree species.&amp;lt;/p&amp;gt;&amp;lt;p&amp;gt;Here, we present the findings of our recent study, in which we used very high-resolution RGB imagery from UAVs in combination with CNNs for the mapping of forest tree species. In this study, we used multicopter UAVs to obtain very high-resolution (&amp;lt;2 cm) RGB imagery over 51&amp;amp;#160;ha of temperate forests in the Southern Black Forest region, and the Hainich National Park in Germany. To fully harness the end-to-end learning capabilities of CNNs, we used a semantic segmentation approach (U-net) that concurrently segments and classifies tree species from imagery. With a diverse dataset in terms of study areas, site conditions, illumination properties, and phenology, we accurately mapped nine tree species, three genus-level classes, deadwood, and forest floor (mean F1-score 0.73). We found that a coarser spatial resolution substantially reduced the model accuracy (mean F1-score of 0.26 at 32&amp;amp;#160;cm resolution) and that larger tile sizes during CNN training negatively affected the model accuracies for underrepresented classes. Additional height information from normalized digital surface models slightly increased the model accuracy but simultaneously increased computational complexity and data requirements. Our results highlight the key role that UAVs can play in the mapping of forest tree species, given that air- and spaceborne remote sensing currently does not provide comparable spatial resolutions. Given the end-to-end learning capabilites of CNNs extensive preprocessing becomes partly obsolete, whereas the use of a large and diverse dataset facilitates a high degree of generalization of the CNN, thus fostering transferability. The synergy of high-resolution UAV imagery and CNN provide a fast and flexible yet accurate means of mapping forest tree species.&amp;lt;/p&amp;gt;&amp;lt;p&amp;gt;In this contribution, we will give an outlook on how the combination of UAV imagery and CNNs can be integrated with multitemporal satellite imagery (Sentinel-1 and Sentinel-2) in order to extrapolate the UAV-based tree species maps to larger areas.&amp;lt;/p&amp;gt;