Marja Ruotsalainen

GAIS: A Method for Detecting Interleaved Sequential Patterns from Imperfect Data

This paper introduces a novel method, GAIS, for detecting interleaved sequential patterns from databases. A case, where data is of low quality and has errors is considered. Pattern detection from erroneous data, which contains multiple interleaved patterns is an important problem in a field of sensor network applications. We approach the problem by grouping data rows with the help of a model database and comparing groups with the models. In evaluation GAIS clearly outperforms the greedy algorithm. Using GAIS desired sequential patterns can be detected from low quality data.

Classification of items in a walk-through metal detector using time series of eigenvalues of the polarizability tensor

During the last decade, the safety regulations of the airports have been set to a new level. As the number of passengers is constantly increasing, yet effective but quick security control at checkpoints sets great requirements to the 21st century security systems. In this paper, we shall introduce a novel metal detector concept that enables not only to detect but also to classify hidden items, though their orientation and accurate location are unknown. Our new prototype walk-through metal detector generates mutually orthogonal homogeneous magnetic fields so that the measured dipole moments allow classification of even the smallest of the items with high degree of accuracy in real-time. Invariant to different rotations of an object, the classification is based on eigenvalues of the polarizability tensor that incorporate information about the item (size, shape, orientation etc.); as a further novelty, we treat the eigenvalues as time series. In our laboratory settings, no assumptions concerning the typical place, where an item is likely situated, are made. In that case, 90 % of the dangerous and harmless items, including knives, guns, gun parts, belts etc. according to a security organisation, are correctly classified. Made misclassifications are explained by too similar electromagnetic properties of the items in question. The theoretical treatment and simulations are verified via empirical tests conducted using a robotic arm and our prototype system. In the future, the state-of-the-art system is likely to speed-up the security controls significantly with improved safety.

Minimizing Fatigue Damage in Aircraft Structures

Aircraft structural health monitoring (SHM) refers to a process in which sensors assess the current (and predict the future) state of a structure in terms of its aging and deterioration to assure users or operators of its safety and performance. In addition to preventing failures, SHM extends aircraft life cycles. Consequently, adopting SHM is strongly motivated not only by flight safety but also by economic considerations. This article focuses on the optimization of aircraft usage as a new aspect of SHM and discusses a knowledge discovery approach based on dynamic time warping and genetic programming. In addition, it points out some of the challenges faced in applying artificial intelligence to aircraft SHM. This novel work reveals that AI provides a means to gain valuable knowledge for decision making on cost-efficient future usage of an aircraft fleet.

Link between Flight Maneuvers and Fatigue

Structural integrity management is in a key role when operating with an ageing fleet. Preventive actions aid in keeping structures healthy and ensuring the designed lifetime for the aircraft. Our research focuses on exploitation of collected usage data by identifying actual in-flight events that cause the major fatigue life expenditure of the fatigue-critical structural details. In order to build a link between the events and damage, we have developed software for flight maneuver identification. As a latest advancement, we have created data models for several flight maneuvers and constructed a model library, referred to as template library. The library instructs the software about what the interesting events look like in the data. Our software together with the template library allows us to perform maneuver-specific fatigue assessment and achieve knowledge concerning the fatigue-criticality of various flight maneuver types. This lays a foundation for detailed analysis of the identified, nominally similar, maneuvers and identification of small crucial actions within the maneuvers that are behind the fatigue. In this paper, we consider the issues related to maneuver-specific fatigue assessment and present analysis results for four structural details of F-18 aircraft with a template library of seven flight maneuvers. We summarize the requirements and prospect of our fatigue analysis approach and prove its applicability.

Reasoning logical rules from multisensor data for lifecycle management of aircraft structures

Maneuvering produces strain loading, which cumulates as fatigue damage to the structures of fighter jets. In order to manage the integrity of structures, critical structural details are monitored with specially designed sensor instrumentation. Monitoring aims at estimating remaining fatigue life of structures, i.e. producing knowledge on how much more fatigue damage each structural detail can undergo before it has to be repaired or changed. Fatigue life expenditure (FLE) caused by flying a maneuver can be estimated using available fatigue life calculation tools. The analyses concerning the usage of aircraft have shown that caused FLE depends not only on the type of the flight maneuver but also on the way the maneuver was flown. In other words, the dispersion of FLE values of nominally similar flight maneuvers can be very high. This paper considers the application of genetic programming (GP) to reasoning such logical rules from flight parameter signals that explain damaging of structural details during certain flight maneuvers. In other words, we are searching for easily interpretable logical rules with which it is possible to predict whether a flight maneuver instance causes minor or major damage to the structural detail in study. In the experiments, GP approach is compared to a tree classifier with the real flight monitoring data of the Finnish Air Force F-18 fleet. The rules reasoned by GP outperform the rules by the tree classifier in accuracy and simplicity. The knowledge deduced from the reasoned rules can be exploited in pilot training when learning to fly maneuvers in less damaging way.

Learning of a tracker model from multi-radar data for performance prediction of air surveillance system

A valid model of the air surveillance system performance is highly valued when making decisions related to the optimal control of the system. We formulate a model for a multi-radar tracker system by combining a radar performance model with a tracker performance model. A tracker as a complex software system is hard to model mathematically and physically. Our novel approach is to utilize machine learning to create a tracker model based on measurement data from which the input and target output for the model are calculated. The measured data comprises the time series of 3D coordinates of cooperative aircraft flights, the corresponding target detection recordings from multiple radars, and the related multi-radar track recordings. The collected data is used to calculate performance measures for the radars and the tracker at specific locations in the air space. We apply genetic programming to learning such rules from radar performance measures that explain tracker performance. The easily interpretable rules are intended to reveal the real behavior of the system providing comprehension for its control and further development. The learned rules allow predicting tracker performance level for the system control in all radar geometries, modes, and conditions at any location. In the experiments, we show the feasibility of our approach to learning a tracker model and compare our rule learner with two tree classifiers, another rule learner, a neural network, and an instance-based classifier using the real air surveillance data. The tracker model created by our rule learner outperforms the models by the other methods except for the neural network whose prediction performance is equal.

Classification of radar data by detecting and identifying spatial and temporal anomalies

For some time, applying the theory of pattern recognition and classification to radar signal processing has been a topic of interest in the field of remote sensing. Efficient operation and target indication is often hindered by the signal background, which can have similar properties with the interesting signal. Because noise and clutter may constitute most part of the response of surveillance radar, aircraft and other interesting targets can be seen as anomalies in the data. We propose an algorithm for detecting these anomalies on a heterogeneous clutter background in each range-Doppler cell, the basic unit in the radar data defined by the resolution in range, angle and Doppler. The analysis is based on the time history of the response in a cell and its correlation to the spatial surroundings. If the newest time window of response in a resolution cell differs statistically from the time history of the cell, the cell is determined anomalous. Normal cells are classified as noise or different type of clutter based on their strength on each Doppler band. Anomalous cells are analyzed using a longer time window, which emulates a longer coherent illumination. Based on the decorrelation behavior of the response in the long time window, the anomalous cells are classified as clutter, an airplane or a helicopter. The algorithm is tested with both experimental and simulated radar data. The experimental radar data has been recorded in a forested landscape.

Decomposing radar measurements through comprehensive response modeling

Radars are used for various purposes, and we need flexible methods to explain radar response phenomena. In general, modeling radar response and backscatterers can help in data analysis by providing possible explanations for measured echoes. However, extracting exact physical parameters of a real world scene from radar measurements is an ill-posed problem. Our study aims to enhance radar signal interpretation and further to develop data classification methods. In this paper, we introduce an approach for finding physically sensible explanations for response phenomena during a long illumination. The proposed procedure uses our comprehensive response model to decompose measured radar echoes. The model incorporates both a radar model and a backscatterer model. The procedure adapts the backscatterer model parameters to catch and reproduce a measured Doppler spectrum and its dynamics at a particular range and angle. A filter bank and a set of features are used to characterize these response properties. The procedure defines a number of point-scatterers for each frequency band of the measured Doppler spectrum. Using the same features calculated from simulated response, it then matches the parameters-the number of individual backscatterers, their radar cross sections and velocities-to joint Doppler and amplitude behavior of the measurement. Hence we decompose the response toward its origin. The procedure is scalable and can be applied to adapt the model to various other features as well, even those of more complex backscatterers. Performance of the procedure is demonstrated with radar measurements on controlled arrangement of backscatterers with a variety of motion states.

Cumulative sum and neural network approach to the detection and identification of hazardous chemical agents from ion mobility spectra

The detection and identification of hazardous chemical agents are important problems in the fields of security and defense. Although the diverse environmental conditions and varying concentrations of the chemical agents make the problem challenging, the identification system should be able to give early warnings, identify the gas reliably, and operate with low false alarm rate. We have researched detection and identification of chemical agents with a swept-field aspiration condenser type ion mobility spectrometry prototype. This paper introduces an identification system, which consists of a cumulative sum algorithm (CUSUM) -based change detector and a neural network classifier. As a novelty, the use of CUSUM algorithm allows the gas identification task to be accomplished using carefully selected measurements. For the identification of hazardous agents we, as a further novelty, utilize the principal component analysis to transform the swept-field ion mobility spectra into a more compact and appropriate form. Neural networks have been found to be a reliable method for spectra categorization in the context of swept-field technology. However, the proposed spectra reduction raises the accuracy of the neural network classifier and decreases the number of neurons. Finally, we present comparison to the earlier neural network solution and demonstrate that the percentage of correctly classified sweeps can be considerably raised by using the CUSUM-based change detector.