Olivier Janssens

Data-driven multivariate power curve modeling of offshore wind turbines

Performance monitoring of offshore wind turbines is an essential first step in the condition monitoring process. This paper provides three novelties regarding power curve modeling. The first consists of illustrating that univariate power curve modeling can be improved by the use of non-parametric methods such as stochastic gradient boosted regression trees, extremely randomized forest, random forest, K-nearest neighbors, and the method of bins according to the IEC standard 61,400-12-1. This is confirmed on both a synthetic data set and a real live data set containing data from three offshore wind turbines. The second novelty consists of an improvement regarding overall power curve modeling results by the use of multivariate models which incorporate the wind direction, rotations per minute of the rotor, yaw, wind direction and pitch additional to the wind speed. The best improvement is achieved by the stochastic gradient boosted regression trees method for which the mean absolute error can be decreased by up to 27.66%. The third novelty consists of making a synthetic data set available for bench-marking purposes. Data-driven non-parametric power curve models perform the best.Multivariate model outperforms univariate models.Data-driven multivariate models capture the phenomena in the data.

Comparing simple video quality measures for loss-impaired video sequences on a large-scale database

The performance of objective video quality measures is usually identified by comparing their predictions to subjective assessment results which are regarded as the ground truth. In this work we propose a complementary approach for this performance evaluation by means of a large-scale database of test sequences evaluated with several objective measurement algorithms. Such an approach is expected to detect performance anomalies that could highlight shortcomings in current objective measurement algorithms. Using realistic coding and network transmission conditions, we investigate the consistency of the prediction of different measures as well as how much their behavior can be predicted by content, coding and transmission features, discussing unexpected and peculiar behaviors, and highlighting how a large-scale database can help in identifying anomalies not easily found by means of subjective testing. We expect that this analysis will shed light on directions to pursue in order to overcome some of the limitations of existing reliability assessment methods for objective video quality measures.

Educational virtual game scenario generation for serious games

In order to help youngsters who are confronted with cyberbullying, the Friendly ATTAC project aims to develop a serious game. In this paper the novel virtual game scenario generation process for this serious game is presented. The goal of the process is to allow non-technical users to model virtual scenarios. After writing the scenarios, the scenarios can be modelled in ATTAC-L and afterwards translated in computer interpretable XML. This XML is then used to automatically build the scenario within the game engine so that it can be played. This paper details the ATTAC-L scenario generation and how it is translated to an in game scenario. Thanks to the presented tools and methods, it is possible to transform a written scenario into an in game virtual game scenario.

Real-time emotion classification of Tweets

Despite adding emotions to applications has proven to enhance the user experience, emotion recognition applications are still not widely available nor used. Within this paper, emotion recognition is done on Twitter tweets using six emotion classification algorithms that are compared on precision and timing. The paper shows that precision can be enhanced by 5.02% compared to the current state-of-the-art by improving the features. Furthermore, the presented algorithms work in real-time.

Enabling Control of 3D Visuals, Scenarios and Non-linear Gameplay in Serious Game Development Through Model-Driven Authoring

Due to the absence of high-level authoring environments and support for non-technical domain experts to create custom serious games, a model-driven authoring framework is presented in this paper. Through model-driven authoring, non-technical people can manipulate the 3D visuals of their serious game, model the scenarios of the game, and even easily add non-linear narrative to the game. The different tools and methods have been implemented and are currently used to build a serious game for the Friendly ATTAC project in order to help youngsters who are confronted with cyberbullying. The presented model-driven authoring framework enables non-technical domain experts to produce serious games easily and quickly, at a lower cost, and therefore lowers the barriers that hinder the production of serious games.

Thermal Imaging and Vibration Based Multi-Sensor Fault Detection for Rotating Machinery

In order to minimize operation and maintenance costs and extend the lifetime of rotating machinery, damaging conditions and faults should be detected early and automatically. To enable this, sensor streams should continuously be monitored, processed, and interpreted. In recent years, infrared thermal imaging has gained attention for the said purpose. However, the detection capabilities of a system that uses infrared thermal imaging is limited by the modality captured by this single sensor, as is any single sensor-based system. Hence, within this paper a multisensor system is proposed that not only uses infrared thermal imaging data, but also vibration measurements for automatic condition and fault detection in rotating machinery. It is shown that by combining these two types of sensor data, several conditions/faults and combinations can be detected more accurately than when considering the sensor streams individually.

A genetic algorithm for interpretable model extraction from decision tree ensembles

Models obtained by decision tree induction techniques excel in being interpretable. However, they can be prone to overfitting, which results in a low predictive performance. Ensemble techniques provide a solution to this problem, and are hence able to achieve higher accuracies. However, this comes at a cost of losing the excellent interpretability of the resulting model, making ensemble techniques impractical in applications where decision support, instead of decision making, is crucial.

Towards intelligent lubrication control: Infrared thermal imaging for oil level prediction in bearings

Rolling element bearings can suffer from energy losses that can be minimized by actively regulating the oil level in the bearings. To regulate the oil level automatically it has to be determined automatically. In this paper infrared thermal imaging is used for this purpose. Several infrared thermal videos are captured of a rotating set-up using various rotation speeds, loads, oil temperatures and flow rates. These infrared thermal videos are given as input to an image processing and machine learning system that can automatically extract the relevant region of interest, features, and subsequently make a prediction regarding the oil level in the bearing. Evaluation showed that the system achieves an accuracy of 96.67 % and misclassifies only one infrared thermal video which is a recording taken during a very high rotation speed which induced unfavorable conditions for the proposed approach.

Dynamic Monitoring Dashboards Through Composition of Web and Visualization Services

In order to present and communicate the condition of monitored environments to supervising experts, a dashboard is needed to present the status of all sensors. The heterogeneity and vast amount of sensors, as well as the difficulty of creating interesting sensor data combinations, hinder the deployment of fixed structure dashboards as they are unable to cope with the accordingly vast amount of required mappings. Therefore, in this paper, the development of a dynamic dashboard is presented, able to visualize any particular and user defined data and sensor composition. By implementing the heterogeneous sensors as semantically annotated Web apis, a dynamic sensor composition and visualization is enabled. The resulting condition monitoring dashboard provides a clear overview of the system kpis in acceptable timing and provides helpful tools to detect anomalies in system behaviour.

A Learning Based Approach for Real-Time Emotion Classification of Tweets

Tmotion recognition can be done in a wide range of applications to enhance the user experience. Because of these many types of applications there are is a large range of different data types that can be processed, such as text, video, speech, sound, accelerometer data and various bio-sensor data types. In order to bring emotion recognition into everyday use, it is important to work with data types and sources that are available to everyone. Therefore in this chapter twitter data is used for emotion recognition. Since emotion recognition applications need to uncover the user’s emotion fast, the focus lies on real-time emotion classification. Sentiment analysis or emotion recognition research often uses a lexicon based approach, though in this chapter a learning based approach is used. Nine emotion classification algorithms are compared with focus on precision and timing. This chapter shows that accuracy can be enhanced by 5.83 % compared to the current state-of-the-art by improving the features and that the presented method work in real-time.