Big Data

Big data focuses on discovering the nuggest of information found in large, industrial data sets. Not only must the information be found, but accuracy and relevancy must be determined. Within the area of big data, the following are topics of interest:

  • Data Quality Assessment: In data quality assessment, the goal is the development of methods and software that will allow a data set with be automatically partitioned into those regions that are useful for modelling and those regions that are not. Not only must the theoretical constraints be considered, but also user-defined constraints, such as known error values or undesired behaviour.
  • Soft Sensors: In many industries, variables such as concentration, density, or composition are difficult to measure in real time with sufficient accuracy and availability using real sensors. Instead, models of the system developed using easy-to-measure variables are used to provide information about the difficult-to-measure values. The development of soft sensors includes not only the modelling, but also their use and online updating. Using soft sensors as inputs to control loops can lead to various unexpected behaviour that needs to be considered when designing the control loop and soft sensor system.
  • Modelling and System Identification: In order to make decisions about how to operate a process, it is necessary to have an understating of how the system responds to changes. This means that models of the system are required. The development of models can occur using different methods and approaching varying from simple linear regression to complex methods requiring artificial intelligence and machine learning.
  • Design of Experiments: Most commonly, models are built using historical data that comes from normal operation of the plant. In certain circumstances, it may be possible to design experiment from which a model of the system may be developed. The goal of design of experiments is to design such experiments that provide with minimal effort the maximal amount of information about the process.

Holistic Control

Holistic control, as its name implies, concerns the development and maintenance of control systems from plant start-up to plant shut-down through all phases of operation. In order to achieve this ambitious goal, it is necessary to not only consider the control of the system itself, but also how to detect process changes, faults or other unexpected behaviour, and component changes. The following topics are of interest in this field:

  • Performance Monitoring: During the course of operation, the performance of a given control loop may change. Performance monitoring seeks to determine why the changes have occurred and what corrective action needs to be taken. Often, this will involve changing the tuning of the control loop to take into consideration the new operating conditions.
  • Fault Detection and Diagnosis: No matter how well a control system is built, there will always be unexpected behaviour, which is often called a fault. Faults can come in many different forms that need to not only be detected, but also properly diagnosed as to their origin. This requires the use of both a priori knowledge and knowledge gained from the available data.
  • Development of Control Strategies: In order to achieve the goals of holistic control, it is necessary to develop robust and appropriate control strategies that can range from simple, proportional, integral, and derivative (PID) controllers to complex control strategies such as model predictive control (MPC). Not only must the strategies be developed, but they must also be implemented on the actual processsing units used in industry and regularly maintained.

Smart World

As the world grows ever more interconnected and it becomes possible to control more from a simple cellphone, the topic of smart world will become more important. These area looks at how to design and implement the key concepts, such as modularity, real-time data analysis, and robustness, in different industrial and everyday areas. Currently, the department is interested in the following areas:

  • Smart Home: Smart home focuses on the integration of smart technology into assisting with everyday tasks. This is especially important for disabled people who may need assistance in performing certain tasks. In effect, smart technology can provide them with the tools to effectively complete these tasks.
  • Smart Factories and Industry 4.0: Similarly, the integration of smart technology into industrial processes will require changes and improvements in how the processes run. At the same time, these processes will be made more efficient, more environmentally friendly, and more productive.
  • Smart Grid: As the requirements for energy increase, the development and implmentation of smart grids will become important. Changes in demand and supply will need to be carefully monitored and assessed to make sure that the grid itself does not collapse.

Computational Linguistics and Natural Language Generation

As the number of computers proliferate and the need to generate grammatically correct sentences at runtime increases, the development and implementation of appropriate methods is required. This area is focused not only on the development of the required software, but also on the development of the grammars and dictionaries required for individual languages. The following projects are currently being pursued:

  • Grammar Engine: Creating the required software for a language-independent grammar processing engine that can take a grammar file that describes the grammar of a language and a dictionary file that contains the vocabulary with appropriate additional information for an arbitrary language and produce the required sentence at runtime. Currently, the focus is on describing Indo-European languages.
  • Grammar Describtions: Creating a grammar of a given language so that fixed rules cover as much of the grammar of the language as possible with minimal input from the user, for example, in German, it is possible to conjugate the majority of the verbs given the infinitive and the verb class. For the other verbs, only the three principle parts will be required. Minimal exceptions exist that can be directly handled.