Tutorial Presentations

Abstract: The current wave of machine learning and artificial inteligence solutions depend heavily on data. Data quality depends heavily on how well calibrated the sensors are which generate the data. In this tutorial, the speaker will start with the basics of sensor calibration. Then he will discuss about blind calibration and conditions that facilitate blind calibration. This will be followed by a review of some of the popular AI based techniques to calibrate sensors and some of the new AI-based techniques for blind-calibration of sensors developed by the speaker’s team. If you are responsible for installing sensors in your project or are concerned about the quality of the data coming from your sensors then this tutorial is for you. Blind calibration of sensors would be a major enabling solution in both Industry 4.0 as well as achieving sustainable development goals (SDGs).

Instructor:

• Amit Kumar Mishra, Aberystwyth University, UK

Abstract: Nowadays, advanced manufacturing and environmental monitoring strongly rely on artificial intelligence (AI) for improving both production processes and the efficacy of proposed solutions. Therefore, it is essential to gain insights into how deep-learning models can operate in challenging real-world scenarios. In these contexts, standard large-scale datasets cannot be employed for spotting anomalies, recognizing objects, or performing change detection, primarily due to the inherently different characteristics of the domains. The adaptation of these datasets to practical scenarios may also be costly. Furthermore, the lack of high-quality annotated data, power-constrained hardware, and privacy-preserving solutions could limit the usability of data-driven architectures, leading to the development of both hardware-and domain-specific solutions, with a great impact on the overall performance. This tutorial will shed light on current challenges in the field of image synthesis in both the industrial and environmental domain. Hands-on activities will spotlight the main characteristics of image generation models tailored for industrial scenarios, with a specific focus on addressing challenges posed by limited annotated data and analyzing the impact of optimized models on resource-constrained hardware.

Instructor:

• Pasquale Coscia, Universitá degli Studi di Milano

Abstract: Many engineering applications require to detect and to discriminate useful signals from noise or nuisance prior to measuring one or more quantities of interest. Some relevant examples of applications in which detection theory plays a crucial role are (but not limited to): digital communication systems, radars and sonars, Non-Destructive Testing (NDT) techniques, power quality event detectors, biomedical measurement systems. This tutorial provides an overview of classic detection theory solutions based on hypothesis testing. Starting from the classic Neyman-Pearson (NP) theorem (that can be applied when the probability density functions of each assumed hypothesis are completely known) the tutorial moves step-by-step towards the more realistic case in which the probability density functions of binary and multiple hypotheses are no longer completely known. Special attention will be devoted to composite hypothesis testing based on the Bayesian approach and to the Generalized Likelihood Ratio Test (GLRT). For each type of detection problem, a possible state-of-the-art solution shall be presented and clarified through application examples in the field of measurement science and technology. 

Instructor:

• David Macii, University of Trento

Abstract: Soft sensors are models that allow estimating the values of a variable based on other process information, without having to measure this variable directly. The main benefits of soft sensors are (1) they represent a low-cost alternative when compared to physical sensors, (2) they can work together with physical sensors, including to identify when they fail, (3) they allow implementation on existing devices, and (4) they provide real-time estimates, being an option for measurements where physical sensors depend on time-consuming analysis. In this tutorial, we are going to learn how to develop a data-driven soft sensor using Python taking into account data-driven techniques such as neural networks, decision trees, and other regression techniques. Besides the Python development, we will learn how to perform attribute selection and feature engineering aiming to improve the accuracy of the model. We will also understand how a neural network and a decision tree can be interpreted and translated to simpler programming form, such as Structured Text, therefore, transforming the generated Python models to be deployed in devices that does not run Python, such as industrial computers. Real data from two processes will be used in this hands-on tutorial, one from a mining process that involves a conveyor belt, motors, and time-displaced scales, and other from a large solar energy area that uses solar irradiance to estimate generated potency. 

Instructor:

• Prof. Dr. Gustavo Pessin, Full Researcher, Instituto Tecnológico Vale (ITV)

Abstract (Part 1 Theory): This tutorial is presented in two parts.  Part 1 presents the fundamental concepts and engineering design principles for the measurement of the gas parameters of pressure, temperature and concentration based on laser absorption spectroscopy. Methods for the extraction of the gas parameters from absorption line measurements are explained in detail with tuneable diode laser sources, quantum cascade lasers and frequency combs. These methods include harmonic ratio detection, acoustic signal detection and dual comb spectroscopy. The relative merits of near-IR or mid-IR operation in terms of cost, flexibility and sensitivity are discussed as well as the practical issues of referencing, calibration and signal-to-noise considerations. The various measurement systems that may be deployed in practice are reviewed, including open-path and multi-pass systems, fibre optic networks and recent developments in silicon photonic chips. The variety of system designs and measurement techniques presented in this tutorial provide the framework for understanding how laser-based gas measurements systems may be deployed in a wide variety of industrial and environmental settings including the harsh environments of combustion and emissions monitoring.  

Abstract (Part 2 Experimental): This tutorial builds on the learning experience gained from the principles and theory of tuneable diode laser spectroscopy (TDLS) presented by Professor George Stewart in Part 1 of the same topic.  Part 2 (here) addresses the practical application and deployment of fully engineered TDLS systems in extreme environments.  It covers system design, component and system characterisation, validation in the laboratory, deployment and use.  Particular issues, such as signal noise, signal to noise ratio optimisation, calibration and calibration-free approaches, and data-acquisition bottlenecks are addressed.  Case studies on fully deployed systems for measurement of methane and water vapour in operating solid oxide fuel cells and carbon dioxide and water vapour in the exhaust plumes of gas turbine (aero) engines under operational testing are presented.  System extension to achieve multi-beam, tomographic species imaging is also addressed, with particular focus on the complexities of upscaling data-acquisition systems and optical distribution.

Instructor: 

• Professor George Stewart, University of Strathclyde
   

Abstract: The purpose of the tutorial is based on idea of presenting topics related to sustainable development, to provide definitions and to indicate the main assumptions of principles of sustainable development in the modern world, providing specific examples of the contribution of Instrumentation and Measurements, such as sensorized buoys for sea water environmental control. 

A general illustration will be given on the sustainable development, a broad term to describe policies, projects and investments that provide benefits today without sacrificing environmental, social and personal health in the future. These policies, described as green, focus  mainly on limiting the impact of development on the environment, as well as extending their benefits on a wide cross section of human health and well-being, including reductions in pollution and environment-related disease, improved health outcomes and decreased stress.

The problem of marine monitoring, of interest for sustainable development will be illustrated. Coastal marine systems are particularly vulnerable to the effects of human activity due to industrial, tourist and urban development, consequently these ecosystems became a primary concern for learning more about the behaviour of the marine environment. It is essential to gather information on large enough spatial and time scales to assure effective monitoring and to be able to produce solutions that as far as possible reduce the negative impact of human activity on these ecosystems.

Instructors:

• Professor Emma ANGELINI, Politecnico di Torino
• Professor Luca LOMBARDO, Politecnico di Torino

Abstract: The growing number of elderly people, the decrease in resources available for healthcare, together with the pathologies (COVID 19) that make contact between doctor and patient dangerous are now a strong signal for a change in the model of medicine currently used. But which new model? Artificial Intelligence (AI) is now the basis of  every change in  life today and promises to be a revolution for both patient care and the healthcare system. But the real challenge is to find a useful model to prevent the growing number of operations required for the elderly: more people but with fewer health problems. Generative Artificial Intelligence may represent the new frontier for creating new data and models for future healthcare systems. These systems can be used to implement and interpret results and generate personalized medicine. Digital twins can be used to develop models to understand the effects of treatment choices in patients using their digital twin. Virtual hospitals will be used to monitor multiple patients in their own houses using AI wearable technology. But a common question is: how these systems can be trusted in medicine? What can we say about the Accuracy, Precision, Sensitivity, ROC curve and F-score of the data generated by a generative A. I.? The second part of the tutorial will make a contribution to these analyses using a metric typical of the Instrumentation and Measurement world.

Instructor:

• Prof. Eros Pasero, Neuronica lab, DET, Politecnico of Turin, Italy

Abstract: The precision agriculture (PA) combines technologies and practices that assure the optimization of the operations associated agricultural production through specific farm management. 

Regarding the employed technologies distributed smart sensing systems characterized by fixed and mobile nodes (based on remote sensing  and Unnamed Aerial Vehicle (UAV))) are used to turn the farming operations into data, and to optimize the future operation based on data driven models. Edge and cloud computing platforms that are capable to run AI/ML algorithms may contribute to help on human decisions. 

The tutorial focusses on digital transformation of the agriculture in the context of heavily uncertainty associated with climate change. The IoT ecosystem technologies for precision agriculture will be discussed including multimodal sensing and artificial intelligence. In-situ and remote sensing are considered special attention being granted to the soil characteristics monitoring (moisture and macronutrients concentration). The agriculture UAV imagery and satellite  imagery solutions as so as the relation between the data coming from the in-situ distributed smart sensors and acquired images using multispectral and thermographic camera and imagery techniques will be part of the presentation. AI multiple sources data driven models for an increased crops quality through the optimization of farming operations as so as examples of data driven models for smart irrigation and nutrients will be discussed. 

Instructors:

• Dr. Octavian Postolache, Instituto de Telecomunicações, Lisboa, Portugal
   

Abstract: Condition monitoring (CM) of components and processes using machine learning (ML) is one of the central promises of Industry 4.0. Many successful examples have been demonstrated under laboratory conditions. However, the transfer to actual industrial application is proving difficult. The main challenge remaining is the data quality required for developing a meaningful and robust ML model: in industrial applications, most data represent the “good” condition, while samples for different fault scenarios are typically scarce. Furthermore, comprehensive training data are required covering all relevant circumstances to allow successful CM under changing environmental conditions and other causes of domain shift. Even if extensive data are available, most effort is spent on their organization to delete outliers, ensure correct labeling etc. The tutorial will address these issues with two main approaches. The first is a checklist to guide users through the complete process of an ML project, starting with project, measurement, and data planning proceeding to data acquisition, checking and pre-processing up to finally building and validating the ML model. This checklist specifically supports users with little experience in ML to be successful. The second approach is classical process optimization based on insights gained using explainable machine learning methods.

Instructors:

• Prof. Dr. Andreas Schütze, Saarland University 
• Tizian Schneider, Centre for Mechatronics and Automation Technology

Abstract: The acquisition of information about physical quantities by means of sensors historically fostered the interpretation of measurement as a merely experimental activity. Conversely, measurement is a complex activity, far more complex than suitably connecting and reading an instrument. Indeed, measurement always requires descriptive activities to be performed prior of the execution of empirical activities to ensure both the correct implementation of the experiments and the interpretation of the obtained information.  
In this tutorial, the basis concepts involved in any measurement are presented and discussed. The tutorial contents support a methodologically correct development of any measurement, regardless the kind of involved quantities (either physical or non-physical) or the field of application.  

At the end of this tutorial the attendee will be able to answer such questions as: Which informative empirical processes can be considered measurements? How do I identify an adequate model for a given measurement? How do I estimate and express the quantity of information I can obtain through measurement?  

Instructor:

• Dario Petri, University of Trento, Italy

Abstract: Autonomous systems are nowadays having an undisputed pervasiveness in the modern society. Autonomous driving cars as well as applications of service robots (e.g. cleaning robots, companion robots, intelligent healthcare solutions, tour guided systems) are becoming more and more popular and a general acceptance is now developing around such systems in the modern societies. Nonetheless, one of the major problems in building such applications relies on the capability of autonomous systems to understand their surroundings and then plan proper counteractions. The most popular solutions, which are gaining more and more attention, rely on artificial intelligence and deep learning as a means to perceive the structured and complex natural environment. Nonetheless, besides the importance of such complex tools, classical concept of metrology, such as standard uncertainty, accuracy and precision, are still unavoidable for a clear and effective understanding of modern autonomous systems applications. 

At the end of this tutorial the attendee will be able to answer such questions as: what are the tools and the methods of major relevance for autonomous systems applications? How do concepts as uncertainty map in the autonomous systems realm? 

Instructor: 

• Daniele Fontanelli, University of Trento - Italy

Abstract: Metal pipes are used in building construction like mechanical supports, plumbing, ducting, heating, harness and electrical engineering. Power systems and RF equipment, radars have metal pipes. Since the performances are affected by change in the internal structure of the pipe such as clogging and deformation, we must know conditions like defects inside the pipe without disbanding it.

We propose two methods that use millimeter-wave band signals applied to both the ends of the pipe. One method is to design two high directive antennas working at the millimeter-wave band and let the antennas positioned at both the ends and send and get the signal between the two sides. This has benefits of being relatively free from impedance mismatch at the two ends due to non-contact, higher resolution obtained at high directivity and small dispersion by the CW signal propagating smoothly The other method is to put special metal caps at both the ends of the pipe and let the signal enter the pipe and move along the channel. The special metal cap has a monopole in front of a slot coupling the millimeter-wave signal from the monopole into the inside of the pipe, which is applicable to practices requiring pipes.

Instructors:

• Sungtek Kahng, Incheon Nat'l University
• Changhyeong Lee, Corning Technology Center Korea

Abstract: This tutorial is designed for industry professionals and researchers interested in quantifying uncertainties in complex systems. Uncertainty propagation allows us to provide estimates of the quantity of interest that are not just single numbers (point estimates) but provide confidence intervals or some other statistical measure. It is a fundamental component of any measurement system. We will provide a practical overview of uncertainty propagation followed by a number of computational examples in Matlab and/or Python. We will start first with the uncertainty propagation through simpler functions such as functions used to compute oxygen saturation in pulse oximeters. We will then show an example of uncertainty propagation through a circuit model of a simple differential amplifier. We will also present a complex model of a blood pressure measurement device that includes the pressure sensor, conditioning electronics and the estimation algorithm that can be based on signal processing or machine learning. In that case, uncertainty propagation allows us to estimate confidence intervals of systolic and diastolic blood pressure. The tutorial is built upon and adapted from the recently published book by Dr. Bolic, "Pervasive Cardiovascular and Respiratory Monitoring Devices: Model-Based Design." 

Instructor: 

• Dr. Miodrag Bolic, Ottawa in Canada

Abstract: The human population in the world is aging fast. More than 15% of the human population will be above 65 years by 2050.  With the increase in life expectancy and aging population, the world may be witnessing health and socio-economic burdens.  As the citizens age, they may suffer from neuro-cognitive impairments leading to dementia, and possible postural instabilities.  Along with frailty, falls, and hospitalization due to falls may become more prevalent. In order not to burden the medical system, it may be necessary to have a reliable privacy-preserving monitoring system that would require no compliance from the elderly.  This could lead to a continuous monitoring of their activities and vital signs.  Further, such a system can also lead to senior citizens aging well in their homes which would help them to retain their independence and avoid depression.  Radar sensors offer privacy-preserving remote monitoring of the elderly.  Using appropriate advanced signal processing techniques, it would be possible to continuously and reliably monitor elderly citizens for their well-being in the comfort of their homes. The time has come to understand and evaluate the various techniques and understand the advantages and disadvantages that come along with them.  This tutorial will motivate researchers in the field of instrumentation and measurements to explore this fascinating area that encompasses multiple areas of engineering. 

Instructors:

• Prof. Sreeraman Rajan, Carleton University, Ottawa, Canada 
• Ms. Ankita Dey, Carleton University, Ottawa, Canada 

Abstract: There has been an astronomical increase in the number of technical paper submissions in the past decade.  Some of the reasons include: 

• pressure to publish, as the success indicator, for promotion and professional advancement,
• universities moving away from the traditional M.S. Theses and Ph.D. Dissertations to instead a compilation of several peer-reviewed journal papers,
• creation of new journals, and
• the open-access publishing “economy”.

Journals are ranked according to certain “indicators” that may or may not be objective. Everyone wants to publish in the highest-ranking journals exasperating the situation for some. However, we wish to think that “Quality” is the number one “indicator” of a journal. “Quality” is not a “measurable” and is difficult to define. However, there are ways by which to positively influence the “Quality” of a journal beyond those indicators.

Instructor:

• Reza Zoughi, at Iowa State University

Abstract: Prepared by a top expert in the field with many developments, real world applications and patents, this lecture will be an invaluable resource for physicists, electronics engineers, teachers, students, technicians and anyone working in the field of sensors. In this presentation, the advantages and possibilities of fiber optic sensors in research and industry will be presented, including two technologies, plastic optical fiber (POF) and the silica fiber. In addition to presenting various technologies that use POF as a sensor, this tutorial will also focus on another type of technology, Fiber Bragg Grating (FBG). FBGs can be found in many industrial applications and we will present our experiences in applying FBGs in many types of sensors for the power industry.

The tutorial will start with the theory of fiber optic sensors using both POF and FBG. It will then cover various practical applications of sensors, including successful field applications developed by our lab in areas such as oil & gas, biotechnology and electrical energy.

The following topics will be presented and discussed in the lecture: Fiber optic sensing technologies; temperature sensing; strain and force sensing; refractive index sensing; high voltage switch monitoring; current and voltage sensing; gas sensing; chemical and biological sensing; oil leak detection; high voltage and high current sensing; gas flow velocity sensing.

Instructor:

• Marcelo Martins Werneck, Universidade Federal do Rio de Janeiro