2021-08-05T15:42:26Z EPrints http://miis.maths.ox.ac.uk/images/sitelogo.gif http://miis.maths.ox.ac.uk/miis/ 691 20 1 2019-10-24T18:24:46Z 2019-10-24T18:29:04Z http://miis.maths.ox.ac.uk/miis/id/eprint/758 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/758 2019-10-24T18:24:46Z

Captured Carbon supply power to Eirgrid by using many small power generators (Single Generators) as if their combined power were a unique large power station (Virtual Generator). When Eirgrid order power from a power station, they assume that power output increases linearly over a period of time (ramp up time), after which the station generates power at full capacity. The main question we tackle is: How can we control the start up time of many small generators in order to get as close as possible to a linear combined ramp behaviour? In this report, we provide 5 different approaches (gradient descent, mixed integer linear programming, bin packing, systematic adjustment and ramp tracking simulation) to the problem using sample data from 20 single generators provided by Captured Carbon. We also discuss briefly the “equity” challenge, consisting of the possibility to have optimal solutions for which not always the same single generator starts first. We conclude with some suggestions for future work.

Paul Beagon Miguel D. Bustamante Mel T. Devine Susan Fennell Jonathan Grant-Peters Cameron Hall Roisin Hill Taulant Kerci Gary O’Keefe 2019-05-11T14:09:37Z 2019-05-11T14:09:37Z http://miis.maths.ox.ac.uk/miis/id/eprint/757 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/757 2019-05-11T14:09:37Z

Globally there is a huge market for herbicides. Syngenta and its competitors spend large amounts of money in trying to develop new herbicides which are highly effective (kills the plants that the farmer wants to get rid of), selective (does not harm the crops that the farmer is trying to protect), safe (not harmful to humans or the environment) and cheap to produce.
Every year Syngenta chemists develop around one thousand new compounds that could be used as herbicides. Accurately evaluating the efficacy, selectivity, safety and ease of production of each one of these compounds would be extremely expensive. Initially synthesizing compounds even in tiny quantities requires a lot of work. Different compounds will then require different levels of dosage, may target certain species of plants better than others, may work more effectively in conjugation with different solvents etc. A huge amount of time and money could be spent optimizing the application of a particular compound, which then turns out to be very poor compared to existing herbicide products on the market.
In order to attempt to select only the best performing compounds in a cost effective way, Syngenta use a screening cascade. This process begins by testing all of the candidate compounds in laboratory experiments, referred to as assays, to see if they can (e.g.) target a particular enzyme, or penetrate a leaf. These lab experiments only require a tiny quantity of each compound to be synthesized. Compounds which perform poorly in these first experiments are discarded from the trial. We call this the first screen.
Compounds which pass this first round are then passed to a second round in which a small amount of the compound is applied to several small pots containing a few different species of plant. Compounds which perform poorly in this second experiment are discarded from the trial process. We call this the second screen.
In subsequent screens larger quantities of the compounds are used in the experiments, which makes them a lot more expensive. By using more compounds the candidate herbicides can be tested on a larger range or species, at different dosages, in conjunction with various different conditions, and with less sampling error.
As the screening process progresses, the screens become steadily more rigorous and consequently more expensive. At the same time the number of compounds remaining in the trial goes down. Finally a small number of compounds are taken to the final level, called the field trial. In this trial the compounds are applied outdoors in the way they would be used if they were eventually developed into a commercial product.

James Hook Kamil Kulesza P. Morawiecki Eddie Wilson 2019-05-11T14:02:32Z 2019-05-11T14:02:32Z http://miis.maths.ox.ac.uk/miis/id/eprint/753 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/753 2019-05-11T14:02:32Z

This report details the work carried out by the Study Group on workflow modelling of con- struction projects.
Data on the progress of about a hundred projects over a single five-year planning period were provided by Heathrow Airport (the client) and their four Tier 1 construction contrac- tors. These data are mapped and analysed. Several unusual features are discovered. For example, most projects undergo several tens of adjustments in their scope and price such that while most projects are technically completed under budget, the price and duration is significantly higher than originally planned.
The main question addressed was whether an optimised scheduling of the project would lead to decreased costs and more rapid completion.
First, a machine learning approach is used to gain insight onto which factors are most significant in predicting the final cost and duration of each project. If more data were available, these methods could be further exploited to allow for predictions to be made on which projects are likely to over-run or go over budget and to examine connections between projects at the subcontractor level.
In addition to the data-centric approach, a complementary mathematical model was de- veloped to gain a better understanding of the effect of resource constraints on cost and price extension due to resource competition of concurrent projects, ignoring the confound- ing effect of scope creep seen in the data. The model takes the form of a discrete time stochastic simulation, whose parameters are fit to the existing data. Tentative conclusions from the model indicate that better outcomes can be achieved by spreading out project start dates, and by prioritising completion of smaller projects.
While more data is needed to validate the model, the results suggested that gains can be made if more thoughtful scheduling of projects is implemented, and also if the prioritisation of projects is monitored and adjusted intelligently.
Our major recommendation to Heathrow Airport is to collect or retrieve more data, as outlined in the report, so that both models can be made more realistic and useful. This would allow Heathrow Airport and their contractors to develop and test strategies to make the system more efficient, ultimately saving time and money.

E. Murphy A. Champneys Hanan Batarfi Edmund Barter Chris Budd Ran Dong Jan Foniok Kamil Kulesza Andrew Lacey Xiaodong Li Francisco Rodrigues Alex Wendland Ambrose Yim 2019-05-11T14:02:04Z 2019-05-11T14:02:04Z http://miis.maths.ox.ac.uk/miis/id/eprint/755 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/755 2019-05-11T14:02:04Z

Computer vision systems can be used to measure pedestrian flow rates, occupancy levels and queue times. It is difficult to assess the accuracy of such methods because the ground truth can be difficult to establish. Human counting is equally prone to error, even when using video recordings with no time constraints and the support of sophisticated software.
In this report, we consider how errors may arise directly from the images recorded by the cameras, due to both occlusion of people and image distortion due to a fisheye lens. We also develop a statistical model of human counting errors and attempt to estimate human accuracy from data. Finally, we attempt to relate human and computer accuracy on the basis of simplifying statistical approximations.

Jonathan Ward Giancarlo Antonucci Edmund Barter Frank Brooks-Tyreman Colm Connaughton Michael Coughlan Ronja Kuhne Marcus Kaiser Victor Wang 2019-05-11T14:01:46Z 2019-05-11T14:01:46Z http://miis.maths.ox.ac.uk/miis/id/eprint/754 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/754 2019-05-11T14:01:46Z

This report addresses the construction of carbon fibre wing boxes and the problems as- sociated with using carbon fibre sheets rather than individual carbon fibre tapes. In the case that the wing boxes are developable surfaces the lay up of carbon fibre sheets is straightforward, since the fibres can follow the contours of the surface without any need for shearing or extension of the fibres. To further expand the potential design space for the wing boxes, this report investigates the lay up of sheets over non-developable surfaces where some shearing of the sheet is required to achieve the desired results. In this report, three analytical approaches are considered, driven by the results from numerical studies on different surface geometries. Each of the approaches offers insights as to the type of geometric perturbations achievable when constrained by a maximum shear angle.

David Barton H. Ockendon Bernard Piette Robert Whittaker Eduard Campillo-Funollet Mike Jeffrey John Ockendon E. Walsh 2019-05-11T13:59:20Z 2019-05-11T13:59:20Z http://miis.maths.ox.ac.uk/miis/id/eprint/752 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/752 2019-05-11T13:59:20Z

A mathematical model is sought for the flow of nutrients in the Hydrosac⃝c growing module being developed by Phytoponics. The basic operation involves long fluid-filled bags with periodic growing zones from which root systems emerge into the bulk fluid. The system is periodically perturbed via two main processes: partial drainage and refilling of each bag with nutrient infused water, with inlet and outlet at opposite ends of the bag; and a more violent oxygenation of the water through bubbles that rise from the pores of an aeration tube that runs underneath the central long axis of the bag.
The aim of the modelling is to determine the key parameters and fluid regimes underlying the nutrient mixing process, to ensure that required nutrient levels are maintained through- out the root zones, and to enable optimal scheduling of the nutrient and bubble flow.
Simple experiments were performed via the injection of dye into an operating Hydrosac⃝c that contained semi-mature plants. This enabled a basic understanding of the time and lengthscales of nutrient flow, and also the extent to which mixing occurs in different zones within the bag. Four different flow regimes are identified. At the scale of a single root, a Stokes-flow approximation may be used. At the scale of the individual plant, a so-called Brinkman flow regime may be employed which is describes a transition between slow porous- medium flow and fast channel flow. These equations may be homogenised into a 1D model that can be used to estimate the macro-scale flow of nutrients along the length of the bag.
A shear flow model is used to predict the extent to which this flow permeates into regions dominated by plant roots. This leads to the requirement to model the bubble-driven flow within a bag cross-section containing a plant. Simplified two-phase flow equations are de- rived and solved within the software COMSOL. The results suggest that the bubble flow is sufficient to drive recirculating flow, which is also found to be consistent with previous literature.
The overall conclusion is that both the periodic flow of nutrients and the aeration are re- quired in order to enable even nutrient spread in the Hydrosac⃝c . Wave effects can be ignored, as can the effect of stagnated nutrient diffusion. The longitudinal nutrient flow enables the whole sack to be reached on the time scale of several cycles of the main inlet flow, while the recirculation from the bubble flow enables enables nutrients to spread within the plant roots. Nevertheless, regions of stagnation can occur via this process near any sharp corners of the bag.
It is recommend that the various analyses are combined into a a reduced-order mathemat- ical model that can be used to optimise the dynamic operation of the Hydrosac⃝c , which can also be adaptable to other geometries and growing conditions.

A. Champneys Graham Benham Stephen Cowley Zoe Dennison Matthew Griffith Alissa Kamilova Atilla Kovács Andrew Lacey Scott Marquis P. Morawiecki John Ockendon Rahil Sachak-Patwa Colin Please Hayley Wragg 2019-01-21T22:44:44Z 2019-01-21T22:44:44Z http://miis.maths.ox.ac.uk/miis/id/eprint/751 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/751 2019-01-21T22:44:44Z

Flat fan nozzles atomize crop protection products, breaking them into droplets. Droplet size matters - smaller droplets give better perfor- mance, but very small droplets drift. We want to use mathematical models to better understand how liquid properties affect droplet size.
There are three types of breakup: wavy sheet, perforation, and rim. In wavy sheet breakup, increasing viscosity or surface tension increases droplet size. To investigate further, we carry out direct numerical simulations of jet breakup, which show that suface tension has little effect, but increasing viscosity leads to fewer droplets. Decreasing the jet velocity also results in fewer droplets, with a wider size distribution.
Each type of breakup involves primary breakup into cylinders of fluid, then secondary breakup into droplets. We thus consider the breakup of a cylinder of fluid. Direct numerical simulations suggest that within the tested parameter range viscosity has little impact on droplet size, however it does influence the timescale on which the instability evolves considerably. Linear stability analysis suggests that increasing viscosity increases the wavelength of the most unstable mode, which we expect leads to larger droplets, and that it reduces the rate of breakup.
Perforations - holes in the sheet - also lead to breakup. We find how the length fraction of the sheet that is void changes with time.
After breakup, the droplets continue to evolve. We develop a model, based on a transport equation, for this process. A key parameter is the breakup rate constant - larger values lead to more breakup, fewer large droplets, and a narrower size distribution.
Together, these mathematical approaches improve our understanding of how droplets form, and can be used to guide experimental work.

Sami Al-Izzi Francesco Broggi R. Cimpeanu Lloyd Connellan Robert Gowers Mat Hunt Davin Lunz Matthew Moore John Ockendon Victoria Pereira James Sprittles Rachael Warrington 2019-01-21T22:43:54Z 2019-01-21T22:43:54Z http://miis.maths.ox.ac.uk/miis/id/eprint/750 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/750 2019-01-21T22:43:54Z

A Railway Traffic Management problem can be defined as forecasting fu- ture progression of trains, identifying conflicts where two or more trains compete for available infrastructure, investigating options for resolution of conflicts, re-planning train schedules to minimise the impact on sy- stem performance. Performance management of complex networks is a problem common to a number of industries and applications. There has been much work over many decades on modelling the generation and optimisation of railway timetables. Much of this focuses on relatively simple railways and services and is therefore quite straightforward. Main line railways have a number of features that introduce significant com- plexity. Traditionally the problem of re-planning a timetable in near real time to manage and recover from service perturbations and disruption is simplified to help arrive at a solution in an acceptable amount of time, but this then can have unintended consequences which can amplify rat- her than reduce the disruption in the network. Resonate are interested in looking at different strategies / models / techniques for dealing with the problem, the likely strengths and risks of these, and how they might be adapted to improve existing solutions. The study group participants undertook a brief survey of recent literature on modelling train delays and found machine learning approaches, network models and a statisti- cal approach to defining the efficiency of a station in dissipating delays which are worthy of further consideration. We then explored total of nine modelling approaches during the study group. The approaches fell broadly into two groups: those that sought to understand the pro- pagation of delays (Approaches 1 to 6) and those that sought to offer strategies for minimising delays (Approaches 8 and 9). Approach 7 pro- poses a way of understanding the propagation of delays and using that to evaluate candidate policy decisions. There are a number of promising approaches here which provide useful lines of enquiry, many suitable for expansion beyond the simple railways modelled, to include variable train speeds, junctions and intersections, temporal differences in usage, such as tidal flows in and out of cities, and resource constraints.

Martine Barons Chris Bick Marco Caselli Antoine Choffrut Vinh Doan Payman Eslami Jan Foniok Laura Guzman-Rincon Roger Hill Emily Kawabata Giovanni Mizzi Chris Norman Norbert Peyerimhoff K. Piwarska C.P. Please Caoimhe Rooney Sofia Trejo Luke Whincop Robert Whittaker Jessica Williams Yuanwei Xu 2019-01-21T22:43:45Z 2019-01-21T22:43:45Z http://miis.maths.ox.ac.uk/miis/id/eprint/749 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/749 2019-01-21T22:43:45Z

By using a high-speed camera, researchers at MIT in 2014 where able to recover human speech from videos of minute vibrations of objects in a room. For example, in one experiment a 2,200fps camera was positioned outside a room behind sound-proof glass, videoing an empty crisp packet on the floor inside the room, while a researcher shouted “Mary had a little lamb” at the crisp packet. By detecting minute oscillations of the crisp packet of 1 μm (0.001 mm), and using hours of computer processing, a ten second audio clip could be produced that was recognisably “Mary had a little lamb” in an American accent.
The purpose of this study group was to investigate whether this tech- nique could be used in practice, with emphasis on the recovery of intel- ligible speech from a video feed of a room. During the week, the group investigated several aspects of the problem, including:
• how much an object vibrates due to sound;
• what can be done to maximize the vibration;
• how the MIT technique detects minute vibrations in videos; • what affects the quality of the resulting recording; and
• how good a recording is needed for intelligible speech.
It was discovered the MIT experiments would not have recovered intel- ligible speech from an ordinary conversation; their success depended on loud sounds and prior knowledge of “Mary had a little lamb”. Camera vibrations were also ignored by MIT; these are expected to be signifi- cant, but the technique could be adapted to be resilient to them. Other possibilities for enhancing their technique, by exploiting resonances or reflections, are discussed in the report. A high-speed low-noise cam- era is essential, and any existing video footage (such as from CCTV) is unlikely to be of sufficient quality. Further experiments with high-end high-speed cameras are needed to assess the feasibility of the technique in practice.

Ed Brambley Helen Fletcher Roger Hill Ifan Johnston Jessie Liu Robert MacKay J. Mathews John Ockendon Bernard Piette 2019-01-21T22:43:33Z 2019-01-21T22:43:33Z http://miis.maths.ox.ac.uk/miis/id/eprint/748 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/748 2019-01-21T22:43:33Z

Gas chromatography (GC) is a popular tool for chemical analysis. Some samples are so complex that a single column does not have enough power to separate all of the analytes. In this instance a higher resolution GC method, known as comprehensive two-dimensional gas chromatography (GCxGC), is used. DSTL want to be able to use data from GCxGC to attribute samples to a particular region or cultivar. However, the nature of the data means that several difficulties must be overcome before being able to do this: noise from sample, peak mis-alignment, and low quantity of samples. In this report, we investigate several methods to overcome such difficulties, and then classify the data. We are very successful in telling apart blanks from seeds, but obtain limited success when trying to classify between seeds. The method that shows the most promise is k-Nearest Neighbours classification by Wasserstein distance. However, this is still quite sensitive to the noise created by the solvent in the sample. Thus, we suggest that more blank runs be obtained, so that the ‘ground truth’ behaviour of the solvent is better understood, allowing us to remove the effect of the solvent from seed data. We also hope that the methods explored here will be more successful on the full raw data than they were on the limited ‘peaks’ data available to us for the purpose of this study.

Matteo Croci P. Morawiecki John Prater Valentin Sulzer Florian Theil 2019-01-21T22:43:22Z 2019-01-21T22:43:22Z http://miis.maths.ox.ac.uk/miis/id/eprint/747 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/747 2019-01-21T22:43:22Z

In recent years, worldwide terrorist strategy has changed from long term plan- ning and high impact attacks, to relatively short-term planning, ’marauding’ attacks. In other words, terrorist attacks from 2010 to present have often consisted of a small number of terrorists in densely populated public spaces actively searching out people to maim and kill with weapons that are har- der to regulate, such as knives and vehicles, with little regard for their own survival. This begs the question, therefore, of how we mitigate against these ’marauding terrorists’. If we assume 3 different types of actors: Public, Ter- rorists, and Responders, and assume the strategy of the Terrorists is to kill as many people as possible before they themselves are killed, how can the strategies of the Public and Responders (police, bouncers etc.) be optimised to minimise loss of life?
In this report, the problem at hand and important information are compiled before 3 approaches to model a terrorist attack in a public space are considered - a Particle Model, a Discrete Network Model, and a Game Simulation model. Whilst this is by no means a complete list of possible models of a terrorist attack, these were believed to be models that could be developed the most in a week at the ESGI130 at the University of Warwick. For each model type, we first consider their assumptions and suitability to the problem, then model the scenario. Finally, we consider possible extensions to each model, and also how they may be used to evaluate the most effective strategies of the Public and Responders given the information available to them at any one time.

Maxim Zyskin Adam Nixon Maria Dolniak Lewis Church Ian Roper K. Piwarska Ben Pooley Simon Baker P. Hjorth Thomas Roy 2019-01-21T22:42:42Z 2019-01-21T22:42:42Z http://miis.maths.ox.ac.uk/miis/id/eprint/746 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/746 2019-01-21T22:42:42Z

This report contains exploratory data analysis of rainfall and Water Resource Sufficiency Index (WRSI) data provided by African Risk Ca- pacity (ARC). The purpose is to assess the predictability of droughts in Africa. We assess the appropriateness of the historical WRSI bench- marks set by ARC members compared to the observed WRSI values for different regions. We conclude that the benchmarks are broadly sensible. We then compare a number of linear time series models based on their ability to fit and forecast the WRSI time series. We conclude that sim- pler models like Simple Moving Average and Moving Median are more appropriate than more sophisticated models containing trends and sea- sonality like Holt Winter and TBATS. We also investigate the use of the SARIMA and TBATS models to forecast the seasonal patterns observed in rainfall data and conclude that both models can generate structured forecasts that reflect seasonal variability. The statistical evidence how- ever favoured TBATS over SARIMA. Attempts to measure the influence of the El Nin ̃o-Southern Oscillation on rainfall levels are inconclusive for the areas studied. Finally we perform a simple application of univariate Extreme Value Theory to rainfall data and conclude that further inves- tigation is necessary to understand how the catastrophic famine that affected Ethiopia in the early 1980’s would be reflected in the data if a similar event were to reoccur today.

Colm Connaughton John Herman Adam Johansen Emily Kawabata Robert Kerr Michael Pegg Jeremy Reizenstein Piotr Sakrajda Nick Tawn Luke Whincop 2018-07-22T17:46:44Z 2018-07-22T17:46:44Z http://miis.maths.ox.ac.uk/miis/id/eprint/745 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/745 2018-07-22T17:46:44Z

The non-profit organisation Florinhas do Vouga support homeless people by providing them with basic necessities such as food, as well as helping them address long term issues, such as mental health problems. This aid is usually provided whilst a homeless person is on the street or in some form temporary accommodation. The Housing First approach is to, instead, first provide a homeless person with their own home, and then, in this more stable setting, provide further support. The social benefits of giving homeless people the opportunity to live in a house are clear. To the individual it brings stability and makes it much easier to find a job, and access other forms of provision, such as healthcare and drug rehabilitation programs. To society at large it reduces the number of people living on the street, potentially increasing tourism and economic productivity in a city’s commercial centre.

The problem is to compare the costs incurred by both the Usual (U) and Housing First (HF) models and determine whether Housing First is a financially viable alternative. The housing status of each homeless person is modelled (for example if they are on the street or in a shelter) and this is used to calculate the costs of each approach. In the simplest model, the state of each person is assumed to be constant with time. The changing state of the homeless population is then modelled (for example how many homeless people are on the streets as opposed to in temporary accommodation or a house at any one time). This is done using a Markov chain approach, and also by direct simulations which incorporate more complex constraints on the state of the homeless population. The different mathematical models provide evidence that the financial costs of Housing First are less than those of the Usual model, especially in the long term (a 36 month period). Recommendations for further work are also given.

A. Avdzhieva Ben Sloman Cristina Requejo Rachel Philip 2018-07-22T17:46:41Z 2018-07-22T17:46:41Z http://miis.maths.ox.ac.uk/miis/id/eprint/744 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/744 2018-07-22T17:46:41Z Gnord Maypaokha Helena Alvelos Joaquim Correia Marina Andrade Mijiddorj Renchin-Ochir Xaivang Xiongshieng 2018-07-22T17:46:37Z 2018-07-22T17:46:37Z http://miis.maths.ox.ac.uk/miis/id/eprint/743 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/743 2018-07-22T17:46:37Z

At the 127th European Study Group with Industry an energy sector com- pany proposed an industrial challenge that consisted on the asset acquisi- tion planning for its liquefied petroleum gas (LPG) cylinder business, one of the most recent business areas in the company. This business area is still in a growing phase and to satisfy the market needs and assure a sustainable growth a very tight control of the main assets, the LPG cylinders, is of paramount importance. Therefore, a detailed planning of all the as- sets acquisition is required, taking into consideration several variables: sales growth rate, seasonality, cylinder rotation and corresponding return rate to the filling plant.
The challenge was to develop a model for the assets acquisition planning. In order to tackle this challenge, it was necessary to forecast the demand. For that purpose, time series techniques were used, in particular, moving averages and exponential smoothing. The results show that the seasonality does not explain all the variation of the demand, therefore it is necessary to use a model that would consider other possible explanatory variables.
According to several authors, gas consumption may be influenced by several aspects, such as, atmospheric temperatures, heliophany (a measure of the day luminosity), wind, relative humidity, rains, minimum and maximum temperatures, demand in previous periods, and prices.
The forecast of bottled propane gas sales and return rate was also addressed through multivariate linear regression. Regression models for the monthly number of bottles of types A and B were obtained, having pre- sented good percentages of explained variability with the variables under study.
The main goal of the challenge, the acquisition plan, was addressed u ing inventory models with reverse logistics. Several deterministic approaches have been considered to enable different aspects in the framework. A new inventory model has been developed to contemplate the three possible des- tinations of returned bottles: cleaning, requalification, or disposal. The models were implemented in Excel and can be tested, using PRIO estimates of holding costs and fixed setup costs, and the forecasts of sales and return rate computed previously.

Aldina Correia Ana Sapata Claudio Henriques Isabel Cristina Lopes Eliana Costa e Silva Fabio Henriques Magda Monteiro Mariana Pinto Rui Borges Lopes 2018-07-22T17:46:32Z 2018-07-22T17:46:32Z http://miis.maths.ox.ac.uk/miis/id/eprint/742 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/742 2018-07-22T17:46:32Z

Eucalyptus productivity is strongly related with climate and soil types of the area where it is planted. To accurately assess the potential productivity of that species in Portugal, plantations were monitored at different management units compartments (MUC) at several locations all over Portugal. Certain indices of productivity of the Eucalyptus at each MUC were recorded, as well as the type of climate and soil characteristics of the region. Both climate and soil, factors that affects Eucalyptus grow, were classified in in ten classes 1, 2, . . . , 10 of expected growing productivity for the Euca- lyptus. Thus, every MUC belongs to a unique pair (c, s), with 1 c 10 and 1 s 10 indicating the type of climate and the type of soil of the region where MUC is located, respectively, and it is expected that to have high (low) productivity indices when c and s are both close to 10 (1).
The aim of this work is to identify regions that have similar productivity levels based on the classifications of soil and climate types and to check if the available data provided by RAIZ show that those factors affect the productivity indices.
During the 5-days ESGI this team worked on the datasets provided by RAIZ, presented an update for the existing MAI productivity chart and developed new clusters for the Density, Yeld and Consumption productivity indices. The definition of some quality measures for the clusters, allowed to compare the different approaches and also point out some fragilities on the datasets. Indeed, a review of classification regarding climate and/or soil characteristics is suggested, as well as the need of a bigger sample for the Density, Yeld and Consumption productivity indices in order to get more reliable outputs.

A. Araújo Alexandra Gavina Ana Tavares J. Orestes Cerdeira Jorge Santos Manuel Cruz Silvia Barbeiro 2018-07-22T16:33:49Z 2018-07-22T16:33:49Z http://miis.maths.ox.ac.uk/miis/id/eprint/740 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/740 2018-07-22T16:33:49Z

EDP Group is an Energy Solutions Operator which operates in the business areas of generation, supply and distribution of electricity and supply and distribution of gas.

The challenge proposed by EDP consists in simulating electricity prices not only for risk measures purposes but also for scenario analysis in terms of pricing and strategy. Data concerning hourly electricity prices from 2008 to 2016 was provided.

Numerous methods to deal with Electricity Price Forecasting (EPF) have been proposed and can be classified as: (i) multi-agent models, (ii) fundamental models, (iii) reduced-form models, (iv) statistical models and (v) computational intelligence models. A recent an exaustive review is presentes in [13].

During this study group different promising Statistical techniques were propose by the study group contributors: ARIMA, sARIMA, Longitudinal Models, Generalized Linear Models and Vector Autoregressive Models. In this report a GLM and a vector autoregressive model are presented and their predictive power is discussed.

In the GLM framework two different transformations were consider and for both the season of the year, month or winter/summer period revealed significant explanatory variables in the different estimated models.
On the other hand, the multivariate approach using VAR considering as exogenous variables the meteorologic season and the type of day yield a multivariate model that explains the intra-day and intra-hour dynamics of the hourly prices. Although the forecast do not exactly replicate the real price they are quite similar.

In both of the approaches here reported a more extensive work would certainly improve the proposed models.
In conclusions, EPF is a growing area that groups multiple different approaches that can be applied. In fact, other approaches from multi-agent models, fundamental models, reduced-form models and computational intelligence models, also present a great space for EPF.

Ana Borges Andreia Monteiro Eliana Costa e Silva Marina Andrade M. Filomena Teodoro Raquel Menezes 2018-07-22T16:33:45Z 2018-07-22T16:33:45Z http://miis.maths.ox.ac.uk/miis/id/eprint/739 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/739 2018-07-22T16:33:45Z

The challenge proposed by PRIMAVERA BSS software company was to find an effective scheduling algorithm that can add new features to their production planning software, with a good performance (being able to run in less than 10 minutes), and that can be su�ciently generic and adaptable to be used by different industries (metal, furniture, wood, textile, and food industry). The requirements configured a NP-hard problem known in the literature as the Flexible Job Shop Scheduling Problem (FJSSP), for which sophisticated mathematical models and heuristic methods are widely available. The wherein proposed approaches consider sequence dependent setup times and different priorities for the operations. Two approaches are considered in the present report. First, two mathematical models were created to address this problem and give insight to the structure of the problem and its constraints. A second approach proposed the use of heuristics. A constructive heuristic used to find initial solutions is followed by the use of an improvement heuristic, which allows to obtain better solutions at reasonable computational costs. The solutions obtained by the heuristics can be used to warm start the optimal solving procedure using mathematical models with commercial solvers.

Adelaide Cerveira A. Araújo Eliana Silva Isabel Cristina Lopes A. Ismael F. Vaz Rui Borges 2018-07-22T16:33:41Z 2018-07-22T16:33:41Z http://miis.maths.ox.ac.uk/miis/id/eprint/738 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/738 2018-07-22T16:33:41Z

The Douro Palace Hotel management wants to develop a software that helps them with the dynamic pricing of the hotel rooms. Revenue Management software is widely used for data analysis of demand and prices. However the hotel barely follows the software recommendations for pricing decisions, mainly due to the lack of understanding of the rationale and the inherent di

Elisete Correira Sonia Dias Fernanda Ferreira Ana Moura Jorge Santos Carla Santos Pereira Paula Nunes Teresa Pereira Filomena Soares Joao Soares 2018-07-22T16:33:38Z 2018-07-22T16:33:38Z http://miis.maths.ox.ac.uk/miis/id/eprint/737 This item is in the repository with the URL: http://miis.maths.ox.ac.uk/miis/id/eprint/737 2018-07-22T16:33:38Z

Savana is a company in the footwear sector which is over 27 years old and has more than 150 employees. This company specializes in children’s footwear of sizes ranging from 18 to 40. Each pair is individually packed in a box which is customized for each client. The ideal size of each box depends on the model of the footwear and the position in which it is placed inside the box. These boxes are ordered from an external supplier that has only certain measures available, so boxes of the same size have to be used for footwear of various sizes and models.

Due to a frequent introduction of new models in the production environ- ment, the box sizes are initially set manually in an experimental procedure (testing), which is often time consuming. Savana challenged ESGI’s par- ticipants to study their packing process, in order to reduce the variety of box sizes, the empty space inside the boxes and to eliminate the need to perform testing, thereby reducing the time and increasing the e�ciency of the packing process.

Furthermore, the footwear ordered by each customer is packed into large boxes, which will henceforth be referred to as containers. With regard to these large boxes, various designs and sizes can be delivered to a single client. The dimensions, weight and forms of these are subject to the customer’s specifications. In this context, Savana intends to determine automatically the sizes of the containers to be sent to each customer and how to arrange the individual boxes for each client’s order.

This report tells how to automatize and speed up the overall process. It describes how to automatically assign shoes to boxes, and gives a manner to pack the shoe boxes, in such way that permit to reduce the size of the card box.

Savana should be aware that this is not yet a ready to use solution, because more data analysis need to be done, in order to improve and make the method reliable. Furthermore, during implementation it may appear new important challenges.

Rui Almeida Jose Duque Flora Ferreira Manuel V.C. Vieira