Study group problems and presentations
Problem 1: Predicting and mitigating the effects of airblasts in mines
Industry: Mining
Industry Representative: Richard Stacey
Moderator: Jim McKenzie
Student Moderator: Cindy Carvalho
Problem Statement:
An airblast in a mine resulting from the collapse of rock can be extremely hazardous and may result in severe damage - mine equipment may be annihilated, ventilation doors and seals destroyed, and conveyor systems and shaft steelwork and equipment ripped out. The source of an airblast is usually the major collapse of rock in an underground mine opening that contains a large volume of air. The collapsing rock acts as a piston to compress the air in the opening. This type of situation can develop in underground coal mines, open stoping mines, and caving mines in which over drawing takes place and the stope back hangs up over large spans. In such cases, sudden rock collapse leads to the displacement of the volume of air, which is then forced along any available pathways – tunnels, shafts, orepasses, access underground roadways, and some will escape “through” the collapsing mass of rock.
As an example of the effects of an airblast, a particularly severe event occurred at an Australian mine. The figure below illustrates the rock that collapsed and the paths that the air followed. Several lost their lives in the event. A Land Cruiser in which two persons were travelling in the mine was totally destroyed, only the engine block being found intact. All ancillary bolt-on fixtures such as starter motor, alternator etc were ripped off.
(from a presentation by Iain Ross, Manager Mining)
The questions considered to be of interest are:
- What pressures and velocities are attained by the air in the pathways taken?
- How easily can air escape through the collapsing, broken rock mass, ie a holey piston?
- How can the effects be mitigated? By alternating excavation sizes, ie artificially changing the roughness of the tunnels? By inducing reversed or counter-airflow, possibly as indicated in the sketch below? What barriers would withstand the pressures and velocities?
Problem 2: Prediction of fracturing and dynamic failure of the roof in tabular stopes in a platinum mine
Industry: Mining
Industry Representative: Richard Stacey
Moderator: Ashleigh Hutchinson
Student Moderator: Despina Zoras
Problem Statement:
Suspected dynamic collapses of the hangingwall (roof) of tabular stopes have occurred in a platinum mine at a depth of about 250m. These stopes typically have a height of about 1,5m. They are advanced on strike (termed breast mining) and each panel has a dip span, between rock pillars, of about 28m-31m. The roof consists of anorthosite, a brittle rock with a compressive strength of about 200MPa. In the problem areas the thickness of this anorthosite is typically 7-10m (spotted and mottled anorthosite combined). It is effectively unjointed and contains no weak layering. The horizontal in situ stress in the strike direction is about 1,8 times the vertical stress at this shallow depth.
Roof falls have occurred, which indicate the effects of stress. The photograph below shows one such fall, illustrating the formation of flat fracture planes and slabs, which are frequently observed. MISG Stope roof fracturing and bursting problem 2014.pdf.
Presentation of report back to industry:
Prediction of fracturing Final Presentation.pdf
Problem 3: Wastage minimization of aluminium lengths in window and door industry
Industry: Glass Design
Industry Representative: Carl Runge and Reiner Runge
Moderator: Montaz Ali
Student Moderator:
Problem statement
Glass Design manufactures doors and windows from sections of aluminum lengths. Each window can have up to 9 different sections or more and each section is extruded in a “standard length”. The supplier of the aluminum lengths offered Glass Design the choice of a “standard length” between 3 m and 8 m. The total weight of the “standard length” for each section ordered must not be less than 250 kg as required by the supplier of the aluminum lengths. Glass Design has to order in bulk for about 1 to 20 different size doors per week where each size of door needs up to 9 different sections or more.
Given a number of different door sizes what will be the optimal “standard length” and quantities for each section, out of up to 9 sections or more per door size so that cutting wastage is minimized subject to a 250 kg minimum order per section length.
Problem 4: Spinners detection
Industry: Real Impact Analytics
Industry Representative: Sebastien Leempoel
Moderator:
Student Moderator:
Problem Statement
Spinners are mobile phone users who change at high frequency of Mobile Network Operator(MNO) to catch the advantage of new promotions. These users often change SIM cards and arehard to detect for MNO's, who would like to detect them in order to measure what fraction oftheir past customers are returning to their network, and at what frequency.To detect which customers are returning customers, MNO's have only few information to use,since the returning customers use a di erent phone number each time. The only informationthey are able to use is what they observe in the calling patterns of their customers: who theycall and where they place calls.
MNO's gather daily the logs of all calls and text messages of all their users, and know hencewho calls and is called by each of their users. In these call logs, they also have the location in-formation of their users at the time of the call, allowing them to estimate their preferred locations.
Two customers A and B are likely to be identical if they meet the following criteria:
- A has made calls until a date d and has never called after that (A is called a churner),and B has made calls only after the date d + e
- The social networks of A and B are very similar;
- The places where A and B are when they make calls are very similar (usually, MNO's usea top x locations to avoid too much variance);
- A has never called B and vice-versa.
The aim of an MNO is hence to build a list of potential spinners, with their old and new identi er.
In this project, the students will perform the following tasks:
- Formalize the problem in mathematical terms. In particular, the students will need to decide how to measure similarity of social networks and mobility between two users;
- Suggest and implement an algorithm to detect spinners on a small synthetic set of Call Data Records (CDRs), which will be provided by Real Impact;
- Test their method on a large synthetic set of CDRs and adjust their method to computational constraints;
- (optional) if their method performs correctly (both in speed and results) on the large synthetic CDRs, their method will be tested by Real Impact on real-world data, to observe the behaviour of the method in real-market situations. The performance of the method will be shared with the students (for con fidentiality reasons, direct access to real-world data won't be possible).
Presentation of report back to industry:
Spinner Detectionv5.pdf
Problem 5: Propositions assignment
Industry: Real Impact Analytics
Industry Representative: Sebastien Leempoel
Moderator:
Student Moderator:
Problem Statement
Marketing strategies have changed over the last decade from mass marketing (all customers are targeted with one main message about anew product or promotion) to personnalized advertising, where specific products and messages are addressed to each customer, to offer the right product to the right customer who needs it. This change has brought up a new problem to marketeers, which is to define how to distribute the available promotions to their customers in order to maximize the potential return of the marketing campaign.
A typical situation can be described as follows:
- the operator has a number of propositions (for example a product at a discounted price) that can be offered to their customers;
- for each proposition, the operator has a limited capacity that they can offer to their customers, the capacity may vary depending onthe proposition;
- each proposition generates a return on investment, which may depend on the type of proposition;
- the operator computes for each customer and for each proposition the probability that the customer will accept the proposition if it is offered to him.
- some proposition cannot be combined with others (if a customer has received an offer for 2Gb of data at 10% discount, he can’t receive after that an offer for 1Gb of data at 10% discount).
RIAMISG2014AssignmentWithout Solution.pdf the full problem statement
Problem 6: Liquid hold-up in a diffuser
Industry: Sugar Cane Processing
Industry Representative: Richard Loubser
Moderator:
Student Moderator:
Problem statement
The preferred method of extracting sugar from sugar cane in South Africa is by means of a diffuser. The cane is shredded and then fed into one end of a diffuser and moved to the other end using an arrangement such as a chain slat conveyor. Water is added at the other end of the diffuser and allowed to percolate through the bed. The sugar juice-water mixture is collected in a tray below the cane bed and pumped to a position closer to the cane feed end where it is sprayed on the top of the bed and allowed to percolate once again through the bed. In this way a counter current extraction/washing process is established. While most of the juice leaves through the intended tray, some of the juice returns to the tray from where it was pumped while other juice bypasses a stage because the cane bed is moving.
The operation of a diffuser is not a steady state operation but rather highly time dependent. The permeability of the cane varies from consignment to consignment. Sometimes the water-juice mixture will be pumped too quickly to the top of the bed causing flooding and uncontrolled mixing of the juice of varying concentrations. This destroys the counter current concentration profiles and reduces the effectiveness of the extraction process. If the flooding occurs near the cane feed area, the juice can overflow out of the diffuser. Operators will often stop the pumps feeding the flooded area. This leads to an accumulation of juice in the tray feeding that pump. This practice can cause overflow of the trays below the diffuser. If there is insufficient juice around the cane then the sugar will not be extracted resulting in losses so it is best to keep the level of liquid in the cane bed just below the critical point where flooding occurs.
The aim of the study is to model the time dependent nature of the flows within a diffuser so as to advise how to keep the juice quantity in the bed at the desired level and what action should be taken when flooding or dry operation occurs. This will require a study of the factors affecting the juice hold-up in the bed such as chain speed, bed height, percolation, recycle and bypass. In addition to these, factors such as pump and pipe characteristics will need to be considered together with the tray capacity and level relationships.
Presentation of Problem Statement:
Diffuser dynamic model Intro.pdf
Problem 7: Airline schedule assignment and construction
Industry: Airline
Industrial Representative: Alan Watson
Moderator: Ian Campbell
Student Moderator:
Problem Statement
Given
- Flight schedule (individual flight legs for a period, typically a season);
- Airline's fleet of aircraft (resources).
Constraints
- Flights must connect geographically (departure airport must be same as previous arrival).
- Flights should have a gap (“recommended ground time”) between one another, which depends on the airport.
- Flights must have a gap between one another (also airport dependent) which is the “minimum ground time”.
- Flight legs of the same flight must have an inter-leg gap of at least the the minimum passenger connecting time if there is an aircraft change (a flight is identified by a single number for all of its component legs, and a “flight date”, being the departure date of the first leg).
- There are carry-in and carry-out flights at each end of the period which must also connect with the flights inside the period (including both geographic and ground time constraints).
Problem 8: Traffic analysis on Bus Rapid Transit (BRT) routes
Industry: Gauteng Traffic Department
Industry Representative: Progress Hlahla
Moderators: Neville Fowkes and Dario Fanucchi
Student Moderator:
Problem Statement:
In order to facilitate the flow of traffic in Johannesburg during peak hours the Bus Rapid Transit System (BRT) is investigating the effect of allowing minibus taxis to use the lanes presently reserved for buses either throughout the day or just during peak hours. One would expect the reduction of flow in the normal lane to result in increased car speeds in this lane and also increased speeds for the minibus taxis in the bus lane, however bus speeds may be reduced and therefore timetables not adhered to. Also there may be disruptions caused by lane changes. The Gauteng Traffic Department asked the Study Group to investigate these issues.