PhD students go FIFO

PhD students Ryan Leadbetter, Kyle Fethers and Braden Thorne visited Roy Hill mine site for two days.  They got first-hand experience of living in a FIFO camp and networking with mining personnel that will use the outcomes of their research.  All were especially pleased to see the equipment associated with their research.

We thank Calvin Tang for organising this opportunity and Mitin Hirani for enabling our PhD industry experience in Roy Hill.

Ryan Leadbetter, Kyle Fethers and Braden Thorne

Collaboration with other ITTC - PhD students attend CIRES inaugural Information Resilience PhD School

PhD students Ryan Leadbetter and Sandy Spiers were offered places to attend the Information Resilience PHD School in Queensland.  This opportunity was made available by ARC Training Centre for Information resilience

The PhD School was a 3-day, in-person event held at the Global Change Institute at the UQ St Lucia Campus.

The school was an excellent opportunity for Ryan and Sandy to collaborate with PhD students from other ITTC disciplines on data-oriented research. During the course, PhD students heard from national and international researchers aiming to address the following data challenges:

• Low data quality, i.e., handling issues like missing entries, false information, lack of samples, class imbalance 
• Limited insights obtained from data, i.e., how can we maximise the value of “big data” with advanced techniques? 
• Unknown reliability and trustworthiness of data analysis tools, i.e., is the automated decision process transparent and interpretable? 
• Anything that hinders the resilient application of research, e.g., ethics, privacy, cyber security, generalisability, efficiency, adoption.

Optimised activity scheduling for maintenance outages

In the  March newsletter , we discussed the  optimal shutdown scheduling software  developed by  Hoa Bui,  Mojtaba Heydar,  Ryan Loxton, and Elham Mardaneh to help the managers and schedulers quickly generate shutdown schedules under different conditions to assist their decisions.

With software development support from Theme 5 and John Hille, Hoa Bui and Mojtaba Heydar are working together to further develop this new scheduling platform.  Ryan Loxton and Hoa Bui presented the outcomes of this work to the Centre's Strategic Risk and Reporting Board and the Nickel West Planning Team. 

The primary goal of the  optimal scheduling platform  is to provide planners with a powerful optimisation system that will alleviate much of the manual work they currently perform. 

Instead of building the schedule themselves, planners will only need to specify the requirements that the schedule must satisfy; the algorithms developed by the CTMTDS team will create the best schedule that meets these requirements. 

This will substantially shorten the time required to create and update schedules and reduce maintenance downtime  through better scheduling and  more efficient resource allocation . 

A key benefit is that planners can quickly create hypothetical schedules under different scenarios and explore how changing the work scope affects the shutdown duration. 

Fast algorithm for solving p-dispersion-sum problems has vast applications in industry

Sandy Spiers, Hoa Bui and Ryan Loxton have published a paper on the mathematical modelling that underlies their industry research titled " Cutting plane methods for solving  p -dispersion-sum problems .

The p-dispersion-sum problem(PDSP) — a classical and challenging quadratic combinatorial optimisation problem — involves selecting a subset of elements from a larger set to maximise some distance or dispersion metric. 

This problem has vast applications in industry. While research into heuristic and meta-heuristic approaches to the p-dispersion-sum problem has gathered significant interest, the development of exact algorithms has fallen behind. 

Researchers in  Theme 3 have developed an exact algorithm using the cutting planes approach for solving (PDSP).  Their method outperforms the state-of-the-art solvers for such problems (see the figure below for the performance benchmark). 

Performance benchmark. P roportion of problems solved:  our algorithm (blue) versus Glover's reformulation with CPLEX (organge)

(For further reading: Spiers, Sandy, Hoa T. Bui, and Ryan Loxton. "An exact cutting plane method for solving p-dispersion-sum problems." arXiv preprint arXiv:2207.10879 (2022) .

Shutdown timing optimisation

Yingying Yang has returned to the Centre after taking some extended parental leave to continue work her research on staging multiple maintenance operations over a long time horizon. 

 Yingying is working on a research paper titled "Shutdown planning optimisation for the iron ore mining industry".   
In her paper, Yingying discusses the model she has developed to support the advance planning of significant maintenance shutdowns. Yingying has developed a time-indexed mixed-integer linear programming model to optimise the long-term shutdown plan for a complex interconnected system, considering all maintenance requirements and total throughput, inventory management and market demands.  Yingying trusts that her model delivers an optimal shutdown plan for an integrated mining system, with the corresponding optimal throughput and stock values.

We look forward to reading her paper.  

Developing exact methods for solving mixed-integer nonlinear programming to support maintenance scheduling problems

Developing cutting-edge algorithms to optimise maintenance scheduling is a primary goal for  Theme 3 . 

Over the past decades, there have been tremendous advances in  linear   integer   programming , culminating in mature commercial solvers such as CPLEX or Gurobi. However, we have not seen progress in  nonlinear integer programming . There is a significant performance gap between linear and nonlinear integer programming methods. Yet, many maintenance scheduling problems faced in the resources and energy industry are  inherently nonlinear.  Hoa Bui's  work on Alcoa's 16-week maintenance schedules is a classic example of nonlinear scheduling models. Using the contemporary nonsmooth analysis theory, researchers in  Theme 3   aim to create fast, exact algorithms in mixed-integer nonlinear programming to tackle large-scale nonlinear problems.

(For further reading, check out: Bui, Hoa T., Qun Lin, and Ryan Loxton. "Cutting plane algorithms for nonlinear binary optimisation." arXiv preprint arXiv:2203.09703 (2022).)