Project software firm nPlan has agreed a deal to provide its artificial intelligence (AI) forecasting and risk management services to the Transpennine Route Upgrade (TRU), a multi-billion-pound programme of improvements to transform the railway network between York, Leeds, Huddersfield and Manchester.聽
The deal will allow TRU to analyse more project schedules, more frequently, and get more meaningful insights and accurate forecasts than it could through any other method of assurance, says nPlan.
After a gradual introductory phase, TRU eventually intends to use nPlan instead of quantitative schedule risk analysis (QSRA) across its entire programme of works. 聽 聽 聽聽
QSRA is used to generate a numerical estimate of the overall effect of risk on project objectives such as cost and schedule. NPlan describes it as 鈥渢he previous gold standard for forecasting and risk management in construction鈥�.聽
Unlike QSRA, nPlan鈥檚 outputs are generated by analysing large volumes of historical project data with AI. At the start of each engagement, nPlan gathers the customer鈥檚 past project schedules and uses an AI technique known as 鈥榙eep learning鈥� to build a model that reflects how the customer executes projects.聽
Among other things , this model captures how the different construction activities the customer undertakes play out in different contexts. Once the model has been trained, it can be 鈥榝ed鈥� schedules for upcoming projects to generate an independent, data-driven and performance-based forecast of how the project as planned will play out, along with detailed information on the risk profile of every activity in the scheme.聽
The outputs of the AI-led process can then be analysed via nPlan鈥檚 web-based platform, enabling customers to home in on the riskiest activities quickly and easily, testing the impact of various risk mitigation scenarios, and visualising their project pathway. 聽
A typical QSRA process depends on human estimates of how long activities will take and which project activities are risky. This determination is inevitably influenced by cognitive biases such as optimism bias, recency bias, and salience bias, says nPlan.聽
By using historical data and AI, nPlan can screen out the effects of these biases to generate a realistic forecast and identify risks that would otherwise have remained hidden from the project team.聽
QSRA is also a resource-intensive process, nPlan says, requiring a lot of people and time to come up with a result. Previously, TRU was only able to perform QSRA on individual projects once a month, and on its whole programme every three to six months. But with its nPlan model trained, TRU can now trigger an analysis whenever it has an updated schedule, ensuring the project team can continue to reduce risk at the rate required to keep the project on track.
Richard Palczynski, head of strategic programme controls at TRU, says: 鈥淭he Transpennine Route Upgrade is an ambitious and complex multi-year programme. Getting the right forecasting and risk management capabilities in place will be critical to finishing on time, on budget, and with the minimum possible disruption to passengers.聽
鈥淲ith nPlan as a delivery partner, we鈥檙e able to take a new approach to analysing our schedules 鈥搑emoving the human bias in favour of learning from actual data and historical performance 鈥� and driving efficiency into the process. Being able to get more frequent analysis done on a larger volume of schedules is a game-changer for us 鈥� we simply don鈥檛 have the resources to use QSRA to generate the insights the programme needs to stay on track.鈥�
Dev Amratia, nPlan鈥檚 chief executive, adds: 鈥淔or some time the rail industry has been signalling its frustration with established assurance methods, which simply can鈥檛 provide the level of confidence required by complex, multi-year rail programmes.聽
鈥淒ata-driven solutions like nPlan鈥檚 represent our best chance in a generation to get rail project execution back on track. The market seems to agree 鈥� we鈥檙e currently seeing very strong demand from rail project owners and contractors looking to upgrade from low quality, resource-intensive methods of providing assurance to an AI-led approach.聽
聽鈥淲e鈥檙e delighted to be partnering with the TRU programme team on this historic project, not least because they share our vision of delivering continuous forecasting and risk management for major rail programmes,鈥� concludes Amratia.
While TRU pins its colours to the nPlan mast, HS2鈥檚 Align JV is working with US-based project management software specialist Alice Technologies to find better ways of delivering its section of HS2 to the north-west of London.
The Align JV, comprising Bouygues, Sir Robert McAlpine and VolkerFitzpatrick, is delivering the C1 package, a 21.6km stretch of railway that includes the 3.37km-long Colne Valley viaduct and the 16km twin-bored Chiltern Tunnel.
In late 2021 Alice and the Align JV set up a six-week pilot to review the programme for the viaduct substructure. This is a critical and complicated part of the programme and was therefore a good test of the Alice system.
The viaduct is especially complicated due to various constraints including local planning restrictions, environment requirements, difficult site access and logistical challenges. There are also a large number of specialist contractors to manage.
Alice uses AI to simulate millions of scheduling options in minutes, helping contractors plan and building complex infrastructure projects more efficiently. Align JV is hoping that it will result in significant reductions in cost, programme and risk.
The team spent four weeks getting up to speed with Alice and building 17 鈥榬ecipes鈥� to analyse. These recipes form Alice鈥檚 instructions and breakdown elements including specific construction tasks that are required. For the pilot, there were 642 separate operational factors to consider. Alice generated six million potential options in just 10 minutes.聽
The remaining two weeks of the trial was spent analysing the simulations and exploring nearly 1,400 scenarios. Examples included testing the mix of teams required on the substructure in order to optimise their utilisation and minimise downtime for labour and equipment. 聽
Alice was also used on specific work packages. For example, the team looked at options for the viaduct鈥檚 top deck and finishes. This helped them to write the scope and package out the various works in terms of delivery windows and key requirements. A key benefit was that it allowed for more detail in contacts and improved price certainty for both HS2 and the subcontractors.
Because Alice was seen as neutral it was also helpful when it came to potential risks and preferred solutions, according to Nick Podevyn, innovation & improvement manager at Align JV.
鈥淭he subcontractor relationship can sometimes be adversarial, but because Alice is independent and based on the data that each contractor and sub-contractor enters, it is a lot easier to discuss the challenges openly,鈥� he says.
鈥淚t also provided subcontractors with value in terms of setting their own programme and providing greater certainty for them.鈥�
The initial six-week trial has since been extended into a 12-month strategic pilot to include the viaduct superstructure, finishes, demobilisation and trial the system鈥檚 鈥榤anage鈥� function.
鈥淎lice has opened people鈥檚 eyes to the way AI can be applied to construction project management,鈥� says Podevyn. 鈥淩unning different construction scenarios using traditional programming software is time consuming and error prone. This innovation has supported collaborative planning with our subcontractors and enabled the project team to rapidly run alternative build sequences.鈥澛�
Got a story? Email news@theconstructionindex.co.uk
