Market access for South African fruit depends on meeting the phytosanitary requirements of our trading partners. Hence, Hortgro established a bespoke phytosanitary facility – PHYLA – to conduct the trials and generate the data needed to obtain and maintain market access.
Probit analyses are a key PHYLA function. These statistical trials provide evidence for the efficacy of treatments such as cold sterilisation. The challenge is that each trial involves three replications of 10 000 infested fruit per treatment, plus more fruit for the untreated control. “It just generates data through the roof because you’ve got all these insects, fruit, treatments, and results,” says Matthew Addison, previous crop-protection programme manager at Hortgro Science.
Organisations such as the European Plant Protection Organisation and the United States Department of Agriculture scrutinise this data when deciding whether to allow the importation of our fruit. “For example, the USA wants to know the genetic origin and diet of your test insects,” says Addison. “That alone is a whack of data. You end up with just spreadsheets everywhere.”
Planning and executing trials generate even more spreadsheets. PHYLA frequently runs multiple probit analyses simultaneously while conducting various industry-funded research projects on postharvest treatments. “At this stage, I need to keep track of everything in my head and on spreadsheets,” says PHYLA manager Dr Renate Smit. “It becomes a bit much when nine trials start in the same month.”
Adaptive software structure
One way to address these issues is to create a digital twin of PHYLA. Such a twin can mirror trials to serve as a planning and execution framework, collect and structure the data from each step, and maintain traceability throughout.
To construct a PHYLA digital twin, Addison turned to Prof. Anton Basson. He heads the Mechatronic, Automation, and Design Research Group at the Department of Mechanical and Mechatronic Engineering at Stellenbosch University and has extensive experience helping companies organise their data. “One of the big challenges with operational and data control is handling changes,” says Basson. “You can design a system super-efficiently if nothing ever changes, but life’s not like that. So, we use a holonic approach to structure software so it can adapt more easily.”
A holon is a stand-alone entity that forms part of a bigger system. Holons can be combined to make more complex holons and hierarchical systems of holons called holarchies. Think of it like this: a fruit tree is like a holon. It’s a stand-alone entity. An orchard is a holon consisting of tree holons, and a farm is a holon consisting of orchard holons. Holons can also represent processes – for example, orchards could have holons for harvesting.
A digital twin based on holons is great for data management. Each holon mirrors a physical counterpart and contains all its data. The relationships between holons structure the data, which can be accessed at the appropriate level of the holarchy. Sometimes, you want data from individual trees, and sometimes, you want an aggregate of the orchard. Process management is where holonic systems shine. “A holon is more than a line in a database,” explains Basson. “Holons are intelligent software entities that are independent but collaborative.”
Holons can organise and optimise processes, based on available resources and user-defined constraints. For example, in a digital twin of a farm, a holonic system could integrate yield estimates and fruit-maturity data from tree holons, availability data from bin-trailer and picker holons, and weather data to optimise the harvest schedule.
Leave it to the holons
As a proof of concept, MSc student Divan Smit created a digital representation of PHYLA under the supervision of Basson and fellow engineer Prof. Karel Kruger, who has since moved to the University of Cambridge. Renate Smit – no relation to Divan – was excited when shown a demonstration of the software. “I’m currently doing this on paper,” she exclaimed. “How soon can I use this software?”
The system has resource and process holons. Resource holons fall into three categories: operators, equipment, and rooms. Cold rooms are differentiated from equipment for several reasons, including that a room can be partially allocated to a trial. So, several trials can be stored simultaneously in a cold room, but something like a penetrometer can only measure one fruit at a time.
For processes, each step is defined, including estimated time and required resources. The complete process is built from these steps, and several processes can be combined to form a trial. These process templates are flexible but make it quick and easy to schedule trials. “We had a pear trial, for example, where we needed specific information. But then my staff did all the pear trials this way,” recalls PHYLA’s Smit. “That was a waste of time because generating those unnecessary data points was very labour-intensive.”
With the software, Smit can select a process template, fine-tune it if necessary, and schedule it. The holons will jump into action, confer, and establish whether the required resources are available to run the trial in the desired time frame. If they are, the system books the resources and adds tasks to the operators’ schedules. If the holons see that resources aren’t available, the system alerts Smit so that she can make another plan.
For example, she may need to hire extra people to perform fruit-maturity indexing during peak periods. Once a trial runs, the holonic system can also keep track of real-time disruptions. For example, the system can reallocate tasks to available operators if someone calls in sick. “You don’t need to depend on your memory or spreadsheets when it comes to all these scheduling conflicts,” says Basson. “The scheduling information also lets you see how much each resource is used. So, you can see whether you need more equipment or the impact of an item breaking.”
Bringing it all together
Traceability is non-negotiable in phytosanitary trials. Every detail must be captured and accessible for PHYLA to ensure our fruit industry obtains and maintains market access. This includes, for example, the orchard where the fruit originated, the quality assurance of insectary diets, and the temperature profile of cold rooms. Temperature data perfectly illustrates how the holonic system can enable traceability and optimise efficiency. Currently, temperature data for each cold room is captured in software specific to the cold rooms.
To generate trial reports, Smit must log into this software, extract and download the data for the relevant period, and rework it. In contrast, the holonic system knows when the fruit boxes for a given trial were in a specific cold room. The holon for that cold room automatically communicates with the cold-room software to obtain the temperature data. When Smit needs a report, the process holon for reports gets the temperature data from the cold-room holon and puts it in the correct format.
“I’m starting to develop a phytosanitary protocol for a new insect pest, so we will have a data package ready should a market request it,” says Smit. “With the holonic system, all the information is linked, and I don’t need to enter or search for data manually.” A digital twin based on a holonic system could allow PHYLA to function more efficiently and reassure our trading partners that our phytosanitary house is in order. Basson and Divan Smit’s work demonstrated that Addison’s vision for the system can be realised.
The next step would be for a programming company to develop the holonic framework into a fully-fledged application. Although customised to PHYLA, it could be adapted to manage phytosanitary treatments and cold storage across the industry.
