Predictive Technology Is the Future of Cold Chain

MaxTrace is a proprietary predictive cloud engine, designed to provide real-time actionable insights about the thermal life of each shipper and it creates a digital twin to accurately predict how long the box can maintain its required internal condition. This engine analyzes temperatures, journey duration, lane risk factors, external weather data sources, machine learning intelligence and more. 

“We came up with a way to develop a cold chain container-specific digital twin, which we then deployed as a predictive cloud engine. It then takes all real-time signals and a bunch of different caveat signals and variables, and it allows us to automate predicting and preventing temperature excursions,” said Kumar.

Kumar listed the following as the four key areas where excursion losses occur:

• conditioning error
• packing error
• tarmac exposure
• shipment delays

To avoid these occurrences, MaxTrace gathers three performance indicators, which are the current state of the packaging, coolants, and insulation; lane parameters; and the expected ambient circumstance the container will face. “If we knew all three different parameters, to a certain extent I can predict what the performance of this container is going to be,” Kumar explained.

The function of the digital twin of the container, developed through a pseudo-analytical model, is to reflect the behavior of the container in real-time. This requires real-time input from IoT devices that are integrated with the container.

Qualification data for the container along with trial shipments in real-time are used to calibrate the digital twin.Arif Rahman, the director of technology at MaxQ Research LLC, said, “The way we calibrate the system is we use qualification data that we have for the container and then we do trial shipments in real-time. With that, we can predict the future state of the container. But a pseudo-analytical model is rigid. To make it more flexible, we ingested external features and parameters that can help improve the accuracy of the model. For that, we added lean metadata, container metadata, weather forecasts, carrier milestones, and logistics disruptions, etc. And obviously you have the real-time data that is coming in through the IoT devices. All those are fed into the cloud engine to predict the performance of the container. This makes the model more agile, more adaptable, and the prediction becomes more accurate.”

This cloud engine is designed to provide predictions up to three hours in advance to allow for preventative actions for both active and passive containers, as well as for business and logistic process integrations, through rerouting and returning the shipments.

The goal is for the whole process to become totally automated through prediction and prevention. Benefits beyond excursion prevention include a scalable architecture that is container-agnostic, IoT device-agnostic, and accepting of a variety of data sources.

Concerns in implementing the new technology

As a new process, there are challenges that the company is striving to remediate. The biggest challenge is validating the digital twin-based predictive model as it uses a wide range of factors. Real-time monitoring programs are also expensive, the agility of logistics processes to execute the prescribed action needs improvement, and it can be a challenge to help people understand that the SOP is dynamic.

Mark Maurice, a solutions consultant at Sensitech, further explained that what’s missing are sophisticated workflows with freight carriers and others within other industries. Someone must be available around the clock in the case the preventative action is required to head off a temperature excursion.

Regulatory compliance issues Maurice touched on are:

• The sustainability of the monitors in using single- or multi-use real-time monitors. In the case of multi-use monitors, the devices would need to be transported back to their origin. There could be a refurbishment process put in place where the device is taken to a local refurbishment location. However, Maurice noted, monitoring systems that are reused will require upgrades in firmware and software as well as cell communications.
• Balancing the inventory of monitors as some may be lost or damaged. There is a cost to resetting and reconfiguring these devices. “You have to couple your sustainability plan with what it’s going to cost and balance that between multi-use in packaging and multi-use in monitoring,” Maurice said.

Maurice also emphasized that the way to manage shipments going through multiple climates is through machine learning. The ability to react quickly when an excursion is predicted is by having the optimal solution provided to the company through the machine learning intelligence. Humans on their own take up too much time to come to the conclusion of what solution is the most efficient, according to Maurice. In the case of a universal pack-out, for example, analytics can lead a company with the most cost-effective seasonal shipper or right shipper considering where the product is being shipped to, managing both quality and cost. “But you can’t do that in your mind, right now, you can’t think of all those variables, the weight of the package, the qualifications, how long, which package to use,” Maurice said. “Predictive models can get us there.”

Multivariable analysis through machine learning is the future of monitoring, according to Maurice, as it would allow for the system to define the pack-out to use for a shipment, driving large quality and cost solutions.  

unPACKed with PMMI: Next Stop PACK EXPO International 2022