Digital Twin of the Air Cargo Supply Chain

Summary. In this paper we develop a digital twin based on the new One Record linked data standard. This enables short-term workload prediction for the various partners in the air cargo supply chain without the need for multiple data exchange interfaces. To the best of our knowledge, it is the first research on the potential benefits of One Record. The concept of the digital twin allows for an overarching optimization of operations in the air cargo supply chain without the necessity of full transparency between all the partners


Introduction
Air cargo is the mode of transport for high value and urgent shipments.The volume of air cargo shipments accounts for less than 5% of the global trade volume in weight, but at the same time almost a third of the global trade volume if measured by its value (IATA 2016).While air cargo is of great importance for global supply chains and transports high value goods it is lagging in digitization, partially due to its complex supply chain structure with many parties involved (Feng, Li & Shen 2015b).
The end-to-end-supply chain considered in this paper begins at the shipper and ends with the delivery to the consignee.To reduce complexity, we primarily contemplate the main parties involved and a direct flight with a single flight leg only (Figure 1).The following analyses exclude integrator and airmail operations but focus on forwarder-driven air cargo supply chains.2023 International Scientific Symposium on Logistics For multiple reasons forwarders use many different airlines.Airlines themselves make use of various logistics service providers, e.g. for cargo or ground handling, as it is not possible or economically feasible to offer own operations at all airports.For this research it is assumed that the forwarder operations are fulfilled by the forwarder itself, but the cargo and ground handling activities are outsourced by the airline which operates the flight only.Some capacity is secured by so called blocked space agreements between forwarder and airline.The rest of the transport volume will be allocated based on pricing and available capacities which is also characterized by volatility (Schutte et al. 2022).This leads to the complex network structures of Figure 2 where typically shippers use services of several forwarders which use flights from various airlines.Airlines themselves have contracts with one cargo and one ground handling service per airport.At least since the description of the Forrester or Bullwhip effect it has become obvious that information exchange is an essential driver of supply chain performance.State of the art in information exchange in the air cargo supply chain is message forwarding, typically to many, but not all parties involved up-and downstream in the supply chain.As often some information is changed at one process step (e.g. a change of weight or volume) parties involved in later steps in the transport chain receive multiple messages for the same shipment with contradicting information.Therefore, the quality of these message is less trusted and operational-wise disregarded until the shipment arrives.
To improve data quality and trust the International Air Transport Association (IATA) started the project One Record (1R) which is based on a linked-data approach (Blaj et al. 2020).This ensures a single truth in combination with a decentralized data exchange.IATA aims for a full implementation of the One Record standard by the airlines in 2026 (IATA 2023a).
The objective of this paper is to conceptually demonstrate how 1R data can be used by the various parties in the air cargo supply chain to improve the near-time workload prediction.This is made possible by becoming aware of certain activities at an upstream partner (e.g. the start of the build-up of an air cargo pallet) which will impact the own workload in the near future.
The paper is structured as follows.After a short literature review on air cargo and digital twins (section 2), the concept and data model of One Record is presented (section 3).This is followed by the presentation of the model and the input data used to feed the digital twin.Section 4 contains the concept of the digital twin and its limitations.In the conclusion the main findings are summarized.

Air Cargo
A comprehensive overview of air cargo operations research is given by Feng, Li & Shen (2015b) while Morrel & Klein (2019) or Schäfer (2020) present a detailed overview of all aspects of air cargo.Merkert (2022)

Digital Twins and Simulation
Digital twins have been widely used in other industries for decades, e.g. the aerospace industry (Haße et al. 2019;Tuegel et al. 2011).A digital twin uses constantly updated data and is also adapted according to the changes in reality (e.g.exchange of a certain machinery or any kind of extension) (Rosen et al. 2015).To enable the digital twin functionality provision of data from various systems, formats and owners is a challenge (Bazaz, Lohtander & Varis 2019).
Pérez Bernal et al. (2012) use a simulation software to model a forwarder-to-consignee air cargo supply chain for the airport of Zaragoza, Spain.They conclude that double-digit savings are possible if an overarching simulation model would be used.

One Record
One Record is an open linked data concept which is set to become the new standard for data exchange in the air cargo supply chain by 2026 (IATA 2023a).It is designed to overcome the above-mentioned deficits of existing messaging services.Instead of using the airway bill (AWB) as the shipment identifier One Record follows a piece-centric approach where all pieces get their unique identifier (Uniform Resource Identifier, URI).Pieces are linked together to form shipments which are identified by the house airway bill (HAWB).Additionally, owner of relevant assets (logistics objects, LO) could assign them URIs -e.g. unit load devices (ULD), trucks and planes.Data elements of a logistics object (e.g. the weight of a piece) can be changed by the owner of this object only.Changes can be requested via a patch function but have to be accepted by the LO owner.This ensures a single source of truth as changes to shipment are transparent to all parties involved.Changes are communicated to all parties involved via a subscription feature.Whenever logistics objects are linked, either virtually by consolidating pieces into a shipment and HAWB or physically in the build-up-process, new data is created, and the partners are informed via the subscription feature.The full documentation of all functions and links to the open-source repositories can be found under (IATA 2023b).

Simulation Model
The digital twin of the air cargo supply chain is build using Siemens Plant Simulation software.With an One-Record-interface the twin can be fed constantly (or regularly) with new data and thereby provides a near-real-time reflection of the relevant part of the air cargo supply chain.The twin relies solely on the One Record data and does not need interfaces to other (legacy) systems or partners -besides the information on infrastructural capacities and workforce available.
The first model completed is the forwarder hub which already includes all relevant landside functionalities (Figure 3).As the virtual creation of shipments (with HAWB) and bookings (Linking shipments to AWBs) is triggered by the forwarder it is the most relevant process step.
Additionally, the forwarder hub takes over the build-up-and break-down-process for a share of the shipment volume.Therefore, the implications of this process step can be already demonstrated by modeling the forwarder hub.

Dataset
For calibrating the simulation model, real shipment information of a local forwarder hub was used.In February 2023 almost 4 000 t with an average MAWB weight of under 1 t/MAWB were handled in the terminal.During the period observed 693 ULDs were build-up.There is a significant higher share of export shipments.From the timestamps of truck arrivals and departures in combination with the available workforce the fill level of the terminal, the distribution of truck handling times and the utilization of different functional areas could be calculated.For validation purposes the output from the simulation was compared against the real output timestamps and volumes.
To model the benefits of One Record, the existing shipment data was transformed into the One Record format (as real One Record data is not available yet).The dataset will be further extended with data from a cargo handling company and from a ground handling service provider.Thereby all basic processes of the simulation model will be validated with real data and ensure that the model behavior reflects real system performance.

One-Record-Driven Air Cargo Supply Chain Model
The benefits of the One-Record-driven air cargo supply chain models derives from the possibility to learn about upcoming logistics objects in their exact quantities, weight, and volume at the time of their creation.We identified six relevant activities that are useful for the workload prediction of the downstream partners in the supply chain (Figure 4).Whenever a shipper creates a piece (which should be shipped via air and therefore gets an URI), the forwarder gets immediately notified about the new data and can plan a pickup whereas today the shipper waits until all pieces are completed, and the forwarder is then informed to pick up the shipment.
With the AWB booking and the subsequent linking of shipments (and thereby the linked pieces) to this AWB, this information can be used by the cargo handling agent to optimize workforce planning -either at the truck dock or for the following processes.
The ground handling agent can learn the number of ULDs to be transported from the beginning of the build-up process (and can assume their time of availability) whereas today this information will be communicated in the best case after a ULD is completed and left the scale for the weight and balance.Based on the typical build-up performance the information is available at least one hour earlier.
Based on the number of ULDs loaded onto a flight the ground handler and the cargo handler at the inbound airport can already plan the resources necessary at the time of departure of a flight.
Without the necessity of checking customs statuses, the forwarder can derive from the start of the break-down process the availability of inbound shipments and can plan accordingly.
The air cargo supply chain will be split in different models for each supply chain partner to avoid the sharing of sensitive information (as in an integrated model all parties involved would have access to all information, e.g.workforce numbers, operational strategies).Based on data of expected incoming shipments and those which are already buffered or processed in the hub, the hub output performance can be simulated.The impact of late arrivals or short staffing or technical issues (e.g.breakdown of an automated storage system) can be simulated and its impact on the output in terms of volume and time will be shown.Based on the identified bottlenecks different operational strategies can be applied.
The relevant output information can be shared with the downstream partners by the use of planned events in One Record.

Limitations
As One Record data is not available today, the twin reflects potential benefits which cannot be validated as a direct comparison is not possible.The concept assumes full adaption from all forwarders and airliners.The research focuses on forwarder-driven air cargo and is not applicable to integrator or airmail operations.The model reflects the processes at an European air cargo hub.Different processes and structures could apply at other airports.

Conclusions
One Record helps to overcome inefficiencies in today's data exchange.Based on the subscription of a limited number of events in the air cargo supply chain the downstream supply chain partners are capable to improve their workload prediction and thereby achieve a more efficient resource planning and allocation.Especially unforeseen disruptions or capacity reductions can be made transparent to the downstream partners without sharing sensitive information.

Figure 1 :
Figure 1: Simplified Air Cargo Supply Chain

Figure 4 :
Figure 4: Relevant activities to predict the downstream workload