Keywords

1 Introduction

While the mining industry remains one of the world’s largest industrial sectors, there are a number of concerns regarding its long-term viability. Exhaustion of mineral reserve, low grade minerals, scarcity of resources, rising energy cost, and commodity market imbalances are the trends that challenge the future of the mining industry. These changes imply volatile, uncertain, complex, and ambiguous environments for firms and affect them across their strategic environment [1]. Various initiatives emphasized the urgency for advanced innovative strategies to tackle those challenges and to support the economic growth and the competitiveness of mining companies on the market.

Along with the adoption of the new generation information technologies in the industry, the term “Digital Twin” has experienced a growth in popularity, and is now recognised as a key enabler for the digital disruption of the industrial operational activity. Digital Twin is defined as the virtual and computerized counterpart of a physical asset, process or system [2]. Digital Twin mirrors the physical systems and the processes that are articulated alongside the system in question, and across its life-cycle, mimicking its operation in real-time. Digital twins consists of several components such as models, data streams coming from the Internet of Things (IoT), sensors, data models, artificial intelligence and machine learning-enabled analytics and algorithms [3]. When combined with information and communication technologies and powerful data analytic algorithms such as artificial intelligence, Digital Twins allow industrials to perform advanced diagnostic, predictive analyses and control during operations [4].

This paper aims to provide the basis for further work in the field of the Digital Twin for the mine value chain optimization. This topic is a part of the Moroccan national project “Smart Connected Mine” that arose after the consortium of industrials and researchers for the digital transformation of the mining industry. The content of this paper addresses Background, definitions, challenges, key enabling technologies and values of Digital Twins, as well as an architecture for the implementation of such concept to smartly control mineral processing is proposed. This paper is structured as follows: In Sect. 2, Digital Twin background, definitions, types and misconceptions are presented. In Sect. 3, the values of a process Digital Twin in a mineral processing plant are listed. In Sect. 4, a generic architecture of the integration of Digital Twins in mineral processing plant is proposed, as well as the different approaches of developing a process Digital Model are detailed. Section  5 and 6 present the challenges and key enabling technologies of process Digital Twin adoption. A conclusion and work perspective are given, by the end, in Sect. 7.

2 Digital Twin: Background, Definitions and Types

2.1 Background and Definitions of Digital Twins

The use of “Twins” dates back to NASA’s Apollo program, when two identical spacecraft were developed to allow for mirroring of the spacecraft’s conditions during the voyage. A Digital Twin can be used to mirror the behavior of an asset for various purposes, exploiting a real time synchronization of the sensed data originating from the field-level and is able to decide between a set of actions with the focus to arrange and execute the whole production system in an optimal way [4, 5]. At first, though, this idea was named the Mirrored Spaces Model, before it was later referred to as the concept of a Digital Twins [6, 7].

Digital Twin is not a completely new concept [8]. It is rooted in some existing technologies [9], such as 3D modeling, system simulation, digital prototyping (including geometric, functional, and behavioral prototyping).

Various definitions of Digital Twins appeared over time. These definitions are reviewed by [10] and [11]. At present, the two most widely accepted definitions were given by Grieves and NASA. NASA defined a Digital Twin for a space vehicle as “an integrated multi-physics, multiscale, probabilistic simulation of an as-built vehicle or system that uses the best available physical models, sensor updates, fleet history and other tools to mirror the life of its corresponding flying twin” [12].

Digital twin extends the concept of twin in Apollo project to virtual space, and creates a virtual product which is similar to the physical entity in external appearance and internal nature by digital means [13]. The relationship between virtual space and physical space is established, so that data and information can be exchanged between them. The concept of virtual instead of real, interaction between virtual and real and controlling reality with virtual are visually and intuitively reflected. With the help of this concept, from a small product to a large workshop, even to a factory and a complex system can establish a corresponding Digital Twin, thus building a “living” virtual space [13].

Similar to a digital model (DM), a Digital Twin is a digital representation of a physical object, but features bi-directional and automatic data flow of operational data and feedback between them [14]. Digital Twins have the potential to improve a manufacturing system by predicting future events, such as machine faults or production bottlenecks, based on its current status and expected behavior and then implementing the appropriate responses to optimize performance in near real-time [15, 16].

2.2 Digital Twin: Types and Nominations

As the Digital Twin concept has attracted strong interests from researchers in the late years, several definitions of Digital Twin has taken place as well as multiple lexical nominations of the concept appeared. The given definitions and nominations of Digital twin differ based the level of data integration between the physical and digital counterpart [14].

  • Digital Model (Fig 1-a): A Digital Model is a digital representation of an existing or planned physical object that does not use any form of automated data exchange between the physical object and the digital object. Simulation softwares are good examples of Digital Models. Simulation softwares contain model-based equations that describes the behavior of a physical system and enable offline simulation with manually integrated data.

  • Digital Shadow (Fig 1-b): A digital shadow is a digital representation of an object that has a one-way flow between the physical and digital object. A change in the state of the physical object leads to a change in the digital object and not vice versa. As the Digital Model receives automatically real time data, it behaves as realistically as the Physical Object and one might refer to such a combination as Digital Shadow.

  • Digital Twin (Fig 1-c): If further, the data flows between an existing physical object and a digital object are fully integrated in both directions, one might refer to it as Digital Twin. A change made to the physical object automatically leads to a change in the digital object and vice versa.

Fig. 1.
figure 1

Digital Twin nominations according to its degree of automation.

Full size image

These three definitions are only to be taken into consideration and not always reliable. One might consider a Digital Shadow as a complete Digital Twin as it already twins the behavior of the physical object associated to. Also this identify the common misconceptions seen in the literature about Digital Twin. Amongst the misconceptions are the misconception that Digital Twins have to be an exact 3D model or just a 3D model of a physical thing [17]. Not to forget the misconceptions and confusions between Digital Twin and Simulation.

2.3 Simulation vs. Digital Twin: What’s the Difference Between Them ?

There are different views on the relationship between Digital Twin and other concepts such as Simulation [18]. Simulation is a digital model that is commonly used during the design phase and often uses computer-aided design software applications. Digital twins, on the other hand, are virtual models created to accurately represent existing physical objects. The physical object is fitted with sensors that produce real time data about different aspects of the object’s performance. This data is then relayed to a processing system and applied to the models encapsulated in the Digital Twin [19]. These digital models are used to run simulations, study current performance and generate potential improvements to be applied back in real time to the actual physical object.

While simulations and Digital Twins each use digital model to replicate products and processes, there are a few key variations among the two. The most important difference to be noted is that a Digital Twin creates a virtual environment able to study several simulations, backed up with real-time data and a two-way flow of information between the physical and digital counterparts, throughout the sensors and actuators that collect and receive data respectively [18]. Moreover, Simulation only focuses on what could happen in the real world (What-if scenarios), in other hand Digital Twin focuses more on what is currently happening by giving a high-fidelity representation of the operational dynamics of its physical counterpart [19]. An other difference is that Digital Twin is considered as active compared to simulation which is static. A Digital Twin will begin much the same as a simulation model, however the introduction of real-time data means that the twin can change and develop to provide a more active simulation.

3 Digital Twin in Mineral Processing: Values and Benefits

The last thirty years have seen a dynamic shift in mineral processing. After years of extraction and exploitation, new deposits nowadays have lower grades, are more difficult to process, and have more complicated mineralogy than older deposits. Moreover, lower grades require greater tonnages to be treated to deliver the same amount of product, which means higher chemical reagents, energy and water consumption. The following section will showcase how Digital Twins, throughout advanced process monitoring, fault diagnosis, self-control, intelligent soft sensing, decision making and value-chain optimisation could stand face to the complexity and variability of the industry and optimise the operational activity of mineral processing.

Real-Time Remote Monitoring, Optimisation and Control. Digital Twins by to their very own nature provide a real time mirroring, prediction and optimisation as a smart control strategy. Throughout sensors and actuators, Digital Twins enable the remote processes monitoring without having to physically be present at specific locations within the plant. The remote monitoring generally involves viewing the operations through a screen or dashboard showcasing the Digital Twins, getting predictions and visualisations of the overall operations performance and sending back actions to be done to the physical process in terms of optimisation and redress. Digital Twins will thus enable the process monitoring taking into account all variations that impact performances and that could harm or vary the outcomes in terms of yield, quality and cost.

Scenario and Risk Assessment. Not to perturb the system, Digital Twins enable what-if analysis by performing and studying in the virtual world what we can’t do, or takes a lot of efforts, time or is very costly and risky, in the physical world. Which is helpful to synthesize unexpected scenarios and study the response of the system as well as the corresponding mitigation strategies.

Predictive Maintenance and Scheduling. Through data captured from sensors and a smart analysis, faults in the system can be detected much in advance. Beyond testing the viability of a component, Digital Twins help with maintenance operations as well. The ability to monitor a whole process using Digital Twins brings additional advantages in terms of planning operational events and improving maintenance strategies. In a situation in which an equipment is expected to fail soon, Digital Twins can assess how this will affect the efficiency of the process and what it will cost.

More Efficient from Operators and Informed Decision Support System. Availability of quantitative data and advanced analytics in real-time will assist in more informed and faster decision makings. The power of the Digital Twin in manufacturing is to facilitate the decision-making process and to enable decision automation through pre-simulation [20]. With easy to use and understandable interfaces, operators and engineers are more likely to monitor remotely the systems, and teams can better utilize their time in improving synergies and collaborations leading to greater productivity. Moreover, being able to explore across the multiple disciplines and fields; e.g., process control, maintenance and R &D, the Digital Twins adoption paves the way to discover a new strategy for operational excellency towards better performances in terms of costs, energy and reagents consumption, recovery, quality and environment impact.

4 Generic Multi-layered Architecture of Digital Twin in Industrial Field: Use Case Mineral Processing

4.1 Data Circulation Between the Physical and Digital Counterparts

One thing that binds most definitions of Digital Twin on literature, other than being a virtual representation, is the bidirectional transmission of data between the physical and digital counterparts. The data to be circulated between the two twins includes quantitative and qualitative data connected to materials, production, process, environmental data, sensors data [21], as well as model’s output data. The journey of interactivity between the physical and digital worlds underscores the profound potential of the digital twin. Thousands of sensors taking continuous measurements that are streamed to a digital platform, which, in turn, performs near-real-time analysis to optimize a business process in a transparent manner.

Fig. 2.
figure 2

Manufacturing Process Digital Twin model.

Full size image

The data-centric model in Fig. 2 illustrates the enabling components (sensors and actuators from the physical process, data, and analytics) as well as the phases of the continuously updated digital twin application. By using sensors to measure the physical process, scientists and engineers are allowed to collect a wealth of data that can be mapped into virtual models after aggregation. This data-charged model perform analysis and provide insights about how the physical thing will respond to the real world. The Digital Twin then sends back actions to be warranted in the real world, by the way of actuators, in terms of refinement and improvement. A Digital Twin helps understand not only how a product or system is currently performing, but also how it may perform in its futuristic behaviors [17, 22].

4.2 Proposed Multi-layered Architecture for the Implementation

There is not a unique comprehensive framework for a Digital Twin for manufacturing. The most global comprehensive model, presented in [23], consists of five layers covering all requirements to represent the physical space in the virtual space. The cyber-physical store data layer, the primary processing layer, the models and algorithms layer, the analysis layer, and the visualization and user interface layer.

Every other published architectural framework of Digital Twins implementation for manufacturing operations comprises 4 main components (data gathering component, data pre-processing and analysis component, models and algorithms component and user interface component). Based on this, we shaped a generic layered architecture adaptive to process Digital Twin of mineral processing value chain in 4 levels (Control, Monitoring, Knowledge and Management). The Fig. 4 represents the proposed architecture and illustrates the levels in question and conceptualizes the data flow circulation between the different layers. In this architecture, the models and algorithms component, from other published frameworks, is divided into predictive and optimisation layers. As shown in the Fig. 4, the shaped architecture comprises a layer for physical systems, a Control Layer, 3 core layers of the Digital Twin (Supervision and Monitoring, Predictive and Optimisation) and a Management layer. Each layer of the architecture packs a specific service and exchanges data with its neighbor layer. Compared to other ones, This shaped architecture handles the data workflow more efficiently and ensures the overall scalability and interoperability of the Digital Twin.

Fig. 3.
figure 3

Digital Twin for Mineral Processing: visualized architecture and data flow circulation.

Full size image

The Control Layer bridges the physical world and their digital counterparts. Controllers and calculators (PLCs, DCSs) are often deployed for the automatic control and are directly linked to sensors and actuators making it possible to know and ensure the continued operation of the production in stable conditions. Besides regulation purpose, controllers transmit information the upper Monitoring Layer. The Monitoring Layer consists of the virtual representation (mostly 3D) of the gathered process information for advanced supervision. The Monitoring Layer, in turn, sends process information to the Knowledge Layers (Predictive and Optimisation Layers) to perform predictive analytics and run optimisation algorithms based on the actual state of the process. The Predictive Layer performs futuristic predictions of the actual operations and forwards them back to the Monitoring Layer for supervision and to the Optimisation Layer. The input data fed to the Predictive Layer are mostly the fusion of processing data from sensors and other information from external data sources such as environmental data and equipment sizing. The Optimisation Layer, in other hand, represents the core of actuating and decision making. With the aid of the neighbor Predictive Layer, the Optimisation Layer evaluates the KPIs and predictions, compares it with the business goals set by the upper Management Layer and runs optimisations algorithms. Optimisation algorithms return the optimal set points to be applied to the operations in order to redress it and to meet the final pre-set plant’s objectives.

Following this conceptual architecture not only allows an efficient orchestration of the services and data workflow after implementation, but also helps in the development phase. By setting apart the Digital Twin into units, it becomes convenient to emphasize a component over an other to start the development journey with. Moreover, this architecture is quite generic as it can be implemented at any fields with data provenance from physical systems and can be extended to other manufacturing operations as well.

4.3 Predictive Layer: Modeling Approaches

The Predictive Layer in the Digital Twin proposed architecture (Fig. 4) encapsulates digital models that are interrogated by the Control Layer. These digital models might be developed through different paradigms and approaches; data-driven approach, model-driven approach or a hybrid modeling approach.

Fig. 4.
figure 4

Modeling approaches of Digital Twin’s Predictive layer.

Full size image

Data-Driven Approach and Model-Driven Approach.

Data-driven methods rely on the previously observed data, making it a paradigm that focuses more on data and variable’s correlation rather than the phenomenon’s physical causation [24]. This includes unsupervised statistical methods or cluster techniques, as well as supervised classification or regression techniques. By analysing the historical operational data of previous performances, Machine learning algorithms can spot connections and patterns that may not even know to suspect, discover features with the most impact and learn how an operation or a system behaves. However, aside its ability of modeling and describing the behavior of a system, the data-driven approach is data hungry and requires a good bit of data to get decent results.

The Model-driven approach, in other side, is actually starting with a solid idea of how the physical system works and re-modeling the system by relying mainly on the physical and phenomenological equations. This way of modeling, as it depends on a deep understanding of systems and processes, demand the collaboration of operators and engineers that understand the physical, mechanical, electronic and all appropriate details of the complex system [25].

Hybrid Approach.

Each approach (data-driven and model-driven) shows advantages and limitations, and one way to get the most of benefits from the both approaches is to combine them towards a hybrid modelling approach [26]. The combination of data-based and knowledge-based modelling is motivated by applications that are partly based on causal relationships, while other effects result from hidden dependencies that are represented in huge amounts of data [27].

The data-driven approach could assist the model-driven approach in various ways. Tiny Machine Learning-based algorithms could be advantageous for the inference of complex measurements that are hard to measure or that are not bound to be directly measured. Statistical methods could also play a major role in studying and investigating hidden parameters dependencies. Moreover, Data assimilation and Reinforcement Learning play an important role in guarantying the Predictive Layer accurate performance after deployment. By continuously injecting observations of the phenomenon, the Digital Twin stays updated preventing itself from data drifts. Inversely, the model-driven approach could as well assist the data-driven approach. In the modeling phase for instance, additional training data could be generated based on physical and phenomenological equations to be fed to Machine Learning models. Furthermore, testing Machine Learning models is less risky if firstly validated in a simulation-based environment that is built with physics and phenomenological equations.

5 Digital Twin in Mineral Processing; Challenges of Adoption

In order to make a Digital Twin completely indistinguishable from its physical counterparts and to ensure the fidelity of the data-centric strategy, many challenges have to be addressed [17].

Mutual Understanding and Integration of Different Domains for the Predictive Layer Modeling

The main objective of adopting Digital Twins in mineral processing is to simulate as realistically as possible the processes and to cover the whole process’s life cycle. The toughest challenges is to own a deep understanding and knowledge across the different domains involved in that process and to encapsulate this knowledge into virtual models. Process Digital Twins must be including all scenarios and changes occurring to products, the key parameters resulting these changes and the complexity of the concerned system. Effectively asset knowledge must include information about the whole production. The abilities and scope of actions must also be clarified by experts in order to come closer to a common optimum.

Data Variety, Security and Connectivity

The amount of data collected from the numerous endpoints existing in a mineral processing plant is massive and various. Such a large variety of data raises the data integration, data cleansing, data fusion and data storage issues [29]. Also, each of the endpoints represents a potential area of security vulnerability. Therefor, manufacturer must be careful not to rush into the adoption of Digital Twins without assessing and updating their current security protocols and access privileges. Moreover, the concept of twinning requires a near zero low latency for the information circulating. It is necessary then to take into account the connectivity and the various data flow control challenges, ensuring that it can be organized and used efficiently to maintain promising, fast and secure delivery.

User Friendly Interfaces for Information Exchange

Human actions and interactions with machines in the production environment are prone to accidents. Human-Machine interaction is a key aspect in adopting Digital Twins for smart manufacturing, which primarily focuses on the issues of communication, interaction, and cooperation between engineers and machines to furnish a remote mode of visualisation and control. Therefore, the efficiency and safety of a Digital Twin relies mostly on how the information will be displayed to the operators in the simplest way allowing an accurate readability, providing understandable decisions and enabling easy remote monitoring.

6 Digital Twin in Mineral Processing: Underpinning Technologies

Internet of Things

The first enabler of the Digital Twin concept is the internet of things (IoT), and specifically industrial internet of things (IIoT) for manufacturing environment. IoT/IIoT devices are made up of sensors, actuators, controllers and Networks. IoT/IIoT have a huge impact on improving manufacturing processes, allowing for tasks to be evaluated with greater knowledge and providing the remote supervision and control through connected devices. From Digital Twin perspective, IoT/IIoT bridges the physical world with its virtual twin. The online data collected by the sensors will eventually serve the virtual twin as an entryway for algorithms to be processed, then control signals are generated for the actuators according to the actions needed to perform.

Machine Learning

The vast majority of target systems have multiple variables and multiple streams of data. Machine learning (ML) algorithm can be trained to recognize process behavior by feeding it thousands of historic observations that have been linked to the process performance. After training and testing, the ML algorithm can then be put to work monitoring the system and alerting when it observes suspected abnormal behavior or deviation. Furthermore, Digital Twin’s consistent efficiency after deployment requires an active monitoring of its performance. Therefore, Reinforcement Learning comes to automate this process by continuously feeding observations for active learning and accuracy regulating.

Heuristic Optimisation

Given their ability to process large datasets based on user-defined objectives, heuristic algorithms can perform complex optimisation using heuristic methods [28]. With a pre-set objectives, system’s behavior equations and the current process state, heuristic algorithms are used to automatically compute optimal values of control variables that might drive the system from its current state to the pre-set objective. These optimal values are systematically sent afterwards to the actuators to act in the physical process in terms of optimisation.

Big Data Technologies

Sensors certainly generates a huge amount of information, which must be stored, treated and analysed for the benefice of the digital twin. The infrastructure for storing and processing high volume data has been advanced considerably over the last decades. And many available platforms are available to handle big data projects in terms of integration, storage, centralized management, interactive analysis, visualization, accessibility and security.

Cloud Computing and Networks

Collecting and transmitting information and ensuring a real time communication and latency represent a big challenge. Cloud and Edge computing, 5G networks and Data encryption techniques are certainly expected to play a major role pushing towards facing these challenges and guaranteeing the efficient interactive aspect of the Digital Twin.

Encryption - Data Protection Techniques

Information and data security are a continuous critical concern for every industrial business. Digital Twin creates valuable information that needs to be protected. Today, different methods and techniques are developed to ensure data safety and protect its availability only for authorized persons and systems.

7 Conclusion and Perspective

Materials and minerals processing Industries are bound to increasingly adopt digital technologies in order to ensure product safety and quality, minimize costs in the face of low profit margins, shorten lead times and guarantee timely delivery of an increasing number of products despite production dead times and uncertainties. The concept of a digital twin put forward in the context of Industry 4.0 encompasses digital models of the production model that imitate the physical system, perform prediction and interacts with it in terms of optimisation.

The Digital Twin concept transforms data in real time into operational information allowing an advanced real time supervision and control. The implementation framework of a process Digital Twin for mineral processing follows a layered architecture; Control Layer bridging between the physical process and its digital counterpart, Monitoring Layer for the supervision, Predictive Layer encapsulating data-driven and model-driven equations, Optimisation Layer encapsulating optimisation algorithms and Management Layer providing goals and objectives to meet.

As part of the Moroccan National Project “Smart Connected Mine”, this topic is an applied research and development axe that aims to provide a Digital Twin prototype to digital disrupt the mineral processing control system. This paper addressed the background, challenges and potential benefits of implementing Digital Twin for minerals processing units, as well as an architecture of the implementation is proposed. The further work will mainly be focused on the development of a flotation process Digital Twin as prototype. The model development and validation of the Digital Twin and the challenges encountered will be tackled in next papers.