Given the definition of smart retrofitting, it is possible to search the literature for reasons why to apply a retrofit or smart retrofit action, and for solutions and benefits to be found.

First, it is good to define what steps need to be taken to properly implement retrofit/smart retrofit actions.

Tantscher et al. ^[1][49] propose a holistic methodology for implementing digital retrofitting from an Industry 4.0 perspective. They identify four superordinate aspects that form the frame of the process: strategy, interdisciplinarity, change management and lean management. A direct relationship relates the four superordinate aspects and the operational process. The procedure includes the following steps:

• IIoT Use Case: optimising asset maintenance, increasing the efficiency of assembling processes and improving machine monitoring.
• State analysis: an analysis of the current state.
• Action planning: specific actions depend on individual requirements and the specific use case.
• Implementation: installation of all hardware and software components and virtualisation of physical objects is necessary.
• IIoT integration: platforms for data visualisation and analytics.
• Verification: if the correct data are measured/transmitted correctly.
• Validation: compliance with the predefined usage targets is checked.
• Standardisation.

Following the eight steps makes it possible to properly evaluate and carry out a machine/plant retrofit. Specifically, with the definition of an IIoT use case, the company outlines the target to upgrade the actual state of the system. Often, implemented use cases aim to optimise the maintenance of assets (condition-based maintenance, predictive maintenance), increase the efficiency of assembly processes (augmented reality applications) and improve machine monitoring (condition monitoring, energy management, Digital Twin) ^[1][49]. This can be simplified into four steps for legacy system renovation. An article by Stock and Seliger ^[2][5] discusses a retrofit action on a legacy system in four steps:

• situation analysis;
• definition of the monitoring strategy;
• data processing;
• implementation of the equipment in a CPS.

For a milling machine retrofit, a Beckhoff PLC converts analogue signals from the acceleration sensor into digital signals for subsequent data processing. The data processing takes place. A human–machine interface realises data visualisation. This milling machine can now be implemented in a CPS. In connection with an intelligent product, the retrofitted machine can program the material flow in a decentralised manner. It can react automatically to machine failures, for example, by informing the responsible worker. Ilari et al. ^[3][38] use a similar implementation methodology to the one above. They conducted a study comparing retrofitting dated machines with the purchase of new ones, focusing on the sustainability aspect (economic, social and environmental). The framework was implemented in five steps: “objectives analysis,” “market analysis,” “retrofitting analysis,” “collecting information” and “selection of the best solution through the AHP”. The evaluations are applied to the case of a drill press. The hardware and software technologies are simple and inexpensive, allowing drill management through a smartphone application. In addition, retrofitted machines demonstrate a significant sustainability advantage over new machines.

Retrofit brings innovation not only to the machine level; sometimes it can be seen as adapting to new market demands and improving working conditions.

Bakir ^[4][11] and Ehrlich ^[5][12] et al. see retrofit as an activity for old machines and an opportunity to make new machines more versatile. This characteristic is necessary because the way of doing business is changing: production quantities, e-commerce and online sites revolutionised business models and customer demand for customisable products. This aspect is crucial when it comes to small and medium-sized enterprises, which encompass people and machines not related to the Industry 4.0 paradigm, generally present a “climate of distrust” in digitisation, cannot bear the onerous costs and long lead times of the eventual replacement of an entire production asset and sometimes lack know-how. On the other hand, sometimes retrofit actions can be the first step toward Industry 4.0 for a small and medium-sized enterprise. Leona Niemeyer et al. ^[6][3] report a simple and inexpensive way to bring companies (especially small and medium enterprises) into compliance with Industry 4.0 standards. They show a simple system integrating smart sensors, IIoT platforms like Thingworx and Kepware and condition monitoring software. These tools enable SMEs to develop new business models and be more competitive for a low investment.

Similarly, Keshav Kolla et al. ^[7][43] report an example of retrofitting that can be applied to SMEs because they have limited resources and implementation time. In this case, a system is implemented (experimentally) to integrate new functions within old machinery. In fact, at the physical level, new sensors and smart devices are installed, which are then connected to PLCs and/or microcontrollers (e.g., Raspberry Pi). Communication is possible by following the UPC UA/ethernet protocols, which communicate with the IIoT layer. This, in turn, is connected with the cyber layer, where the databases and monitoring/analysis systems (implemented in MATLAB, Grafna, InfluxDB) are located.

A case study, reported by Burresi et al. ^[8][17], concerns Fonderia Gelli, a steel mill producing industrial cast iron and steel components. The first step in the configuration is to interface with the Samsung S7 PLC system to enable a connection via the OPC standard. A proprietary Siemens OPC-DA server runs on a Windows machine—a UDOOx86 board—allowing a custom Python application to obtain data from the PLC and forward them to remote services for data analysis. An IP camera is also connected to the gateway to allow external services to access the video stream. It is supported by imaging software and computer vision (CV) algorithms. This makes it possible to visualise part defects, stop production and report anomalies to the operator, who can take corrective action based on the information reported on the user dashboard. An article by Generosi et al. ^[9][16] also discusses a retrofit action for a worker identification system to prevent workplace accidents and improve worker conditions. The hardware retrofit uses Raspberry Pi 4, a touchscreen display, an embedded webcam and LEDs.

As mentioned, different drivers require smart retrofit action on the old assets. Sometimes the necessity to introduce the I4.0 paradigm in the system with limited investment makes it possible to also apply smart retrofit actions to new machines that are not 4.0.

As mentioned above, SMEs often do not have the time, money or knowledge to completely change their business assets. At the same time, they are faced with the need to introduce the Industry 4.0 paradigm. To realise this step, they need a solution such as smart retrofitting.

In an article written by Mazzuto et al. ^[10][50], these factors are evaluated, and the implementation of a Digital Twin and machine learning algorithms are identified as the solution; in this way, plant failures can be predicted and detected. This approach requires limited investment and reasonable time (suitable for SMEs because it uses “economic” algorithms, software and a connected sensor for control and monitoring), enables innovation, fits into I4.0 and allows an old plant to be transformed into a smart plant. In addition, the plant can quickly communicate with operators to increase their safety and aid in their decision-making.

Ruppert T. et al. ^[11][55] also point out in their article that manufacturing industries, especially SMEs, often lack the knowledge and resources to develop I4.0 solutions. In these situations, a smart retrofit can provide a quick and cost-effective way to improve productivity and competitiveness. For these reasons, they propose Digital Twins of several old machines to retrofit them and extend their useful life.

A case proposed by Hassan could be a retrofit solution applicable in the context of an SME: the article shows the construction of a DT of a drilling machine. However, the solution requires the use of simple and inexpensive devices such as Arduino, anomaly detection algorithms, a camera for operator recognition and other cheap and easy-to-use technologies adequate for an SME. Furthermore, the application of a “digital triplet” can also answer another driver: the possibility of reusing the old asset and valuing the experience of workers can create a climate of “acceptance” towards adopting Industry 4.0 technologies and put the workers in a “central position” in the company.

Another aspect is sustainability, particularly environmental sustainability. In order to be able to reuse a machine, it is possible to apply retrofit actions through the construction of a DT and a “digital triplet”. Burresi et al. ^[8][17] report an example that used the same PLC (Siemens PLC S7), Raspberry Pi, camera and OPC UA standard of the article reported in the section before by Michael M. et al. ^[12][56].

Ermini et al. [13] analyse the relationship between man and machine in the industrial sector (human-centred retrofitting) and how introducing data collection devices, sensors, Digital Twin, PLC, and machine learning algorithms represents an opportunity to redesign this relationship in production processes. The proposed DT implementation solution aims to allow the plant to increase efficiency, decrease machinery faults and produce rapid and low-cost technical improvements. This type of implementation enables environmental sustainability and social sustainability for workers.

Mazzuto ^[10][50] and Bevilacqua ^[14][57] propose similar considerations and developments: the approaches enable predictive maintenance applications. In this regard, the proposed Digital Twin can create virtual modelling of maintenance processes, thus preventing high-risk events for operators.

The importance of desktop/Web applications and human–machine interfaces is emphasised by García et al. ^[15][14]. There is a modular infrastructure that is composed of different microservices to store and process data. All information from the various tiers is connected using Web APIs. There are three parts: a portable IIoT infrastructure with non-intrusive sensors, heterogeneous data streams to the edge tier and software interfaces; a cloud-based service architecture, hosting a common information connectivity layer and data models to the cloud tier; and end-user human–machine interface (HMI) management, with interactive human–machine software tools and asset health condition-based strategies providing knowledge models to the business tier. The DFM (data flow management) module contains information about system configuration and status, real-time data, data analytics, augmented contents and dashboards. This module is valid for anomaly detection analysis and advanced system monitoring. The results of these analyses and all data can be displayed in real-time on the operator’s device. The technology implemented in an article by Kumar et al. ^[16][9] proposes to retrofit an analogue water meter. Such a study could be helpful and re-applicable to industrial systems with similar devices. An external structure is built to be mounted on the analogue meter with an LED, Raspberry Pi and a router for internet connection and data transmission. The software component is added to the hardware implementation using a deep learning algorithm for image recognition. The algorithm uses a convolutional neural network based on the Visual Geometry Group (VGG) architecture. The system allows real-time and remote readings of meter data.

Panda et al. ^[17][36] apply a retrofit regarding hardware, communication and the cloud. They implement a device layer and integration layer. The first layer comprises one or more field devices attached to sensor nodes. The second layer of the architecture consists of a physical device capable of communicating with the sensor nodes that will provide the value of process parameters. OPC UA is identified as the ideal candidate for middleware communication protocols as it allows for platform-independent, service-oriented architecture with a standardised information model. AWS Greengrass facilitates interaction between Raspberry Pi and the AWS IoT core. Further, the acquired data can be analysed in AWS IoT analytics and used for forecast learning and predictive analysis. For a food company, the quality of ingredients can be determined and enhanced by adding these features to reduce waste, resulting in cost-effectiveness.

All these techniques for secure transmission, visualisation and storage of data related to retrofitting can also be applied when developing a DT, as mentioned in Hassan Alimam’s article ^[18][53].

Chang, in his article ^[19][58], examines a crucial aspect that characterises Industry 4.0: IIoT and security issues in data transmission. The report highlights how retrofitting activities can also focus on implementing more secure connections and integrating security systems.

Arjoni et al. ^[20][39] report examples of retrofits of different devices. In these cases, they use devices such as Arduino Uno, Raspberry Pi 3, RFID (and related readers), communication protocols such as OPC UA and UPC and software such as Labview and Tecnomatix. They conclude that the retrofit strategies have a good cost-benefit ratio, permitting old equipment to be used as elements of advanced manufacturing with little programming and mechanical adaptation effort. Haskamp et al. ^[21][31] aim for a Flexible Manufacturing System by exploiting PLC, sensors, four different cameras, RFID, UPC UA communication protocols and RAMI 4.0.

In addition, a digital interface has been developed that allows the operator to check the system status, real-time data, historical data and system control. Also used is a dataFEED OPC Suite that allows writing/reading to PLCs and Azura IoT Edge to run programs and cloud services. Kofi Atta Nsiah et al. ^[22][26] describe how the NIKI4.0 Toolkit is implemented and how it can enable process monitoring and real-time visualisation. In addition, OPC UA securely allows data exchange, and MySQL allows the storing of historical data. Finally, Oks et al. ^[23][8] discuss the retrofit activity to be implemented on a PID4CPS (Portable Industrial Demonstrator for Cyber-Physical Systems). They equipped the PID4CPS with a camera, a Raspberry Pi system, UPC UA communication protocols, imaging software and neural networks. All combine to return a system capable of autonomously detecting the degree of shelf filling.

Strauß et al. ^[24][28] focus on low-cost retrofits, especially concerning sensors (e.g., RFID). Arduino, Raspberry Pi and Banana Pi are mentioned as embedded systems that can interface with different types of sensors and cloud services. The cloud service can be implemented with Microsoft Azur IoT Device SDK. Already known architectures such as RAMI, ARM or IIRA are also reported for IIoT architecture implementation. Again, Azure Cloud is mentioned in the part concerning the cloud platform.

A cloud data collection system then provides the input to the “data processing, analytics and management” systems in which storage, stream processors, analytics and machine learning, business integration connectors and gateway components are present that then connect with the “presentation and business connectivity” layer for integrating personal mobile devices and business systems. Etz et al. (2020) also reported similar considerations and implementations ^[25][44] in their article.

The use of neural networks, machine learning algorithms and artificial intelligence is also applied in several cases. The use of neural networks, machine learning algorithms and artificial intelligence is also applied in several cases. The use of machine learning algorithms can can be employed to solve common retrofit-related problems. In their article, Pupăză et al. ^[26][7] analyse a prevalent problem when performing retrofits or smart retrofits: overheating of systems. Overheating can lead to system damage and incorrect evaluations of acquired data. Then control logic and machine learning algorithms based on deep learning neural networks are implemented. The acquired data are then processed and graphed onto a model. This application makes it possible to extend the life of electronic systems, reducing overheating and subsequent failures.

Since there are many cases in the literature where the Digital Twin represents a retrofitting solution, the following is a section on smart retrofit actions through DT.

IIoT, cloud, communication architectures, machine learning and augmented reality are utilised in the “digital triplet”, where, from the interaction of the Digital Twin with human knowledge, it is possible to create more immersive virtual spaces, more user-friendly interfaces and simulative environments in which one can insert avatars of one’s operator and work safely. These possibilities enable various aspects such as environmental, social and economic sustainability, greater security in the training and work of the worker and more. The digital triplet of the four-level hierarchy was used for the retrofit of a drilling machine in the study reported by Alimam Hassan et al. ^[18][53]. The article highlights how the “digital triplet” and DT are intelligent tools for the retrofit of older machines at low cost. The three interactive components of the digital systems are the digital master, which is the anticipated state of the system’s validity and adjustment by devoting (machine) learning; the Digital Twin, which is the duplication of the system where mixed data, information, models, methods, tools and techniques are present; and the digital prototype, which is the envisaged state of the emulated system. The advanced stages of the traditional Digital Twin paradigm are the intelligent activities layer and the master component of the digital system. In order to create an innovative study of the numerous tactics, the “digital triplet” concept, which was created as an implication of intelligent evolution, encourages the fusion of the actual, virtual and intelligent activity worlds with human awareness.

Thus, the “digital triplet” requires a four-level aggregate to the Digital Twins. The hierarchy was created using the four levels of complex advanced decision-making, using machine learning based on human intelligence and creativity, the control of physical system behaviour prediction and emulation, maturity in terms of iterative observation of actual physical system behaviour using real-time data and a level for visualising and simulating virtual features. The hierarchy clarified the intelligent actions based on transforming and extracting the human awareness required to advance the digital triplet paradigm’s apex, improving convergence and facilitating the exchange of data and information across cyberspace, physical space and people.

The digital triplet decreases financial risk while improving the viability, security and resilience of smart retrofitting. Through smart actions, the Digital Twin enhances human–machine integration and skill transformation. The evaluation showed that the digital triple hierarchy could be retrofitted to user-preferred intelligent solutions, utilising the creativity of human experts working with intelligent Digital Twins and reducing the complexity of the paradigm through interaction between various levels of the digital triplet and human awareness, as well as by clarifying the functionality of pertinent information for each Digital Twin across all levels. All are applied to an outdated drilling machine on which new sensors, devices and cameras are applied, and on which a Digital Twin is built. The control is realised with an inverter (Schneider electric Altivar machine ATV320). At the same time, the data transmission is done with an Arduino Mega with an ethernet shield for control and data transmission with an ethernet protocol. To collect vibration data, a vibration sensor with Arduino Nano 33 IoT and wireless transmission protocol was installed in the machine. A node-based JavaScrip server enables real-time data transmission between the physical and digital parts. MySQL was utilised for data transfer and as a repository for digital data produced by the camera and IoT subsystem. AI-based decision-making and anomaly detection systems were built to analyse the behaviour of vibrations during machine operation. The intuitive human–machine interface and the Android application for identification were developed and connected to the system.

In their article, B. Ralph et al. ^[27][22] report on the construction of a six-level digitisation architecture that can be integrated into a CPS to perform a smart retrofit on older machines and transform them into a low-cost and user-friendly system using machine learning algorithms. Building a DT and applying smart sensors and controllers make it possible to access condition, process and management data and realise a perfect low-cost, user-centred CPS.

Mazzuto et al. ^[28][59] also used artificial neural networks (ANNs), swarm intelligence algorithms and a smart sensor plant to make a Digital Twin (DT) and monitor the operating status of a multiphase flow ejector. These algorithms allow for analysing the ejector’s operation according to the set conditions and evaluating malfunctions.

Many other authors see the Digital Twin as an effective retrofit solution. For example, an article by Hegedus ^[29][27] shifts the focus to the supply chain and warehouse, analysing the relocation and storage of raw materials, work-in-process inventories and even the shipping of finished products. The systems used are Real-Time Locating Systems (RTLS), the Indoor Positioning System (IPS) from AITIA, Ultra-Wide Band (UWB) ranging technology, Bluetooth Low Energy (BLE), Radio Frequency Identification (RFID), Arrowhead SOA systems and a local cloud. These systems solve many problems regarding workstation automation, tracking of the entire supply chain, the possibility of creating a Digital Twin for manufacturing production in the best possible way and application within existing facilities without disrupting machinery and production. Moreover, in this way, it is also possible to exchange helpful information with other systems.

In their article, Di Carlo, Mazzuto, Bevilacqua and Ciarapica et al. ^[30][2] set goals to improve working conditions, achieve better quality processes, achieve better communication and collaboration, improve productivity, efficiency, flexibility and agility and reduce costs. They seek to implement a system capable of anomaly detection to achieve most of these goals. The anomaly detection software is implemented through data from a Digital Twin. In fact, as a first step, a DT is implemented through a supervised approach. DT allows the behaviour of the real plant to be replicated in a digital environment. Then, knowing the standard behaviour under various conditions, it is possible to evaluate the abnormal operating condition (with proper tolerances) and report it. The Digital Twin then makes the digital counterpart of an existing physical system. A description of how a Digital Twin can be developed is provided by Bevilacqua et al. ^[14][57].

It may be more useful to leverage or combine augmented reality technologies. For example, in a case described by Mazzuto et al. ^[31][54], more innovative technologies, such as IIoT (combined with connections, HMI and cloud systems), Digital Twin and augmented reality, are implemented.

Mourtzis et al. ^[32][42] propose a solution designed to work with Microsoft HoloLens that integrates a machine learning algorithm. They define three levels for the implementation of the framework. The first layer includes the back-end application, which contains the infrastructure, the communication protocols used and an online database for handling all the data needed for the framework operation (the Cloud SQL database was used for cloud information management). The second layer consists of the front-end application used by the users. Finally, the third layer identified is the augmented reality (AR) module, which combines all the pertinent data to visualise the digital models in the user’s physical environment. The main benefits found were low-cost implementation, extended machine life cycles and implementation of AR systems for improved performance.

Hassan Al-Maeeni et al. ^[33][4] also implemented the same technologies for the same purposes but in a real manufacturing plant.

A similar structure is presented by Mazzuto et al. ^[31][54]. The article describes how a pair of smart glasses (Vuzix Blade) can guide an operator through a facility and its status. Specifically, a notification is shown on the lens that provides an alert about the status of an element of a plant and the actions to be taken. Additionally, Settimi et al. ^[34][52] talk about a retrofit applied to a machining bench where a drill is used to drill holes. In addition, using a handheld IR-based scanner (FARO^® Frestyle2), the timber stock (in batches) was digitised to obtain digital replicas of physical objects. Integrating the drill retrofit within the technologies installed on the wood pallets makes it possible to achieve automated workflow flexibility and simultaneously provide an asset digitisation opportunity to small companies with low-level entry technology by leveraging the already present know-how and pre-existing manual tools.

In addition, in the context of Digital Twin, the use of Industry 4.0 reference architectures such as RAMI 4.0, NIKI 4.0 and OPC/OPC UA is reported on. Lins et Oliveira ^[35][10] propose a method for retrofitting within manufacturing plants a CPPS (cyber-physical production system). The entire process is based on Reference Architectural Model for Industry 4.0 (RAMI 4.0). First, the infrastructure of the company is evaluated and an attempt is made to implement an embedded board component that can support various connections and IoT device components. After that, they move to the communication level, where they must evaluate the network component and protocols to allow them to adapt to the OPC/OPC-C/SDN-C communication standards. Last but not least, functional elements of the CPPS are implemented: a database component; a remote access component; a Web service component; a monitoring component; and a cloud component.

Many benefits have already been mentioned above or can be inferred from retrofit applications. However, it is possible to explore them further in this section by answering an additional question: why should a company adopt retrofitting or smart retrofitting solutions?

The previous section discusses different aspects that drive a company, not necessarily an SME, to undertake retrofit actions. Regarding new business models, it has been shown that there is a growing trend to produce more customised products; such products have, in principle, the same characteristics, but they can vary in secondary aesthetic features (customisations). Sometimes it is costly for a company to purchase different machinery to enable customisation. On the other hand, not having this option makes the company less competitive. Therefore, one can think of applying a retrofit or smart retrofit activity to new machines as well. Additional mechanical devices with related control systems and software could be integrated (imaging software and connected cameras) for smooth production management of the same essential product, with different customisations but the same quality.

Moreover, “new business models” also mean the quantity produced. In this regard, being able to sensitise a machine or an entire system, build a Digital Twin or simply monitor the performance of a machine/system allows people to understand what the critical points are and the areas where there is a need for maintenance, and to properly manage resources (both material and energy). It is precisely for this reason that it is possible to say that the Digital Twin appears to be one of the most cost-effective smart retrofit solutions, as it allows for the integration of anomaly detection systems, simulation systems and communication systems with other Industry 4.0 devices within old machinery.

Moreover, among the new business models, the recent trend of selling a service instead of a physical product should be noted. Technology is advancing day by day, and it is difficult for those who buy machinery to be able to keep up. Generally, a company buys machinery carefully, planning depreciation and machine usage time. Therefore, it might be a good business strategy to provide, along with machinery, a series of “upgrades” based on retrofit actions. All of this should also be conducted with a view to savings and sustainability. Retrofitting enables social sustainability in the sense of the well-being of the worker who is accustomed to working with old machinery: if old machinery is integrated with smart devices, the operator will be able to gain benefits in terms of decision-making, equipment monitoring and safety, but without having to disrupt his knowledge and skills. In particular, in the case of “digital triplet” integration as a smart retrofit action, the knowledge of the operator and engineers is crucial for constructing technologically advanced and high-performance systems.

Retrofitting also refers to economic sustainability: it brings benefits on the production side because new production, quality and resource management functions can be enabled, but it also brings innovation on the implementation side of these technologies. Ilari shows that retrofit actions cost far less than purchasing a new machine ^{[3][17][36][37]}[35,36,37,38]. Often, the issues of efficiency and time-to-market in developing, experimenting, testing and deploying new products are the preserve of large companies. However, this scenario, thanks partly to the cloud, is changing. At this time, detailed analysis of risks, knowledge of possible threats and the ability to seize all simulation development opportunities with the utmost precision appear increasingly within reach of the SME world ^[38][60]. Retrofitting also enables environmental sustainability because there is no need to dispose of old machinery to purchase new machinery: technology is implemented directly all over the dated machinery. In addition, retrofitting encourages environmental sustainability, especially with the construction of Digital Twins or simulative spaces: they digitally simulate system operation and performance without wasting resources and time. Therefore, applying retrofit or smart retrofit actions on a machine or system can bring many benefits, especially since having smart sensors, software systems and DT systems enables the extension of various functionalities without requiring great economic or time efforts. The capacity to create from this retrofit action a product or, better yet, a service could be the key for many companies to free themselves from building a physical asset in favour of a more digital one.