Even after three decades of research in modernizing legacy systems, it is notable that many remain in operation. This is due to the fact that these systems are generally very comprehensive (Brooke & Ramage, 2001; Ramage, 2000). They interoperate with other processes or subsystems, only remain in operation due to their technical complexity of replacement and/or adaptation and criticality in the organization's operations, in such a way that remains in constant activity. Every system is likely to become a legacy at some point and its data is characterized as valuable since its history can be used to understand its behavior in search of optimization (Batlajery et al., 2014). However, to remain competitive, companies must continually change their processes, sometimes radically, and legacy systems can delay modernization processes and directly influence the company's business strategy (Liu et al., 1998; Matsumoto et al., 2020; Moeuf et al., 2018; Morariu et al., 2016).

Two architectures were bases to allocate legacy systems into the conformities of I4.0 in a coordinated way. They adopt structures that organize evaluative attributes in perspectives that portray the adequacy of maintenance systems, considering their interoperability barriers.

The first, Framework for Enterprise Interoperability (FEI) (Chen et al., 2007), was considered by the premise that interoperability might be a relevant metrics to understand what can or cannot be implemented to a system. This possibility is feasible because FEI relates conceptual, technological, and organizational barriers linked between the enterprise layers, that could be generated by systems trying to communicate. Coupled with that, the prerogative that interoperability barriers could difficult the insertion of technology seems feasible once legacy systems and other adjacent systems/processes may share communication dependence.

The second is the Reference Architecture Model for Industry 4.0 (RAMI4.0) (Plattform Industrie 4.0, 2015), converging multi-stakeholder views on how I4.0 can be accomplished based on existing communication standards and functional descriptions (Pedone & Mezgár, 2018). Analogously to the FEI, the RAMI4.0 presents a similar enterprise’s layers perspective. Considering that this research investigates interoperability barriers that might appear by implement I4.0 technologies in legacy systems, those frameworks were compared (see Figure 2).

Figure 2
FEI barriers x RAMI4.0 layers compared frameworks.

This composition considers interoperability barriers into an I4.0 referential architecture. The following subsections explain, firstly, how this relation generated a maturity As-is view of a legacy maintenance system, and after, how Industry 4.0 technologies could enhance that system, expressed in a To-be view.

2.2.1. System maturity for Industry 4.0

The authors propose a maturity view through the lens of RAMI4.0/FEI architecture in early studies. It aims to understand maintenance systems' maturity by the relations between its attributes and functional requirements (Cleland-Huang, 2007). The present work defines Attribute as something that qualifies a concept, in this case, maintenance. The definition adopted for Functional Requirement is something that supports the Attribute to which it is related. Figure 3 illustrates how these elements are related to each other.

Figure 3
Relation between attributes and functional requirements.

The purpose of the attributes is to qualify maintenance within the RAMI4.0 layers. Using the Assets layer as an example, the attributes raised have a bias to guarantee the functionality of the acquisition system and to ensure the quality and the way that the sensing in the equipment is carried out. In the case of functional requirements, they must support the attributes, so that they are met. Again, using the Asset layer as an example, the functional requirements are related to the needs of a good sensing system, what should be sensed and what these sensors should monitor. Table 1 presents all 25 attributes raised in the literature and their descriptions, follow by the 62 functional requirements derived from the attributes, therefore using the same literary base indicated by the ID column.

Table 1
Attributes and Functional Requirements description and its references.

Those maintenance attributes and functional requirements give the present work directions in how to analyses maintenance systems in an a priori state (i.e., As-is).

The legacy systems addressed in this work were constrained to industrial maintenance. Maintenance is currently seen as a complex management process that combines several organizational processes, such as production, quality, environment, risk analysis, and safety. Bearing in mind that nowadays maintenance management is a key part of the organizational composition, it is important to keep its processes in line with the company's strategy. An appropriate maintenance strategy not only reduces the likelihood of equipment failure but also improves the working condition of the assets, resulting in lower maintenance costs and/or higher product quality (Sipsas et al., 2016; Vaisnys et al., 2006). In an exploratory character, a partial review of the literature with three research rounds was carried out, focusing on recent technologies for the maintenance sector.

2.3.1. First research round

The first research round provided a general context of I4.0 technologies. For that, the most cite reports with frameworks already formalized in the literature were used (see Table 2).

Table 2
Technology consultancies and its reports.

The objective was to gain an overview of I4.0 technologies, with the perspective of different technology consultancies.

2.3.3. Third research round

Finally, in the third round, the technologies highlighted for the industrial maintenance were filtered and allocated into groups. The whole literature database ended with 69 articles and reports. From it, nine Maintenance-4.0 technology groups were identified: Big Data, Analytics, Artificial Intelligence and Cloud Computing, formalized as cyber-physical subgroup; Advanced Machines, Advanced Materials, Flexible Connection Devices and Digital-to-Real Representation (i.e., encapsulating Digital Twin applied in maintenance activities), formalized as application subgroup; and Sensors (i.e., encapsulating IoT and Smart Sensors, formalized as the bridge to digitalize physical operations). Table 3 details each group.

Table 3
Maintenance-4.0 technology groups, characteristics and applications.

Various concepts have been developed to increase maintenance effectiveness. One of the most commonly used concepts in organizations around the world is Total Productive Maintenance (TPM). The TPM emphasizes proactive and preventive maintenance to maximize the operational efficiency of the equipment. Production losses, together with indirect and hidden costs, make up the bulk of the total production cost (Kodali et al., 2009). Developed to support TPM initiatives, Overall Equipment Effectiveness (OEE) is a metric that identifies the percentage of planned production time that is truly productive. The OEE loss of availability, loss of performance, and loss of quality can be subdivided into what is commonly called TPM Six Big Losses (Vaisnys et al., 2006), the most common causes of lost productivity in manufacturing.

In order to achieve I4.0 adequacy for the maintenance sector the six big losses were considered (Ahuja & Khamba, 2008). For those losses, the model in Figure 4 formalizes courses of action, meaning that for each loss there is a course of action based on an I4.0 solution.

Figure 4
Maintenance-4.0 referential model based on TPM.

Such referential architecture was based on a digital asset management platform. With operations in more than ten countries and more than 15 years of know-how in the maintenance area, it can be considered a commercially validated source, reliable in defining applications. Because the scientific literature varies widely from organization to organization, this platform was chosen as a tool to define maintenance in Industry 4.0. Moreover, those courses of action are categorized into three main maintenance approaches: predictive, preventive, and corrective (Dhillon, 2002). Therefore, the spheres, or Maintenance-4.0 functions, represent enablers for predictive, preventive, and corrective approaches based on the technologies reviewed in the previous section (2.3). The 32 functions are shown (ranked) as a product of the case study in section 4.

In resume, aiming to guide maintenance processes to zero waste using disruptive technologies, this proposed model serves as a To-be guide, for the presented As-is analysis (2.2.1), due to interoperability barriers. Alternatively, what is needed to implement (i.e., disruptive technologies) according to what is possible to be implemented (i.e., interoperability barriers).

Not used as a theoretical dimension but as part of the scientific scenario in a tooling bias, multicriteria decision making/analysis (MCDM/A) methods emerged in the search for solutions to complex problems that are difficult to measure, already demonstrated in the maintenance domain (Ruschel et al., 2017). This strategy is used as tools for more assertive decisions in systems adequacy, also following a couple of referential researches which applies decision-making to assessment in the dimensions of interoperability and Industry 4.0 such as (Battirola Filho et al., 2017; Lazai Junior et al., 2020).

Four elements characterize MCDM methods: Set of “alternatives”, from which the decision is chosen; set of “criteria”, or factors related to making a good decision; the “preferences” of the decision-maker, being clear, the problem becomes more understandable; and the “result” of each choice, measured in terms of criteria according to the decision maker's preferences.

Two different MCDM are used for the three steps framework, detailed in the next section. For Step 01 and 02, the Analytic Hierarchy Process (AHP) (Saaty, 1987) is used in order to derive priorities based on sets of peer comparisons, thus it is structured on the intrinsic ability to ponder their perceptions or ideas hierarchically (Forman & Peniwati, 1998). This method uses a compensatory characteristic, weighting the positive and negative attributes of the considered alternatives and allowing positive attributes to offset the negative ones (Elbok & Berrado, 2020). This article also explores the AHP possibility to combine geometric means, thus, aggregating the decision-makers responses according to the approaches presented in (Ssebuggwawo et al., 2009).

In Step 03, the Preference Ranking Organization Method for Enrichment of Evaluations (PROMETHEE II) is used. It is characterized as an interactive method designed to deal with quantitative, qualitative criteria, and discrete alternatives. This method can classify alternatives that are difficult to compare due to a commitment to standards of evaluation as non-comparable alternatives (Athawale et al., 2012). It suggests a non-compensatory strategy, eliminating alternatives that do not meet a particular criterion (Banihabib et al., 2017). According to (Brans & Mareschal, 2005) it have been applied in varied fields such as industrial locations, labor planning, investments, medicine, chemistry, tourism, and ethics.

Although the two methods applied are based on different strategies, they meet the evaluative requirements of each step of the proposed framework. Also, the use of a hybrid MCDA approach offers more robust results than isolated MCDA methods (Liou et al., 2017). The next section details the framework.

Once is confirmed the organization's strategy to optimize its systems to an I4.0 scenario, in Step 01 an assessment of its maturity concerning the desired requirements is carried out. For this, engineers and maintainers who know in depth the maintenance processes and systems to be evaluated must be available, answering the proposed AHP model. They will be in the role of decision-makers. Figure 6 reflects such a model by constructing classification structures from the six layers of RAMI4.0/FEI. Working as a maturity assessment, this model describes the decomposition of a machine in its structured properties, enabling its virtual mapping.

Figure 6
Step 01 – Maturity assessment (AHP model 1).

The name of the analyzed layer will be located at the top level of the decision model, representing the model’s objective. The intermediate level will consist of attributes and functional requirements belonging to the industrial maintenance domain, distributed among the six layers to be analyzed. In the end, the lower level presents the alternatives: meets, partially meets, and does not meet; related to each functional requirement of the intermediate level. The relation attributes/requirements qualify the analyzed system (Justus et al., 2018).

Before this decision support method, a questionnaire aims to answer the importance (i.e., weight) of the elements to be raised. This is done based on the know-how of the chosen engineers and maintainers. Then, performing the AHP's peer review, the three alternatives are ranked, thus providing the result of the maturity assessment for each layer of RAMI4.0/FEI. When all six layers are evaluated, it will be possible to obtain the degree of maturity related to the requirements of Industry 4.0.

Having delimited the areas with a major lack of industrial maturity in Step 01, the objective of Step 02 is to prioritize maintenance functions. Those functions will be parameters in the process of implement I4.0 technologies to the legacy systems analyzed. The AHP method will be used again, but in another model (see Figure 4), aiming to gather the functions' weights solely and not support a decision. In other words, this AHP model will be used for assigning weights to the functions according to the preferences of the decision-makers, not regarding alternatives, as done in the previous Step 01. After that, those weights will be used to support the last decision step.

At this stage, another questionnaire, now based on the Figure 7 model, reflects Maintenance-4.0 expectations. It presents decision-makers a series of maintenance functions and their application in the light of I4.0. Based on the TPM's six main losses, the engineers and maintainers must consider their decisions regarding predictive, preventive, and reactive approaches that will guide maintenance processes to zero waste. At the end of this step's comparison, each function of the Maintenance-4.0 model is ranked by weight.

Figure 7
Step 02 – Maintenance-4.0 functions ranking (AHP model 2).

Based on the maintenance functions weighted in the previous stage, Step 03 objective is the prioritization of I4.0 technologies that best suit those functions. Here, the decision model does not require the organization’s engineers and maintainers, leaving the role of decision-maker to a maintenance-4.0 specialist. Considering it, the literature review on I4.0 technologies under the maintenance domain (section 2.3) serves as a base.

Step 03 decision model uses the PROMETHEE II method. The weights of each function of Maintenance-4.0, from the previous step, will be input and related to the nine technology groups from the literature review on maintenance technologies, as shown in Figure 8. The decision-maker is responsible for analyzing the technologies necessary to cover the maintenance functions.

Figure 8
Step 03 – Maintenance-4.0 technologies prioritization (Promethee II model).

Specifically, the technologies suggested for implementation are intended to increase the maturity of legacy systems, at the same time, ensuring interoperability due to barriers applied to FEI/RAMI4.0 layers. After completing all the framework's stages, there will be enough information to develop an assertive I4.0 compliance plan. Such a plan suggests that: The Maintenance-4.0 technologies selected in Step 03 enable the functions prioritized in Step 02, which will act on the diagnosed areas arising from the Step 01 maturity assessment.

In Step 01, the industrial maturity assessment made with the AHP method (seen in Figure 6) according to the maintenance managers’ questionnaire represented in Figure 9 resulted in the analysis from Figure 10.

Figure 9
Maturity questionnaire (Step 01) example.

Figure 10
RAMI 4.0 layers’ interoperability assessment from Step 01.

In order to clarify any possible doubts regarding the questionnaire, one of the authors followed the professionals’ considerations in person without any interference that was not requested. All the Consistency ratio of each layer comparison were accepted for being below 10%: Asset: 0.08380; Business: 0.05787; Communication: 0.09363; Functional: 0.06948; Information: 0.05362; Integration: 0.04954.

For the Asset layer, the maintenance professional highlights the functional requirement “Supervise equipment performance” and in the Business layer “Assist in an easy and quick way in individual decision making”. The functional requirement in Communication layer “Allow connection and exchange of information on mobile devices” and in the Functional layer “Record failure prediction learning based on maintenance history” were highlighted. In the Information layer was highlighted “Ensure data storage capacity” functional requirement and finally, the Integration layer stands out the “Allow the ability to connect with different industrial protocols” requirement.

With deeper analysis, even though Business and Information layers meet the level of maturity, in both cases the alternatives “does not meet” and “partially meets” together exceeds 50%. This means that the AHP method is pointing out the preference (i.e., acknowledgment) of decision-makers, that the factory is at a level that “meets” the requirements, but with more uncertainty in comparison with the Functional layer, for example. The analysis is presented in Appendix 1.

In Step 02, the functions of the Maintenance-4.0 model were ranked by relevance. A graph with the prioritization of its courses of action is presented in Figure 11.

Figure 11
Maintenance-4.0 best courses of action.

The 32 maintenance functions are ordered according to their respective weight in Table 4, resulting from the normalization of the AHP method (seen in Figure 7). The overall course of action comparison was acceptable with a Consistency ratio: 0.04973.

Table 4
Maintenance-4.0 Decisive Functions’ Rank.

The application of Step 02 took place in a second meeting, through a second questionnaire. A summary of the results obtained in Step 01 was made available to the maintainer, aiming to direct him to attribute less relevance to decisive maintenance functions poorly related to the target areas. The analysis is presented in Appendix 2.

For Step 03, one of the authors played the role of decision-maker as a specialist/consultant. This was possible because of the knowledge acquired by the literature review on I4.0 technologies in the maintenance context. The weights of the functions obtained in Step 02 were imputed in the Promethee II method (seen in Figure 8). Then the alternatives, Maintenance-4.0 technology groups, were analyzed by their level of need i.e., syntactic graduation from 1 to 9. Table 5 presents the ranking of the most relevant technology groups to meet the Maintenance-4.0 functions. The phi, represents the preference index used by the method.

Table 5
Most relevant technologies analysis.

Technologies at the cyber level were predominant: Analytics, Artificial Intelligence, and Big Data; along with Sensors at a physical level. They are responsible for enabling actions that are lacking in the factory, as established in Step 02.

The results of the framework’s application in an automobile factory show that it was possible to provide guidelines for adequacy plans. Although it has been positively validated, its complexity is evident. Among the main difficulties encountered are the long questionnaires that need to be filled out in the steps, as many judgments are necessary. However, it was confirmed that the proposed digital functionalities correspond with the organization's strategy of elevating efficiency and performance standards.

The purpose of the framework was to promote a new way of solving the application of technologies that support Industry 4.0 in legacy and maintenance systems. For that existing frameworks’ concepts were used to define such a non-trivial digital transformation strategy. It contributes in three distinct points, defining maintenance in I4.0; relating system’s adaptation and interoperability; and, how MCDM organize problems, supporting subjective decisions encountered in digital transformation projects.

*andre.venancio@pucpr.edu.br