Recently, the evolution in information technologies such as the Internet of Things (IoT), Cloud Computing (CC), and Cyber-Physical Systems (CPSs), besides development in communication technologies, are revolutionizing the system approach to transmitting information between various sources. Undoubtedly, the revolution in all aspects of life nowadays is due to digitalization. Because of this advance and rebellion in Information Technology and Industry 4.0, the Digital Twins (DT) has come to be one of the talented ideas [1–8]. The DT is a digital model of real products from the perspective of CPSs. Therefore, a virtual complement of a convinced system mimics its real behavior. During full lifecycle development, the digital data from the virtual system, which is coupled to a real system, characterizes the information from the physical system. Consequently, integrating digital with its physical twins makes it an efficient method to handle, control and support the decision-making when the real system is running [9]. The DT records all data collected from physical sensors as well as the execution system response. As a result, the critical role of DT is to predict and diagnose the behavior of the physical system to foresee malfunction or fault, and feed data to the system to provide the best possible maintenance [10].

The virtual representation of the system is the major goal of DT to enhance productivity, increase operation flexibility, and decrease maintenance time and costs. On the other hand, DT can be run depending on the requirement, either on a cloud-hosted system or at the edge layer, to address the issues of Industry 4.0, which arise from the production of linked components. DT, on the other hand, supports both a clarification of the behavior of the actual system and the best possible solutions for the physical model. To improve the control action, foresee the system performance, and endorse decision making, DT uses a fundamental modeling system, simulation procedures, and transparent simulation [11–14].

The digital twin concept was first introduced by Grieves in 2002 [15], and then National Aeronautics and Space Administration (NASA) was the first to apply it to construct virtual space machine models. Existing DT works and implementations are still in their early stages, according to the literature, and require a lot of effort. However, it is well-integrated into a wide range of applications, such as biomedical systems, manufacturing, aerospace, agriculture, smart cities, and weather forecasting [1,2]. In addition, a number of distinct specialist engineers and computer scientists are needed in this essential subject to design an efficient DT system for any physical system. Their responsibilities will include building and designing the essential product prototype as well as creating a detailed explanation of the virtual system.

With the advance of modern technologies currently, the digital twin is integrating with other technologies like cloud computing, Artificial Intelligence (AI), Big data, Blockchain, and IoT and sensor data fusion to develop a digital data structure that is reorganized and modified when its physical counterparts alter [7,12,16–20]. In recent years, IoT integration with DT has allowed for greater visibility of physical Twins and their current status. It also simplifies the connection and documentation of the system function so that the performance of the physical devices may be understood and clarified. Furthermore, Artificial Intelligence is continually improving DT’s abilities by processing data from its physical counterpart and surroundings. Data analysis and prediction result in the detection of patterns, data classification, and model identification.

The introduction of modern technologies, such as Blockchain, has recently made it easier to use Digital Twins (DTs) in practical applications [21–23]. Also, emerging AI technology like deep learning with CPSs is utilized and accompanied by digital twins to improve performance and enhance manufacturing production [17]. It proposes a concept for smart manufacturing and Industry 4.0 through the use of conceptual schema and modern technology. In [24], they presented a novel scheme for fault detection and prognosis based on digital twins and deep learning for fault diagnosis. The deep learning system may easily detect failure in the process. Digital Twins are used for defecting diagnosis for an electrical machine, which is proposed in [22]. A new DT with the Industrial Internet is introduced in [25] to improve the vision for smart production [26–28].

Thus, this research paper presents an efficient fault diagnosis framework based on GA and ML algorithms. The contributions of this paper are concise as follows:

Developing a hybrid adaptive ML and GA algorithm for automatic and efficient fault diagnosis in industrial IoT applications.

The remainder of this work is structured as follows. Section 2 discusses the previous related works. All details and descriptions of the suggested framework are given in Section 3. Then, the high-level structure of the suggested framework with further details is presented in Section 4. The achieved outcomes for examining the performance of the suggested framework are offered in Section 5. Section 6 presents the main concluding remarks and some future directions.

This study aims to create and use a DT model for the Triplex pump by using simulated data to detect and diagnose faulty equipment. To increase efficiency, safety, and continuous production, fault detection and diagnosis have been carried out for various industrial processes. Several intelligent fault diagnosis technologies have been presented in recent decades to improve the durability and safety of complicated equipment [29,30]. The model-based approach, in which a precise model of complex equipment is created using various analytical expressions, is a well-known and successful method for fault diagnosis [31–34]. The well-established model-based approaches have been effectively deployed on numerous processes for complex equipment based on physical and mathematical knowledge of industrial processes. But developing an accurate model with a thorough understanding of the underlying physical processes is a problem because of the enormous complexity of some equipment.

Another powerful way to predict or diagnose fault states in complex equipment is the data-driven approach [35–38]. The mathematical correlations reflect the equipment’s complicated features; without any physical model, the features of a fault condition can be learned. Recently, machine learning algorithms have gained great attention and are used in mechanical fault diagnosing operations. The use of a backpropagation neural network based on the Levenberg-Marquardt training algorithm to develop an effective fault detection strategy for proton exchange membrane fuel cells (PEMFCs) systems is presented in [39]. Another fault diagnosis based on a multiple kernel support vector machine for the distillation column is developed in [40]. Dynamic distillation column simulation generates data in normal and faulty operations and then uses the proposed approach to provide a high fault detection rate. Deep learning based on LeNet-5 is proposed in [40] for fault diagnosis of motor, centrifugal, and hydraulic pumps. The Convolution Neural Network (CNN) is proposed in [41] for detecting Photovoltaics (PV) inverter open-circuit problems. Finally, the edge cloud paradigm is used to perform real-time diagnosis.

It is observed that work presented many works on faults diagnosis. However, there are rare contributions to the literature on digital twin-assisted machine learning algorithms in automated industrial IIoT applications. Consequently, this paper suggests an intelligent fault prediction framework that composes four steps precisely: (1) Data Acquisition Step (DAS), (2) Data Handling Step (DHS), (3) AI-based Model Training and Testing (AIMT2), and (4) Fault Classification and Detection Step (FCDS). This research proposes to build an innovative smart prediction digital twin-assisted AI framework using a hybrid genetic algorithm using various ML algorithms such as Logistic Regression (LR), Naïve Bayes (NB), and Support Vector Machine (SVM) that can effectively classify faults.

In this work, we presented a novel automated fault diagnosis framework for discovering the status of digital twin-assisted AI. Fig. 1 demonstrates the steps of the suggested framework. It composes different algorithms, such as GA, LR, NB, and SVM methods. Additionally, the suggested hybrid automated framework comprises various significant steps to achieve the diagnostic system, which are clarified as follows:

Step 1 (Data Gaining and Gathering): This step is responsible for data collection from DT sources to be utilized in the next steps in the suggested framework.

The layout of the employed genetic algorithm used in the suggested DT-based AI framework is given in Fig. 2. Genetic algorithm (GA) is utilized to optimize the features of the model through minimizing the fitness function of the error between the predicted and the actual values of the measured data.

The suggested framework has the objective of professionally fault diagnosis for machinery. So, the diagnosis process is mainly accomplished by exploiting the DT, IIoT, CC, and AI concepts for real-time fault diagnosis and monitoring of industrial systems. Thus, this proposes enlightening the efficiency of industrial control systems by delivering a more advantageous intelligent system that enables early observation of the faulty process. The proposed framework consists of three key phases that cooperate to accomplish system purposes. Fig. 3 explains the suggested fault diagnosis framework that composes three different phases.

Phase 1: In this phase, the IIoT system is utilized for real-time DT data gathering.

This section offers the simulation results, their analysis, and discussions of the suggested diagnosis framework.

With the advancement in industry 4.0, many technologies such as Cloud Computing (CC), Digital Twins (DT), Artificial Intelligence (AI), and Industrial Internet of Things (IIoT) can be utilized to improve the performance of industrial systems. Therefore, this work presented an efficient fault diagnosis system utilizing the integration of DT and AI to accurately identify the early faults in industrial machines. In the suggested framework, the DT model of the industrial machine was constructed by generating the simulation testing model and further updating incessantly measured data from the physical asset. Then, different ML algorithms were employed in the suggested fault detection and diagnosis framework for data classification. The suggested hybrid framework was examined d through a case study of triplex pump fault diagnosis with different generated datasets for learning purposes. The experimental results demonstrated that the hybrid GA-ML method elasticities outstanding results compared with using ML methods individually. It is observed that the suggested hybrid framework accomplished a high detection accuracy of 95% in the case of using the hybrid GA-SVM classifier. Therefore, the suggested fault diagnosis framework can assist industrial operators in making proper decisions regarding fault analysis for industrial control systems. In the future, different efficient deep learning algorithms must be employed in the suggested framework to further improve detection performance and diagnosis efficiency. In addition, different industrial IoT scenarios and applications can be investigated for developing other DT-based automated diagnosis systems.