*1.2. Blockchain*

Blockchain technology is one of the famous areas for trust and safety, which can apply in any related topics to keep the information and data private. Similarly, it is a novel technology for decentralized and distributed computing architecture that keeps the dataset with encrypted blocks in a chain [12–14]. Digital information related to transactions, date and time, amount, etc., which are elaborated in the transaction process, is all stored in blocks. The saved data are available within the distributed network, containing nodes' participants to validate the transaction. All nodes throughout blockchain are linked with each other and support the crypto and transaction codes. Another important feature in blockchain technology is the mathematical algorithms, which are very strong in this network. It provides block validation to minor nodes without any effect on data through the blockchain network, which is why blockchain is secure and transparent [15–23]. There are many of research requirements for addressing the security problems and recommendation systems based on blockchain and knowledge discovery technology [24–29]. This process needs to carry out the integration of blockchain and IoT. Similarly, the security issues which are mentioned by many authors specify the blockchain as a good solution. In [30], blockchain's key features are defined as trust, security, programmability, etc. A blockchain can be one of three different types—a public blockchain, a private blockchain, or a consortium blockchain. The public blockchain is famous for digital currencies. The main objective of a consortium blockchain is to combine the stakeholder and service trading entities. Li et al. [31] presented the energy trading system based on a consortium blockchain. Min [32] proposed to leverage blockchain methods to enhance supply chain flexibility in risky situations. In a business trading system, blockchain technology can be assumed for IoT applications for implementing private blockchains. In [33], an IoT-oriented data

exchange system was designed based on the Hyperledger Fabric to overcome the automatic maintenance of a distributed managemen<sup>t</sup> system problem.

#### *1.3. Internet of Things, Industrial Internet of Things, Industry 4.0, and Cyber-Physical Systems*

The growth of the IoT system provides substantial support for the digitalization environment. Furthermore, the IoT applications cover different perspectives—smart farms, smart cities, traffic monitoring, etc. Similarly, the machine-to-machine (M2M) techniques are also covered by IoT systems, which is a way forward of digitalizing the manufacturing system [34]. The abstraction of Industry 4.0 becomes apparent when IIoT meets the cyber-physical system (CPS), which is the best solution for improving the efficiency of productivity in smart manufacturing. Yang et al. [35] presents the IoT applications and issues in the smart manufacturing system. The conclusion of the proposed work shows that IoT visualized the interconnection of the physical world and cyberspace. On the other hand, in [36], a cyber-physical production system (CPPS) was proposed to authorize the dataset efficiency transferring based on the intelligent network and trustworthy communication technology. The Industrial Internet Consortium (IIC) is one of the most famous techniques launched in US top five companies—GE, AT&T, Cisco, Intel, and IBM. This technique mainly points to the standardization of network innovations, applications, and constructions; data circulation growth; and industrial digital transformation. The IIoT sub-concept was first launched in Germany by the name of Industry 4.0 and globally partial CPS facts based on artificial intelligence in smart manufacturing. In short, CPS shows the relationships between information and the physical world, relying on the interconnection of things. The IoT technology selects the interconnections between physical address objects to check if they are related to the industry or not. Table 1 shows the studies related to smart manufacturing systems. Ten studies are compared based on the industry sector, internal equipment, external equipment, and concept of creation.


**Table 1.** A taxonomy of smart manufacturing applications.

Table 2 presents the recent challenges on the integration of blockchain and IoT technology in the smart manufacturing industry. The comparison shows the techniques applied in this research, the main contributions of the presented methods, the usage of

blockchain and IoT, the challenges of the proposed systems, and the limitations of the research.


**Table 2.** Challenges of blockchain and IoT integrated methods.

The development of smart manufacturing underpins integrating information technology, data technology, and operational systems. The ever-increasing facilities and devices are leading to data processing and application challenges in existing technology. To reduce this issue's effectiveness, multi-access edge computing was extracted from cloud technology as a solution for the mentioned problems and for its ability to simplify the data processing in the Industrial Internet of Things and industrial cloud computing [57]. Another issue in the smart manufacturing system is the transmission of data and business transactions. Blockchain technology is a suitable answer to overcome this issue, which stabilizes data transmission and business transactions by using the distributed control mechanism [58]. Smart manufacturing systems' immense data processing causes the issues mentioned in [59,60]—high dimensionality, feature space, etc. Deep learning allows the data processing to automatically go through complex feature abstraction using multiple layers, and similarly provides advanced data analysis for smart manufacturing. The challenges mentioned above are being analyzed using state-of-the-art machine learning

techniques and smart manufacturing applications. Figure 1 shows the data-driven role in the smart manufacturing system. The data-driven process is divided into three main layers named data-driven, manufacturing system, and benefits. The data-driven layer contains machine learning, deep learning, artificial intelligence, the Internet of Things, big data, and cloud computing techniques. After data-driven, the manufacturing system layer contains three main steps, named technology in manufacturing, network, and advanced analysis. This step's important information includes the design, process, equipment, records, customers, suppliers, parts, and workforce information. The last layer of the data-driven system has the manufacturing system's benefits: quality, energy, cycle time, etc.

**Figure 1.** Smart manufacturing's data-driven roles.

The main contributions of this paper are:


The rest of this paper is divided up as follows: Section 2 presents the proposed integrated model's conceptual scenario in smart manufacturing. Section 3 presents the final result and validation of the system's performance, and we conclude this paper in the conclusion section.

#### **2. System Architecture of the Proposed Smart Manufacturing Environment**

The integration of edge computing, blockchain, and machine learning can simplify data processing and transactions in s smart manufacturing system. The following steps present the details of the proposed method in a smart manufacturing system.

#### *2.1. Prototype System Based on Edge Computing*

The edge computing system's main concept is to apply the computing technique as close to a data source as possible. Figure 2 presents the edge computing architecture in the smart manufacturing system. The local infrastructure is used to process the data in an edge-computing system, and it takes the cloud server to the hardware. There are three main layers in the edge computing system named the physical layer, network layer, and application layer. The physical layer consists of sensors, robots, actuators, etc., organizing the physical layer's main components. The second layer contains the various edge servers, which process the terminal devices for the third layer's input. Unlike a cloud server, an edge server provides a computational service limited to capacity and resources. The root of enterprise-level applications is IIoT cloud server data processing, all done in the application layer. Enterprise information systems (EIS), supply chain managements (SCM), and smart contracts (SC) are some application layer examples. Applying edge computing in smart manufacturing is far greater than cloud server supplementary resources. Edge computing's prosperity is highly based on virtualization technologies. Virtualization technology contains virtual machines and containers. The main differences between them are the implementation and level of isolation; in the virtual machine, the implementation needs hardware visualization. In the virtual container, the performance is based on light-weight visualization.

**Figure 2.** Overview of edge computing architecture.

#### *2.2. Service Validation Based on Blockchain*

The blockchain technology in smart manufacturing consists of two main contributions. The first one is IIoT, and edge computing servers' smart manufacturing changes from cloudcentered to the distributed system architecture. In this process, the blockchain system is applied to strengthen data integrity and decrease data transmission risk to authorize the validation key and identification in a distributed manner. To avoid operation defectiveness,

the data transactions should be time-stamped through the hash code and refrain from positioning the fake data in the linked chain. The second one is the consensus mechanism, which is used to decide whether adding a validated block into blockchain is possible or not. Smart manufacturing digitalization recommends manufacturing virtualization, leading the cloud manufacturing service from another point of view.

Figure 3 presents the manufacturing system based on two main fields, contents and metadata: identify the unique service and give a detailed description of the process. The service block was created based on the manufacturing system abstraction, and similarly broadcasting the distributed manufacturing in-network service to further validate network entities. The service transaction block creation is based on purchasing and querying the manufacturing service. The transaction block is in the same manufacturing system network, and validates based on the other peer-to-peer entities. Similarly, the transaction block adds to the blockchain transaction system too. In contrast, blockchain's transaction process organizes the smart contract between the business partners, facilitates the inner protocols, and verifies a contract's performance.

**Figure 3.** Overview of a blockchain service.

#### *2.3. Machine Learning-Based Smart Manufacturing*

Based on the recent new technologies—big data, IoT, etc.—smart facilities are positively developing intelligence manufacturing to impact the cross-organization in smart manufacturing systems. The manufacturing system is experiencing an unexampled data extension based on the data collection from sensors in various formats, structures, and semantics. Data collection is based on the multiple manufacturing systems, e.g., lines of product, manufacturing equipment, processes, etc. Hug data in the manufacturing system need data modeling and analysis to handle the high-volume dataset growth and support the real-time data-processing. Machine learning techniques contain some advantages for improving smart manufacturing: cost reduction, security, fault reduction, increasing production, operator safety, etc. These advantages include a grea<sup>t</sup> and strong bond for the operating procedure. Furthermore, the system's fault detection is one of the decisive components for predictive preservation, and it is essential in the case of industry. Figure 4 presents the overall architecture of smart manufacturing based on the integration of edge computing, blockchain, and machine learning. Each of these methods is well-known, but the integration between them has a huge effect on the manufacturing industry regarding safety, cost reduction, increasing production, etc. The edge computing section is based on the physical, network, and application layers. The physical layer provides the smart sensors connected to the IoT platform for real-time data collection and monitoring. Similarly, in

this layer, the ability to check the condition of machines is also available. The network layer updates the information and tracks the dataset over time. The application layer corresponds and reviews the data quality, and finally measures and reports the monitoring results. The edge computing process's final report moves to a blockchain service for securing the collected information in blocks. This information is in terms of assets, design, and block security. The process moves to the machine learning section to control the quality of the service and fault rate prediction. In this section, there is a various level of data analysis. This process contains predictive analysis, diagnostic analysis, descriptive analysis, and prescriptive analysis. The main goal of descriptive analysis is to give the product manufacturing process and operation information, capturing the environmental conditions and parameters. If the product's performance decreases, the diagnosis analysis examines the issue and presents the reason for the problem. The predictive analysis operates the statistical models and predicts the possible future equipment and products based on a historical dataset. The final analysis is the prescriptive analysis, which further recommends actions and measures the identification to improve the rates of outcomes, solve the problems, and present each final decision outcome. Based on the advanced machine learning analysis, the smart facilities are highly optimized. This process's benefits are reducing the costs of operation, meeting changing consumer demands, improving productivity, and reducing downtime.

Equations (1) and (2) present the evaluation of cost reduction in manufacturing industry based on machine learning prediction process. In the first step is a derivation function applied to decrease the error of cost function. The cost function is evaluated below:

$$B = \frac{1}{m} \sum\_{n=0}^{m} (\mathbf{g}\_n - (\mathbf{x}h\_n + d))^2 \tag{1}$$

where *gn* is the predicted value and *xhn* is the actual value of the cost prediction process. *α αx* represent the partial derivative values. *d* and *e* are representing the intercept, and *x* represents the slope of the evaluation.

$$\frac{\alpha}{\alpha \infty} = \frac{2}{M} \sum\_{n=1}^{M} -h\_n(g\_n - (xh\_n) + \varepsilon) \tag{2}$$

The predictive accuracy evaluation is based on two main metrics: mean absolute prediction error (MAPE) and normalized root mean square error (NRMSE). Equations (3) and (4) present the MAPE and NRMSE evaluations.

$$MAPE = \frac{1}{m} \sum\_{n=1}^{m} |\frac{\mathcal{g}\_n - \mathcal{g}\_n}{\mathcal{g}\_n}|\tag{3}$$

$$NRMSE = \frac{1}{m} \sqrt{\sum\_{n=0}^{m} (\frac{\mathcal{g}\_n - \mathcal{g}\_n}{\mathcal{g}\_n})^2} \tag{4}$$

The MAPE evaluates the prediction's total error compared with initial values, and NRMSE evaluates the normalized squared errors.

**Figure 4.** Smart manufacturing overall architecture based on an integrated system.

#### *2.4. Fault Assessment Diagnostic Analysis*

Generally, the manufacturing system faces failures based on abnormal and degradation operations. The failing causes high costs, disqualifies the product, and causes lower productivity. Based on the implementation of a smart manufacturing system, it is necessary for smart factories to monitor the condition of machines, identify the primary defects, recognize the root causes of failures, and finally combine the information for manufacturing system production [61]. Based on the data collected from sensors, there are many machine learning algorithms to investigate the fault diagnosis and classification [62]. The convolutional neural network (CNN) combines feature learning and identification into one model and has been applied in many sectors— wind generator [63], rotor [64], bearing [65–68], etc.
