Swarm Intelligence

Generally, the swarm intelligence (SI) approach is a famous process among artificial intelligence algorithms. Two main strategies follow based on this algorithm named approximate and non-deterministic to consider and utilize the searching spaces to find the near-optimal solutions [69]. SI contains various approaches; among them, the artificial bee colony (ABC) algorithm demonstrates SI's classic features. The importance and required process for intelligence performance, self-organization, collective behavior, and decentralization of SI are sufficient [70]. Moreover, the mentioned three features contain the simple mechanism control, which is tuned with only two parameters. The bee colony's size determines whether the solution can be dropped or whether there is no need to drop it. Figure 8 shows the process of solving the computational task problem based on the artificial bee colony workflow.

**Figure 8.** Artificial bee colony workflow.

#### *3.4. Performance Evaluation*

The test models are generated based on the following patterns. The task length follows a uniform distribution in the range of [1, 10] million specifications. The data volume is defined as 100 KB to 10 MB. The time delay is 100 milliseconds to 10 s. The average processing performance based on the edge server is defined as 10 million instructions per second (MIPS). The cloud volume is 1000 MIPS. Edge server and device connections work through wireless communication. The edge server and cloud connections go through broadband. Tasks are specified to the evident edge server, which forwards the information to the cloud. This causes the edge server to be limited to processing enough resources for

the under-processed task during the delay time. Figures 9 and 10 present the analysis of parameters for abandonment and solution number (SN) criteria of *α* for incoming tasks (200).

**Figure 9.** Performance evaluation of various solution number (SN) settings.

**Figure 10.** Performance evaluation of various *α* settings.

To show edge computing's effectiveness in the system of smart manufacturing, Figure 11 shows a various number of incomes based on three main frameworks, i.e., cloud, edge, and mixed-mode. The meanings of these three scenarios show the computational task between them. As shown in Figure 11, the mixed-mode shows the combined outperformance of edge and cloud. Similarly, it is increasing the number of tasks along with the cloud mode's average processing time. When the tasks are less than the cloud server's capacity, there is a decrease at a certain level; on the other hand, if the number of functions increases, then the edge server does not modify the processing time appropriately.

**Figure 11.** Performance evaluation of various scenarios.

Figures 12 and 13 present the machine learning techniques applied in this process: k-means clustering algorithm (IKCD), k-means clustered deployment (KCD), and random deployment (RD). Figure 12 shows each algorithm's delay rate in different edge computing nodes (ECNs). When the number of nodes increases in the edge computing system, the amount of equipment production also reduces due to the network's delay reduction. Based on the presented results, the network delay in the IKCD process is the shortest one, and RD is the worst among the compared methods. Based on the ECNs in the system, when there are between 1 and 3, the IKCD method is better than KCD, and when the number of ECNs is more than three, based on the network latency, the differences between IKCD and KCD decrease.

Figure 13 presents the system cost deployment differences for the ECNs. We can see the ECNs incurred greater costs based on the system node increases. Based on the deployment of ECNs for the higher costs, the costs for all three methods increased. Comparing the three methods, IKCD had the highest and most outstanding performance. The RD method's deployment used a number of ECNs randomly and did not deploy any node in the production node. This process caused the node to be chosen without consideration and constraints. The deployed nodes recorded in the KCD process are based on the Euclidean distance between the devices, which is not sensible and causes the network delay and access time communication for data processing in real-time requirements.

**Figure 12.** Changes of network nodes based on edge computing.

**Figure 13.** Cost dependency deployment on edge computing nodes.

Figure 14 presents the relationship between system cost and edge computing nodes based on the compared methods.

**Figure 14.** Total cost of IKCD and edge computing node relationship.

The results show the advantages of the IKCD method, and similarly, show the reduction of network delay based on the ECNs and increasing the computing cost of ECNs. The sum of the process decreases at the start and then increases.

#### *3.5. Challenges and Opportunities of the Smart Manufacturing System*

Data management, performance evaluation, and standardization of edge computing in the IIoT system are briefly explained in the above sections. Analyzing the proposed system based on the integrated methods reveals grea<sup>t</sup> opportunities and challenges for edge networks, data processing, security, etc., in IIoT technology related to edge computing. The below information explains several challenges and opportunities for smart manufacturing systems.


#### **4. Conclusions and Future Research**

Smart manufacturing is a favorable movement for the evolution of the manufacturing industry and production in a new industry. The manufacturing system's implementation causes the support of data technology, information technology, and operational technology, surrounded by the development of integrated edge computing, blockchain, and machine learning based on the Industrial Internet for operational processes in the manufacturing environment. This paper's proposed system was designed based on integrating edge computing, blockchain technology, and machine learning to support the manufacturing system's design. The assignment problem of the system was formulated based on the optimization model. Unlike other research in edge computing and IIoT, the presented method's stresses illuminate the integration method's importance in future developments. The future research plan is to improve the manufacturing system more and analyze it in more detail. The blockchain system's applications can be quantified and further analyzed. Other technologies can be incorporated to enhance the development of the manufacturing system. The experiments and results can be analyzed with an on-site dataset to identify the possible impact factors and regulate the proposed model's configured parameters.

**Author Contributions:** Conceptualization, Z.S.; Formal analysis, Z.S.; Funding acquisition, Y.-C.B.; Methodology, Z.S.; Writing–review and editing, Z.S.; Investigation, Z.S.; Resources, Y.-C.B.; Methodology, Z.S.; Project administration, Y.-C.B.; Supervision, Y.-C.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was financially supported by the Ministry of Small and Medium-sized Enterprises(SMEs) and Startups(MSS), Korea, under the "Regional Specialized Industry Development Program(R&D, S2855401)" supervised by the Korea Institute for Advancement of Technology(KIAT).

**Conflicts of Interest:** The authors declare no conflict of interest.
