Special Issue: Smart Service Technology for Industrial Applications

1. Introduction

With the gradual maturity and popularization of the Internet of Things (IoT), technologies of measurement and analysis for production data have also been continuously advanced, realizing the collection of large production data. Effective production data analysis and applications can help to enhance process management technology, as well as process quality level [1,2,3]. Not only does it accelerate the development of intelligent manufacturing for Industry 4.0, but it also facilitates the improvement of process capabilities. In addition, the rapid development and advancement of emerging technologies, such as production data analysis technology and artificial intelligence, have promoted innovation and fierce competition between businesses around the world. This has led many manufacturing industries to become more service-oriented [4,5] in order to provide not only new innovative and valuable products, but also smart service. Smart service is a new type of digital service that uses and combines the growing internal and external data of the manufacturing industry to create personalized solutions for customers. Smart service offers a variety of new possibilities for the manufacturing industry. Therefore, the topic of this special issue is “Smart Service Technology for Industrial Applications.” In this special issue, 26 papers were submitted, and 14 of them were published. These 14 papers can be roughly divided into three subfields. First, Section 2 focuses on four papers introducing the subfield of machine learning, deep learning, and artificial intelligence application. Second, Section 3 emphasizes six papers addressing the subfield of performance evaluation and applications in smart manufacturing and service technology. Third, Section 4 concentrates on four papers presenting the subfield of production data analysis and statistical decision-making. Last, conclusions are made in Section 5.

2. Machine Learning, Deep Learning, and Artificial Intelligence Application

In this subfield, four papers were published, presenting highlights as follows:

(1)

Control Chart Concurrent Pattern Classification Using Multi-Label Convolutional Neural Networks:

This paper used multi-label convolutional neural networks to construct a classifier for concurrent patterns of a control chart. The results of this study showed that the recognition performance of multi-label convolutional neural networks is better than that of traditional machine learning algorithms. This study also applied real-world data in order to demonstrate the applicability of the proposed method to online monitoring, contributing to the further realization of intelligent statistical process control [6].

(2)

High-Dimensional, Small-Sample Product Quality Prediction Method Based on MIC-Stacking Ensemble Learning:

This paper selected eight algorithms as basic learning models, used the maximum information coefficient (MIC) to obtain the correlation between the basic learning models, and selected the model with low correlation and strong predictive ability in order to build a stacking ensemble model. Furthermore, a model based on stacking ensemble learning and measurement was proposed, which would effectively avoid overfitting and obtain better prediction performance. Taking enhancing prediction performance as the optimization goal provides a new method to ensure the accuracy of quality prediction for the final product [7].

(3)

Multitask Learning with Knowledge Base for Joint Intent Detection and Slot Filling:

This paper proposed a joint model based on knowledge base and slot-filling. This model was based on long short-term memory and convolutional neural networks to obtain shared parameters and features between these two modules. Subsequently, the knowledge base was introduced into the model to improve its performance, and a weighted loss function was built to optimize the joint model [8].

(4)

Evaluation of Deep Learning-Based Automatic Floor Plan Analysis Technology: An AHP-Based Assessment:

This paper suggested that a newly developed technology could handle complex floor plans. First, an evaluation framework was proposed to evaluate the comparison between the newly developed technology and manual digitization. The results of the study revealed that this newly developed technology can systematically compare the technical values between the automatic floor plan analysis and traditional manual editing. In addition, a qualitative evaluation was conducted, for the first time in any existing research. It can be said that the evaluation has guaranteed the effectiveness and practicality of the automatic floor plan analysis used as a reasonable technique for receiving indoor spatial information [9].

3. Performance Evaluation and Applications in the Smart Manufacturing and Service Technology

In this subfield, six papers were published. Highlights are individually addressed below:

(1)

Applying ANN and TM to Build a Prediction Model for the Site Selection of a Convenience Store:

This paper used the techniques of Artificial Neural Network (ANN) and Back Propagation Neural Networks (BPN) to build a systematic and reliable prediction model for selecting locations of the convenience store chains, in order to improve the existing decision-making method for choosing locations, which is currently done by experienced managers. This model adopted the Taguchi method (TM) to find the optimal parameters of the BPN. The study results demonstrated that the prediction accuracy and decision-making quality of this model are higher than those of the existing manager-oriented decision-making method [10].

(2)

Multi-Relational Graph Convolution Network for Service Recommendation in Mashup Development:

This paper solved the problems of data scarcity and cold start faced in service recommendation, and proposed a novel framework of Multi-Relational Graph Convolutional Networks (MRGCN) used for service recommendation. This study revealed that the method proposed by this paper, under certain conditions, can outperform the state-of-the-art method in the aspect of service recommendation [11].

(3)

Parallel-Structure Deep Learning for Prediction of Remaining Time of Process Instances:

This paper derived a new deep learning method consisting of a convolutional neural network (CNN) and a multilayer perceptron (MLP) with embedded layers in the context of the growing popularity of deep learning and the need of exploiting heterogeneous data. This method performed experiments with real-world datasets, and the result showed that the proposed method can achieve more accurate predictions [12].

(4)

Simulation and Optimization for a Closed-Loop Vessel Dispatching Problem in the Middle East Considering Various Uncertainties:

Due to uncertainties such as weather and unexpected incidents, as well as the lack of effective management techniques, the operation of the closed-loop shipping system often lags behind. Therefore, this paper developed a detailed and realistic simulation model to evaluate the economic and environmental performance of the closed-loop vessel dispatching system, and also took into account various uncertainties of weather and port operations. In addition, this paper incorporated the optimization model into the simulation model and proposed a new vessel dispatching policy which first applied to large vessels, in order to achieve cost savings and reduce greenhouse gas emissions [13].

(5)

Fault Location and Restoration of Microgrids via Particle Swarm Optimization:

This paper first introduced the fault location algorithm and the recovery algorithm, respectively. The fault location algorithm was based on the network connection matrix to form a new system topology. After the fault section was assigned, the particle swarm algorithm was used to realize the multi-objective function, and the optimal recovery was carried out under its constraints. Finally, a series of simulations and analyses were performed for the sample system. The result showed that the proposed optimal algorithm can effectively solve the problem of fault location and recovery in the micro-grids [14].

(6)

Analysis of the Effectiveness of Promotion Strategies of Social Platforms for the Elderly with Different Levels of Digital Literacy:

This paper aims to examine strategies for social platform promotion. Based on empirical results of the innovative diffusion research, an agent-based model was created to analyze the effectiveness of social platform promotion strategies for the elderly with different levels of digital literacy. The result demonstrated that the proportion of passive information receivers (PIR) has a direct negative impact on market penetration and a moderating effect on the effectiveness of various promotional strategies [15].

4. Production Data Analysis and Statistical Decision-Making

In this subfield, four papers were published, introducing the following highlights:

(1)

Fuzzy Quality Evaluation Model Constructed by Process Quality Index:

In this paper, a model of fuzzy quality evaluation was developed by the process quality index. This model can cope with the problem of small sample sizes arising from the need for enterprises’ quick response, which means that the accuracy of evaluation can still be maintained in the case of small sample sizes. Moreover, this fuzzy quality evaluation model was built on the confidence interval, enabling a decline in the probability of misjudgment incurred by sampling errors [16].

(2)

Process Quality Evaluation Model with Taguchi Cost Loss Index:

Many statisticians and process engineers are dedicated to research on process capability indices, among which the Taguchi cost loss index can reflect both the process yield and process cost loss at the same time. Therefore, in this study, the Taguchi cost loss index was used to propose a novel process quality evaluation model. In addition, this paper adopted the mathematical programming method to find the confidence interval of the Taguchi cost loss index, and then employed this confidence interval to perform statistical testing and to determine whether the process needed improvement [17].

(3)

Novel Service Efficiency Evaluation and Management Model:

This paper proposed a standardized concept with a service operation efficiency evaluation index. This index is not only convenient and easy-to-use, but it also has a one-to-one mathematical relationship with the performance achievement rate. The advantage of the simple and easy-to-use point estimate can be maintained, and the risk of misjudgment caused by sampling errors can be reduced as well, which is helpful for the service industry to move towards the goal of intelligent innovation management. This method is not only applicable to the performance evaluation of the multi-workstation service operation process, but also to the performance evaluations of other service operations [18].

(4)

Statistical Hypothesis Testing for Asymmetric Tolerance Index:

An asymmetric tolerance index is a function of the average of the process and the standard deviation. It is difficult to obtain the confidence interval of the index. Therefore, this study adopted Boole’s inequality and DeMorgan’s theorem to find the combined confidence region for the average of the process, as well as the standard deviation. Then, this study adopted the mathematical programming to find the confidence interval, and also employed this confidence interval for statistical hypothesis testing [19].

5. Conclusions

Effective production data analysis and applications of the process management technology are relatively beneficial to the enhancement of the process quality level. At the same time, they can also accelerate the development of smart manufacturing for Industry 4.0. This Special Issue was proposed based on such a concept. It is hoped that by means of discussing emerging technologies, such as production data and artificial intelligence, more advanced production data analysis and intelligent service technology will be proposed to the industry for reference, so as to speed up the innovation and development of the industry. This Special Issue had a total of 26 papers submitted, of which 14 papers were accepted and published, with a rejection rate of nearly 50%, which can be said to be quite successful.