Yokogawa Technical Report / Vol.59 No.1 (2016)

At this site technical articles published at the YOKOGAWA technical report are introduced.


Expectations for Data Engineers

  • Manabu Kano*1

*1 : Department of Systems Science, Graduate School of Informatics, Kyoto University

Changes in Data Analysis Technology in the Manufacturing Industry in the Midst of the Third Artificial Intelligence Boom

  • Akio Nakabayashi*1
  • Hidehiko Wada*2

*1 : Industrial Solution Center Dept. 4, Solution Business Division, Yokogawa Solution Service Corporation
*2 : Technology Strategy Department, Business Development Center, Marketing Headquarters

   The arrival of the third artificial intelligence boom has been announced recently. As a background, parts of human intelligence are being replicated by processing big data thanks to the improvement of data infrastructure such as networks and cloud environment, and the advancement of data analysis technologies including statistics, pattern recognition and machine learning. This new IT trend is gradually changing the ways of utilizing data in the manufacturing industry. This paper introduces recent major IT trends and shows that they depend on big data and advanced data analysis technologies. The paper then describes the transition of industrial data analysis in the manufacturing industry affected by the IT trends, and discusses its future.

Modeling Technology for Optimizing Energy Consumption and Product Quality, and Example Applications

  • Kenichi Kamada*1
  • Mitsunori Fukuzawa*1

*1 : Incubation Department, Innovation Center, Marketing Headquarters

   Yokogawa has developed a modeling technology that can extensively extract all characteristics of process variables from historical operation data of actual plants. Yokogawa aims to improve the operation of customers’ plants with this technology that can create models elucidating relationships among energy consumption, operation cost, and product quality, and derive optimum operating conditions of plants through Yokogawa’s unique optimization algorithms. Whereas conventional plant modeling requires advanced technical knowledge and considerable man-hours for engineering, the new modeling technology greatly mitigates such issues. This paper describes the modeling technology and examples of its application to diagnosis using actual plant operation data.

A Method for Quickly Searching Similar Waveform Patterns in Historical Process Data

  • Tomohiro Kuroda*1
  • Tetsuya Ohtani*2
  • Hidehiko Wada*3

*1 : Incubation Department, Innovation Center, Marketing Headquarters
*2 : RTO Department, Advanced Solutions Center, Advanced Solutions Business Division, Solutions Service Business Headquarters
*3 : Technology Strategy Department, Business Development Center, Marketing Headquarters

   Plant operators often rely on their experience when operating plants. However, they could make more appropriate decisions by quickly referring to plant data similar to the current condition among a huge archive of process data. In such cases, efficient referencing of data is crucial.
   This paper introduces a technique to satisfy this need by focusing on the waveform patterns in current data and a certain period of historical data. This method can quickly identify and retrieve waveform patterns that match the target ones.

Data Analysis for Stabilizing Product Quality and the Mahalanobis Taguchi (MT) Method

  • Kikuharu Arai*1
  • Takeo Ueda*1

*1 : APS Department, Advanced Solutions Center, Advanced Solutions Business Division, Solutions Service Business Headquarters

   Recently in the manufacturing industry, there is a strong need to stabilize product quality while responding to diverse market needs. Plant operation should respond flexibly to external changes and requirements such as variations in raw materials quality and rigorous quality requirements from customers. However, it is difficult to satisfy quality only with final inspection processes; new approaches are needed to improve operating management by determining target quality values to be controlled in each process or at each production unit.
   Under these circumstances, Yokogawa has been proposing solutions for stabilizing the quality of products in various process units of customers such as quality management, manufacturing, and production engineering. This paper introduces a solution for quality stabilization that uses the Mahalanobis Taguchi (MT) method as a core data analysis technique.

Quality Stabilization of Formulation Process by Using Mahalanobis Taguchi (MT) Method and Applications to Continuous Drug Production

  • Masaru Konishi*1
  • Masanobu Sudo*2
  • Joji Murakami*3
  • Naoto Fujisawa*4

*1 : Market Development Group, Pharmaceutical Department, Sales Division 1, Yokogawa Solution Service Corporation
*2 : IA Strategy Department, IA-MK Center, Marketing Headquarters
*3 : Technology Strategy Department, Business Development Center, Marketing Headquarters
*4 : Consulting Department 1, Solution Business Division, Yokogawa Solution Service Corporation

   A paradigm shift in manufacturing processes has been proceeding mainly in formulation processes in the pharmaceutical industry. Pharmaceutical manufacturing, in which product quality is ensured under the control of good manufacturing practice (GMP), is facing the challenge of poor productivity compared with other industries. Thus, pharmaceutical manufacturers are making efforts to increase productivity while maintaining high product quality by transforming their production from batch to continuous processes. For this purpose, technology that ensures ideal continuous production while maintaining quality is required. This paper introduces a case example of a quality stabilization approach by using the Mahalanobis Taguchi (MT) method, and describes technical strategies necessary for constructing a continuous drug production system.

Machine Learning Applied to Sensor Data Analysis

  • Go Takami*1
  • Moe Tokuoka*1
  • Hirotsugu Goto*1
  • Yuuichi Nozaka*1

*1 : Filed Digital Innovation Department, New Field Development Center, IA Platform Business Headquarters

   IoT and big data have already become household words and people are becoming increasingly interested in how to effectively use enormous amounts of collected data. Meanwhile, the third boom of artificial intelligence has emerged, and relevant technologies including machine learning are evolving rapidly. Yokogawa believes that technologies for machine learning can enable advanced maintenance such as degradation diagnosis and predictive maintenance by being applied to sensor data analysis in the field of industrial automation. This paper outlines the theory of machine learning and introduces an application example of analyzing the data of pH sensors.

Real-Time Data Extraction Technology for Effectively Supporting Work Improvement Cycles

  • Nobuhiro Niina*1

*1 : Solution Co-creation Department, Business Planning Center, IA Platform Business Headquarters

   Yokogawa has developed a real-time data extraction technology for effectively supporting work improvement cycles. This technology can extract in real time each process data from consecutive data such as temperature and electric current. Using these data enhances the investigation of causes, analysis of tasks, implementation of solutions, and their evaluation, thus effectively supporting work improvement cycles. This paper describes the basic concept of this technology and introduces an example of its application to a film factory.

Evolution of Exaquantum to Accelerate Effective Use of Plant Data

  • Masaru Kimura*1

*1 : PA System Planning Department, Systems Business Center, IA Platform Business Headquarters

   Exaquantum is Yokogawa’s plant information management system (PIMS) software package. This is a core component of manufacturing execution systems (MES), and has evolved over more than 15 years since its release. The package was mostly used for applications in narrow areas such as using accumulated process data of distributed control systems (DCS) for monthly and daily reports. In the last few years, however, Exaquantum has been used by customers to accumulate data across multiple plants and analyze large amounts of process data and alarm messages in order to increase the operation efficiency of entire plants. Exaquantum R3.01, which was released in October 2015, has significantly enhanced data collection throughput and increased the maximum number of tags that can be handled while ensuring a real-time characteristic for improving plant operation efficiency. This paper describes technical features and improvements in R3.01, and introduces its solutions.

New Products