This series introduces trendy, popular themes from among those covered in Fujitsu Research Institute's information magazine Chisounomori by interviewing consultants who are working on business projects. This article's theme is "Mathematical Optimization Technology." (Interview Date: January 13, 2017)
Speakers (honorifics omitted)
Akihito Ichikawa (Director, Supply Control Center, Department of Supply Control and Disaster Management, Tokyo Gas Co., Ltd.)
Takeshi Iijima (Deputy Director, Supply Control Center, Department of Supply Control and Disaster Management, Tokyo Gas Co., Ltd.)
Jota Ugai (Field Innovator, Field Innovation Unit, Fujitsu Limited)
Tomoko Sunadoi (System Division IV, Social Infrastructure Systems Unit, Fujitsu Limited)
Mieko Moki (Chief Senior Consultant, Business Analytics Consulting Group, Fujitsu Research Institute)
Minami Watanabe (Interview Moderator: Head of the Business Analytics Consulting Group, Corporate Executive Officer, Fujitsu Research Institute)
*This article is a summary of pp. 8-17 of Chisounomori 2017 Vol. 4. All affiliations and titles are as of the time of the interview.
The Supply Control Center Ensures a Stable Gas Supply
――There is a growing movement to accelerate business digitization by actively using new ICT services, such as AI, big data, and IoT. Also, there is a need to work on initiatives like "Industry 4.0." Drastic productivity improvements in business activities are expected to promote smart management, which in turns improves business scalability and ensures reasonable management decisions. First, could you tell us about the Supply Control Center?
Ichikawa: We use various facilities to deliver city gas produced at plants to customers. The Supply Control Center's 37 maintenance workers perform maintenance and control for 280 facilities in the Tokyo metropolitan area, including electrical instruments at high-pressure governor stations located along high-pressure pipelines originating from liquefied natural gas (LNG) terminals as well as electrical instruments and machinery at pressure regulation stations for gas holders (also called gas tanks) in towns.
――What is important when performing such difficult tasks?
Iijima: I think the most important challenges are ensuring a "stable supply" and "security." It is easy to say so, but operation standards must be observed when carrying out operations, and stable supply can be realized by steadily carrying out work procedures in accordance with inspection service manuals defined under the operation standards. We instruct our personnel to strictly adhere to the operation standards and work procedures so that they do not rush and fail to observe procedures.
――Do you mean the Company emphasizes education and training for work?
Iijima: Yes. It is difficult for new employees to always be aware of ensuring a "stable supply" and "security," so we raise awareness by repeating these words out loud as a chorus during work to help them understand the concepts' importance.
From Defensive to Offensive Maintenance
――In addition to ensuring a "stable supply" and "security" as absolute requirements, a small group of elite professionals must efficiently maintain and control the large number of gas supply facilities scattered across a wide area, correct?
Ichikawa: Currently, one center monitors 280 facilities in the Tokyo metropolitan area, and we expect the area covered and number of facilities will increase in the future. It was obvious that the workload would get out of hand sooner or later if we relied exclusively on the conventional maintenance method. Around the time we noticed this, we received a proposal on Field Innovation (FI) (*) from Fujitsu, which analyzed our operations to find new maintenance approaches--in other words, to achieve business reforms.
――What is the new maintenance method?
Ichikawa: There are two key points. Under the conventional method, maintenance workers left to visit individual facilities from a single control center located in Senju, which means that the travel distance increased as the facility area expanded. To cope with this challenge, we set up maintenance centers to serve as relay bases so that operators could visit facilities in remote maintenance areas. The second is a change in maintenance systems--from equipment-based to facility-based maintenance. In the past, we planned maintenance on a per-equipment basis and visited all facilities that had the target equipment; now, however, we plan maintenance on a per-facility basis, so the various types of equipment located in a given facility are checked when operators visit said facility.
――What kind of scheduling system supports the Supply Control Center's business innovation?
Iijima: We have a system for forming teams that have the necessary levels of operational knowledge and calculating the optimal maintenance schedule by deriving the most effective order for visiting the 280 facilities. Previously, it took about a week for us to work out the schedule for the following month, but now, the system works it out for us in just 50 minutes.
Sunadoi: Actually, the maintenance operations and the amount of time and workers needed to conduct maintenance differ depending on facility size and equipment age. I think it was actually impossible for us to take all of these factors into account when manually preparing maintenance plans.
――Could you please tell us a little about the background to the business reforms?
Iijima: Under the "Challenge 2020 Vision" (**) formulated after the Great East Japan Earthquake, the Pipeline Department was tasked with "ensuring safe, stable, and efficient energy supply." Given the movement to expand and use LNG in various areas, as of 2012 we could not be certain about the future scale of expansion in equipment and areas. At that time, Fujitsu brought us a proposal for sharing its wisdom on how to improve our operational efficiency, so we started to review our work.
――That was Fujitsu's proposal of FI.
Ugai: At that time, everyone felt uneasy about further area expansion, saying, "It seems impossible to continue in the same way, but we have done this for 30 years." Based on the onsite work volume survey, it was obvious that one-day round-trips took a long time, and many forward-looking opinions were expressed about business reforms regarding topics such as "shortening travel distances" and "facility-based maintenance." Therefore, we suggested two business reforms: "visiting facilities from remote relay bases" and "a facility-based maintenance style," which shortens travel distances to enable the same number of workers to perform facility maintenance across wider areas. Tokyo Gas had not changed their maintenance method by themselves because they had difficulty realizing efficient scheduling that reduced travel times while adjusting for complex conditions, so they requested that we develop a new system.
――What did Tokyo Gas think about the proposal?
Iijima: To be honest, we were unsure at first. However, as I listened to the proposal and looked at what it showed for the next year, and the year after next, and 20 years from now, it numerically indicated that our method would fail in the year 20XX. For us, improving work efficiency meant "defensive maintenance" that focused on longer inspection cycles and shorter inspection times by simplifying operations, but this proposal motivated us to take on the challenge of "offensive maintenance" by drastically changing our conventional approach.
――I see. Tokyo Gas's always high awareness of future risks and the FI proposal perfectly meshed with each other.
Applying Mathematical Optimization Techniques to Determine the Order of Facility Visits and Team Formation
――Under the concept of changing the conventional way of operations, you were involved in developing a viable scheduling system. What did you do?
Moki: Fujitsu Research Institute was responsible for function design and logic development to use AI to achieve mathematical optimization of the scheduling system. This was right after implementing FI, and we had clearly identified the following challenges facing the client: "responding to future requirements to handle equipment maintenance across wider areas with the same number of workers" and "leveling out operational workloads and reducing travel distances." These challenges were not so simple that they could be resolved merely by creating a monthly scheduling system. Rather, we developed a three-stage logic consisting of a medium- and long-term plan to level out operational workloads over the several decades to come, an annual plan, and a monthly schedule created by inputting these plans. Therefore, we used mathematical optimization technology to solve the problem of how to allocate maintenance work on a monthly basis while minimizing the number of facility visits and leveling out the work volume over the next 20 years. Mathematical optimization is good at solving combination problems to find a balance point given various purposes. Next, we built a logic to determine efficient routes for visiting facilities and formed teams so as to minimize travel distances. Both the order of facility visits and team formation are combinational optimization problems, and we also used mathematical optimization technology here.
――Did you verify your plans' viability by simulation before development?
Moki: Yes. We used simulation for verification before starting development and concluded that maintenance operations could be carried out by the current 37 workers. In addition, the simulations verified the quantitative benefit that the number of annual facility visits could be reduced from approximately 1,500 to 500 per year by changing the maintenance style from equipment-based to facility-based and preparing plans using mathematical optimization technology.
――Given various constraints, such as inspection cycles and travel times, it is difficult for humans to conduct simulation manually, isn't it?
Moki: When I received this offer, I thought the inspection cycles' characteristic constraints perfectly matched the capabilities of mathematical optimization technology, so I was confident that we could derive feasible plans to meet the target requirements. It is quite rare to have an opportunity such as this in which mathematical optimization technology can be adapted to all aspects of medium- and long-term plans, annual plans, and monthly schedules, so I really wanted to tackle and solve this problem. Since mathematical optimization is basically numerical calculation performed by computers, it sometimes produces system-centric results, which may not always be acceptable for field maintenance workers.
――As ongoing maintenance operations had to be switched over to the new plans abruptly at a certain point in time, inspection cycles for some operations had to be accelerated. Of course, this encountered resistance because some maintenance operations had just been done recently. However, this issue was settled through internal negotiations while looking 5 to 10 years ahead.
Aiming to Develop a System to Enable Flexible Responses and Continued Service in Field Operations
――Please tell us about system development.
Sunadoi: I observed maintenance operations in action and knew that situations could arise in which emergency responses would be required and workers thus may suddenly become unable to make maintenance visits. Therefore, I thought that field workers might not accept the optimal results derived using mathematical technology as is, and instead strived to develop a system for flexibly responding to various situations. For example, we determine the personnel assignments ("who goes where") without mathematical technology. Instead, we use the mathematical technology to form teams and assign individuals to each team. Also, we only determine which facilities should be visited within a given week; the order of operations is left up to field workers. Moreover, we developed a system for flexibly responding to emergencies.
――So you changed from "system-centric mathematical optimization" to human-centric optimization?
Moki: In this project, we assumed various operations scenarios, carried out simulations of multiple patterns, visualized the results, and asked the client to compare and examine them in order to achieve results that were as close as possible to what the client desired. We plotted monthly facility visit simulation results on a map, showed it to the client, and extracted feedback, such as "We do not visit mountainous areas in the north during the snowfall season" and "We do not visit sightseeing areas during holidays." Also, I simulated the balance of "leveling out operational workloads" and "minimizing travel distances," which have a trade-off relationship, and explained the differences in results quantitatively.
――You made great efforts to make the system human-centered.
Moki: Thus far, we have applied mathematical optimization to planning systems for various types of industries and operations. Based on this experience, we expect that mathematical optimization will tend to result in system-centric optimization regardless. I talked with Sunadoi about a system function that could cover that point and proposed that the system include an "adjustment function."
Sunadoi: Yes. We originally received a request for achieving business reforms, not functional requirements for developing a new system. Since the business reform scenario from the FI phase, including the background, was shared with all people on both the customer and Fujitsu sides, we were sure that we could meet the client's expectations. In system development, often requirements confirmed in the requirement definition phase get changed in the design phase, or clients change specifications after the design phase. However, in this project, almost no such change requests were made. I think that was because the focus was simply on "leveling out operational workloads" and "improving efficiency."
Iijima: Honestly speaking, we have yet to use the system 100%. However, we are becoming more aware of the system's effectiveness and will use it even more actively in the future.
――How do you see the current situation, Director Ichikawa?
Ichikawa: First, I think the system has worked well for the past year. If we use the system more actively, we can carry out operations more efficiently and easily. The challenge for the future is whether the people in charge can make full use of it. I myself sometimes visit facilities for maintenance as well as to educate young personnel. I hope the system will be used successfully while taking into account the points we have discussed here.
Aiming to Be a Slim, Powerful Operator
――Lastly, I would like to ask you about the Supply Control Center's future goals.
Ichikawa: First, we want to get the scheduling system's operation on track to respond to expansions in the number of facilities and operations while verifying that we can handle the maintenance of facilities to be opened next year and those under construction without trouble. Second, although the scope of this scheduling system covers only maintenance and periodic inspection operations, we now want to include other operations, such as facility renovations, in order to optimize the efficiency of all operations in the future. Third, as our president declared in recent remarks, Tokyo Gas will aim to be a slim, powerful pipeline operator. In response to increasing the number of management facilities and expanding pipelines, we must consider not only operation quality and efficiency but also sophistication of human resource management and operations support as important challenges.
――So Tokyo Gas's business assumes further expansion of the gas supply network in the future?
Ichikawa: The number of facilities we maintain will clearly increase, given that they are used for transportation networks that comprise social infrastructure. Also, equipment where it used to take only 10 minutes to perform maintenance may be changed to ones that require an entire day to complete maintenance. In consideration of such points, the time may come when we need to review our operations again.
Iijima: In breaking away from the parameter that "an increase in the number of facilities leads to increased work volume" during the requirement definition phase, Fujitsu may not have expected a large increase in the number of facilities. However, approximately 20 parameters have already been added, so this was a miscalculation. For operational improvement, we must increase the number of workers who can handle the system other than myself and add interfaces for system maintenance. Once this scheduling system is applied, it will be difficult to stop or change to another one. Therefore, going forward we want to continue to use this system while maintaining it through collaboration with Fujitsu.
――Based on the fact that full liberalization of the gas retail market will start in April 2017, it can be expected that Tokyo Gas's business environment will also change significantly. Given such circumstances, I think the importance of its business as gas supply infrastructure and social infrastructure will further increase. From the viewpoint of safety and security in establishing and maintaining infrastructure as well as improving service levels, future operations will require still further sophistication. Thank you very much for your time today.
*: Field Innovation (FI): Activities to visualize business challenges together with customers to promote reforms in ICT, work methods, skills, and operational processes.
**: Tokyo Gas Challenge 2020 Vision
***: Tokyo Gas Co., Ltd. website
* Corporate Vice President Watanabe resigned on March 31, 2017.
* This interview is a reproduction of an article that appeared in Fujitsu Research Institute's information magazine Chisounomori 2017 Vol. 4 with some corrections.