Utilization of AI, which is expected to solve problems in the field

Today, all industries are facing various issues such as "heavy work load", "aging equipment", and "labor shortage".
By 2025, most ethylene production facilities in Japan will have been in operation for over 40 years. Expectations for AI are increasing as we work to somehow solve these problems.

With AI, we will be able to obtain various effects such as inheritance of know-how, standardization of judgment standards, reduction of workload, and improvement of productivity.
However, in order to do so, it is important to take into consideration the characteristics of AI, and to set the issues and the effects of introduction in advance and work on them.

Although we would like to proceed with the introduction of AI as soon as possible, there are three main obstacles to its implementation.
There are "management issues" such as organizational development and lack of human resources, "technical issues" that make it difficult to ensure a high level of technology, and "project execution issues" represented by the difficulty of setting goals and explaining reliability.

“What is particularly difficult is how to explain the reliability. Then, how should we deal with the lack of explanation of AI? "Hybrid AI" is able to incorporate these three measures." (Kusunoki)

In recent years, NEDO (New Energy and Industrial Technology Development Organization) has listed "AI for semantic understanding" as one of the "AI technology developments that should be addressed", and is considering commercialization, showing the importance of AI. and expectations for the future.

What is hybrid AI that solves the lack of explanation?

Machine learning and deep learning, which have been widely used in the third AI boom, are called "numerical AI" and learn rules by finding patterns between input and output. It allows the machine to make predictions and decisions based on the sample data it was trained on without being explicitly told how to perform the task.
However, the problem is that it is a "black Box" that does not explain why it was inferred that way. On the other hand, “Symbolic AI” is a knowledge model that humans have and is an AI that can be explained.

Kusunoki: "By using pattern recognition with numerical AI to discover features from large amounts of data, and combining human experience, knowledge, and knowledge with symbolic AI, it is possible to create a white Box and support decision-making."

数値AIだけの意思決定プロセスは、センサーデータを分類もしくは予測し、アラートメッセージを出すことはできますが、どうすればいいのか具体的な行動につながる情報は多くありません。
ハイブリッドAIでの意思決定プロセスでは、熟練者のナレッジやガイドラインをもとに作ったシンボリックAIを組み合わせることで具体的な理由や解決方法を添えて意思決定をサポートするリコメンドが可能になります。つまり、現場の担当者は、確証ある行動をとることができます。

Hybrid AI spreading in various fields

ハイブリッドAIは、すでに適用領域が広がっており、さまざまな領域において社会実装の実績があります。

At power plants, it is used to improve combustion efficiency and predict equipment abnormalities.
In the medical field, after numerical AI detects abnormalities in vital data acquired from sensors, symbolic AI incorporates the knowledge and findings of clinicians to make final decisions.

製油所では、計画された生産スケジュールと実際のオペレーションとのギャップが課題でした。
そこでハイブリッドAIによる意思決定のサポートを行うことで、プラントの製造工程全体を見すえた解決策の提示や、タイムリーで透明性の高いレコメンドを提示し、操業効率や収益の向上に貢献しました。

“Hybrid AI also contributes to optimizing the placement of gas leak detection sensors. It incorporates network topology, sensor data and inputs to chemical distribution systems, and deep learning, physics-based symbolic AI and numerical AI. By utilizing hybrid AI based on , it is possible to discover network sensors with insufficient coverage and advise on the optimal placement of sensors in the pipe.The optimal placement of sensors reduces costs, It seems that the risk of downtime and interruption has been greatly reduced.” (Kusunoki)

Hybrid AI is also used to improve the yield of ethylene production.
Ethylene is one of the world's most sought-after petrochemicals, with global production tripling since 1980.
However, the plant is still operating at full capacity at this time. Compressor fouling or blade imbalance can cause excessive vibration above thresholds and lead to unplanned shutdowns, resulting in huge losses.

“Normally, after a problem occurs, the manager receives a call from the site, and after a specific engineer checks the site, the replacement parts are ordered and prepared, so it takes a lot of time.Hybrid AI introduction With this, symbolic AI based on operational knowledge can present recommended actions to managers when a problem occurs.The manager can immediately present the parts that need to be replaced to field technicians, It led to a reduction in time." (Kusunoki)

Design thinking is also necessary for social implementation of AI

Hybrid AI can effectively combine the advantages of numerical AI and symbolic AI.
However, there are some companies that find it difficult to suddenly introduce AI, or that they are not yet working on AI. In such cases, please use "Re:Alize" provided by Macnica.

“Re:Alize is an AI social implementation service that designs an experience that users will want to continue using. AI is not the purpose of creating it, but how to incorporate it into business and management strategies is a very important point. Re:Alize is a service that allows us to design and work on such experiences together.” (Kusunoki)

The Re:Alize service is divided into three steps: "Re:Concept", "Re:Creation" and "Re:Experience".

“In the first step, ‘Re: Concept’, we organize the issues and set goals together. After that, we assess whether we should use numerical AI, symbolic AI, or hybrid AI.” (Kusunoki)

In the next step, “Re:Creation,” we actually analyzed the data and carried out AI modeling. Then, we will develop applications that utilize AI.

"In the last 'Re:Experience', we will have users actually experience it. Then, we think that opinions such as usability will come out, so we will give feedback on it and further set the task. We will continue to improve usability by repeating this process.” (Kusunoki)

Design thinking is key to this effort. Effective use of design thinking is effective when tackling social and management issues with AI. Efforts to connect AI to social implementation using design thinking are becoming very important.

Data scientists and engineers alone are not enough to implement AI in society.
At Re:Alize, members including UI/UX designers work together as a member of the customer's team. If you have any problems, please feel free to contact us.

AI has the power to accelerate change