WHY AI IN MANUFACTURING IS A NECESSITY

AI is expected to radically change many industries. Working with AI in the field of manufacturing myself, I have seen how fast the transformation goes, especially when the technology is well suited for the field. In this post, I will briefly recap on the advancements within the field of AI, where AI/optimization in manufacturing stand, use cases and what to consider when implementing AI.

The last five years have shown lots of new applications transforming many industries. Many of these are driven by giants like Facebook, Google, Uber, and Amazon. However, what are the drivers behind these applications?

In short, you can summarize the drivers behind these applications to be explained within these three areas:

1. A huge amount of data

For many companies, lots of data have become available within the last decades as a result of the advanced use of ERP/CRM/MES or similar systems. Data can also be acquired from external sources to enrich data and IoT are obviously a huge source of data itself.

2. The rapid development of cheaper and faster processing power

Processing power can be installed on premises at reasonable costs or acquired from the larger SaaS cloud providers.

3. Advanced machine learning algorithms supported by researchers and open source

Open sourcing is specifically interesting. For instance, in 2015, Google made their machine learning libraries and models TensorFlow accessible to the public. Why so? Well, data is king. These companies sit on a treasure trove of data, meaning that they can still outplay smaller companies. By open sourcing their models, they get feedback from engineers and scientists around the world. It also helps these big companies attract new talents, which is a scarce resource — but we all benefit from the open source.

Let’s take a look at how this also influences AI in manufacturing.

For decades manufacturers have been investing in automation and sensors which also fuel further data acquisition for AI analytics. Processing power can be rented from either cloud or hardware providers as mentioned earlier. Data science work can be supported through engagement with consulting firms, universities or by building up in-house capabilities.

Working with manufacturing is a continuous race for establishing sufficient capacity for new or existing products and/or reducing production costs. So how can AI help here? Why not rely on well-known optimization methods?

Let’s have a closer look at AI and traditional optimization projects. You will need to do a lot of data retrieval and possible cumbersome value stream maps in order to identify areas ideal for optimization. This is time-consuming and data might not be easily available — if available at all.

Subsequently, working with data, you traditionally apply statistics to understand root causes, make Pareto graphs and much more. Business warehouses and dashboards are perhaps built to help monitor the process in real time — though this is all stationary. Within the field of AI, you will also use statistics to understand data. But more so, historical data will be used to identify hidden patterns and train AI algorithms so that they will be able to predict outcomes.

So, in short, you will move from not only understanding the past but based on the past and the present, you can start predicting the future and take actions in due time. Predicting the risk of a heart attack for a person with a specific health record is just one example. Within manufacturing an example could be that you can start predicting how well your equipment will be running, by understanding how input parameters influence the output of products. Based on this, adjustments can be done manually or automatically to maximize output. The new insight into data will also support you in deciding which long-term improvement initiatives will increase equipment output and what the increase would be.

The biggest difference from a traditional optimization project thus lies in the nature of an (adequately designed) AI project. Understanding the business challenge is a prerequisite for both types of projects. But the data preparation part for an AI project should be done with a long-term view. The opportunity to retrieve the right data will add tremendous value to the company and enable the development of supporting tools for decision making and prescriptive guidelines.

To summarize, the benefit from embarking on an AI journey is that data that adds long-term value will be identified and used proactively. AI models will thus enable prediction of the future on an ongoing basis, giving you a powerful tool to maneuver.

Finally, with this, you will be able to initiate optimization projects based on already available data, without having to start from scratch every time and even be able to predict the benefit of such initiatives.

When all of the above is said, companies still need to have the infrastructure in place to manage data science projects, curate open source code and ensure traceability to be able to share and monitor the developed models. In my opinion, you need to start getting experience with AI and machine learning and consider working with AI as a long term inevitable investment.

Related posts:

According to OptinMonster, 77 percent of internet users read blogs. And among these blog readers, many are looking for content to solve a problem.

Are you looking to start your own business? It’s not always an easy task, especially if you’re expecting to go it alone. There are so many people in America with an entrepreneurial spirit, but that’s…

Mawmaw Interior adalah sebuah firma interior yang berdomisili di Depok, Jawa Barat. Firma ini bergerak dalam bidang desain, building dan consultant interior. Pada proses ini saya mencari dan…