Periodic Reporting for period 1 - DATASET (Advanced manufacturing though data analytics and intelligent insights)
Reporting period: 2019-12-01 to 2020-04-30
Problem: Modern factories are unnecessarily inefficient. A typical factory operates at only 66% productivity, which can cost industrial manufacturers an estimated $50 billion annually. Factories are wasting energy, water and raw materials, while simultaneously contributing with 56% of global greenhouse gas emissions. That does not mean factories are not utilizing technology. With the rise of Industry 4.0 factory owners have begun equipping their facilities with sensors to collect production data across all kinds of assets. Unfortunately, these same factories are not prepared to utilize these massive amounts of data for smarter and sustainable operations and hence manual analysis is impossible. The good news is that automated methods utilizing Artificial Intelligence have shown exciting potential for tackling this problem. Unfortunately, current solutions available in the market require too many resources and too much high-tech competency for factory owners to implement. Nor can these available solutions provide the level of insight necessary for factory owners to trust and deploy at scale. Modern factories need a solution which can easily integrate with their existing data acquisition systems and identify bottlenecks in the production process without requiring expert intervention.
Society: Managing operations in an environmentally and socially responsible manner – “sustainable manufacturing”– is no longer just nice-to-have, but a business imperative. Companies across the world face increased costs in materials, energy, and compliance coupled with higher expectations of customers, investors and local communities. Most of the Fortune 500 companies have already started to take important steps towards green growth – ensuring their development is economically and environmentally sustainable. However, many manufacturing companies (having a big share of the global greenhouse emissions simply due to heavy usage of fuels for energy) have not yet embraced these great opportunities. Industrial manufacturing accounts for almost 56% of the total greenhouse gas emissions in the world. Examples show that artificial intelligence can reduce energy and CO2 emissions with as much as 40% without reducing productivity or quality. Insight generated using Intelecy advanced analytics engine can be used to identify what process steps contribute most to the total CO2 emissions and best possible measures can be devised to minimize greenhouse gas emission.
Objectives:
The overall objective of the DATASET project is to mature, scale-up and demonstrate Intelecy with reference customers in different countries. The specific objectives of the SME Instrument Phase-2 will include:
- Testing and demonstrating the system with customers in different regions and industrial vectors, to continuously improve the user experience and shorten the on-boarding process for new customers.
- Enabling integration of Intelecy with market leading control systems which will increase our TAM by 5-fold.
- Demonstrate and document the benefits of these improvements with at least 5 end-users outside Scandinavia to increase market awareness and make the system ready for international market roll-out.
Conclusions:
Development, integrations, testing and demonstration has been successful, but because of the global COVID-19 crisis demonstrations outside of Norway has not been possible. But gladly Intelecy has been able to expand into new industry verticals such as metals, minerals and mining.
Society: Managing operations in an environmentally and socially responsible manner – “sustainable manufacturing”– is no longer just nice-to-have, but a business imperative. Companies across the world face increased costs in materials, energy, and compliance coupled with higher expectations of customers, investors and local communities. Most of the Fortune 500 companies have already started to take important steps towards green growth – ensuring their development is economically and environmentally sustainable. However, many manufacturing companies (having a big share of the global greenhouse emissions simply due to heavy usage of fuels for energy) have not yet embraced these great opportunities. Industrial manufacturing accounts for almost 56% of the total greenhouse gas emissions in the world. Examples show that artificial intelligence can reduce energy and CO2 emissions with as much as 40% without reducing productivity or quality. Insight generated using Intelecy advanced analytics engine can be used to identify what process steps contribute most to the total CO2 emissions and best possible measures can be devised to minimize greenhouse gas emission.
Objectives:
The overall objective of the DATASET project is to mature, scale-up and demonstrate Intelecy with reference customers in different countries. The specific objectives of the SME Instrument Phase-2 will include:
- Testing and demonstrating the system with customers in different regions and industrial vectors, to continuously improve the user experience and shorten the on-boarding process for new customers.
- Enabling integration of Intelecy with market leading control systems which will increase our TAM by 5-fold.
- Demonstrate and document the benefits of these improvements with at least 5 end-users outside Scandinavia to increase market awareness and make the system ready for international market roll-out.
Conclusions:
Development, integrations, testing and demonstration has been successful, but because of the global COVID-19 crisis demonstrations outside of Norway has not been possible. But gladly Intelecy has been able to expand into new industry verticals such as metals, minerals and mining.
Technical:
1. Built prototype and tested new integrations with systems supporting the industrial OPC-UA communication protocol standard. This prototype is improving the integrations with control systems like Siemens and Iconics.
2. The company has received consent from most users to analyze how they use the product; this brings valuable insight to usability and overall user experience using A/B testing on selected features.
3. Our UX expert has conducted interviews with new and potential customers to identify areas of improvement.
Commercial:
1. Geographical expansion: Established new partnerships in both in the ASEAN, Middle East and African regions. From several talks with partners we see that the maturity levels of the manufacturers vary a lot, both between regions and within regions. Intelecy have together with our partners defined a specification describing the most attractive customer profile and identified potential first movers in all regions.
2. Demonstrations were planned and confirmed in Asia, but the plans were postponed or cancelled because of travel restrictions and lack of resources for the manufacturers.
3. Even if the planned global activities stopped, we have been able to focus more on our Norwegian market and is now seeing a growth in demand for more data driven production solutions.
4. In parallel with the project we have hired a Communications Manager to improve the company’s communication and to increase market awareness outside the Nordics, directly and through our partners.
1. Built prototype and tested new integrations with systems supporting the industrial OPC-UA communication protocol standard. This prototype is improving the integrations with control systems like Siemens and Iconics.
2. The company has received consent from most users to analyze how they use the product; this brings valuable insight to usability and overall user experience using A/B testing on selected features.
3. Our UX expert has conducted interviews with new and potential customers to identify areas of improvement.
Commercial:
1. Geographical expansion: Established new partnerships in both in the ASEAN, Middle East and African regions. From several talks with partners we see that the maturity levels of the manufacturers vary a lot, both between regions and within regions. Intelecy have together with our partners defined a specification describing the most attractive customer profile and identified potential first movers in all regions.
2. Demonstrations were planned and confirmed in Asia, but the plans were postponed or cancelled because of travel restrictions and lack of resources for the manufacturers.
3. Even if the planned global activities stopped, we have been able to focus more on our Norwegian market and is now seeing a growth in demand for more data driven production solutions.
4. In parallel with the project we have hired a Communications Manager to improve the company’s communication and to increase market awareness outside the Nordics, directly and through our partners.
As part of the project we have worked with existing and new customers on many use-cases to validate the value creation from the project. Here are some concrete examples:
1. Identified and confirmed anomalies related to operators overriding security barriers to save time, but with negative impact on quality. With anomaly detection this was detected by the process engineers and corrective measurements were taken.
2. Identified quality related anomalies in cheese production related to heating and cooling. The value gained is to be able to stop production so customers don’t waste energy and time on products that will not meet the quality standards at an early phase and well before it reaches the customers.
3. Identified anomalies related to blocked load-cells. In this use-case some of the material could fall of the conveyer and get jammed causing the weight measurement to give a false number. If this isn’t detected fast enough it can cause error in the batch/recipe and a thus creating quality defects in the product.
The reason for using Machine Learning rather than pure threshold alarms is that in many cases the value will not be less/more than the normal operating range. With ML methods it can detect unexpected measurements that a threshold-based alarm method will not.
4. In this use-case a large pump was trained on behavior in the first 9 months after the bearing was replaced and temperature was between 50-55 in normal operating modes. Some months later temperature in the same bearing is now in the high 60’s. The ML model is still predicting all the behavior, but with a steady offset. When this offset reaches 15 degrees it means it’s time to schedule maintenance.
5. Compliance and environmental impact. In this use-case a model is predicting temperature on the wastewater 60 minutes into the future. When you have a good understanding of what will happen without acting, you can have operators act before it’s too late to adjust the process.
1. Identified and confirmed anomalies related to operators overriding security barriers to save time, but with negative impact on quality. With anomaly detection this was detected by the process engineers and corrective measurements were taken.
2. Identified quality related anomalies in cheese production related to heating and cooling. The value gained is to be able to stop production so customers don’t waste energy and time on products that will not meet the quality standards at an early phase and well before it reaches the customers.
3. Identified anomalies related to blocked load-cells. In this use-case some of the material could fall of the conveyer and get jammed causing the weight measurement to give a false number. If this isn’t detected fast enough it can cause error in the batch/recipe and a thus creating quality defects in the product.
The reason for using Machine Learning rather than pure threshold alarms is that in many cases the value will not be less/more than the normal operating range. With ML methods it can detect unexpected measurements that a threshold-based alarm method will not.
4. In this use-case a large pump was trained on behavior in the first 9 months after the bearing was replaced and temperature was between 50-55 in normal operating modes. Some months later temperature in the same bearing is now in the high 60’s. The ML model is still predicting all the behavior, but with a steady offset. When this offset reaches 15 degrees it means it’s time to schedule maintenance.
5. Compliance and environmental impact. In this use-case a model is predicting temperature on the wastewater 60 minutes into the future. When you have a good understanding of what will happen without acting, you can have operators act before it’s too late to adjust the process.