Research: What are obstacles, advice for integrating AI, industrial automation?
More than half of respondents were challenged when trying to integrate artificial intelligence (AI) and industrial automation, according to Control Engineering and Plant Engineering research. See advice.
- Learn how implementing AI in industrial automation has challenges, according to respondents to a Control Engineering and Plant Engineering survey; respondents also offered advice.
- Understand AI considerations and how to overcome challenges to AI use for industrial automation.
- Review other areas the AI research covers and methodology.
AI for industrial automation insights
- Implementing AI in industrial automation has challenges, said respondents to a Control Engineering and Plant Engineering survey.
- Some survey respondents offered advice about how to overcome challenges to AI use for industrial automation.
- Understand other areas that the AI research covers and its methodology.
More than half of respondents (57%) to a Control Engineering and Plant Engineering survey of use of artificial intelligence (AI) for industrial automation have had a challenge or obstacle when integrating AI technologies into existing industrial automation systems; 14% have not and 29% have yet to implement AI.
Implementing AI in industrial automation
Internal champions for increasing use of artificial intelligence (AI) in respondents’ facilities are found in various departments and titles. Among leading titles are end-user engineer, director and project manager, according to the September research, “Artificial Intelligence in Industrial Automation.”
Results showed a three-way tie when deciding about AI use, considering margin of error for the research.
Same level of scrutiny as they would any other new technology.
Take a closer look due to lack of AI experience.
Great deal of scrutiny given AI’s potential issues.
AI considerations: Operational efficiency, ROI, better use of data
When asked what factors are considered when deciding whether or not to implement AI-based automation solutions in industrial processes, the leading responses were: Operational efficiency gains; cost and return on investment (ROI); and availability and quality of data. Other important factors (about half or more of respondents also checked off) technological infrastructure as well as workforce impact and skill requirements.
How to overcome challenges to AI use for industrial automation
Some survey respondents explained how to overcome those AI integration challenges in a write-in question, lightly edited below.
Acceptance by management without technical skills.
Lack of client knowledge regarding AI. It takes a lot of training and educating clients before deployment, even before owner engagement.
Operational technology system integrators have no knowledge in machine learning (ML) and AI to understand how it works as the following: 1) What is an ML model? 2) What is an ML algorithm? 3) What is ML training and inference? 4) What are the critical data, information and outcomes for selected topics? 5) How to define the valuable topic and return on investment (ROI)? 6) What are labels and features? 7) How to store time series data of data sources from industrial field to information technology (IT) systems, database (DB) or data lake.
IT workers have no knowledge in physical world and industrial automation. Be the consultant to interview the C-level and plant manager for valuable topics from managing matrix then collect the field data from PLC tags to OPC [Unified Architecture] server to DB then import as data source for ML utility training.
Selection of the features, final customer needs not clear
Gain culture and mindset changes to start with and then get buy-in of the senior management with active participation of stakeholders. Let the things go as it is …. this type of mindset is the real issue. Another issue is right problem statement with expected benefits. There is a requirement of single vendor who can take up turnkey job till establishing ROI. Data availability outside the distributed control system and programmable logic controller (DCS/PLC) network with its security is also a matter of concern.
AI cannot be used for process automation because is not safe enough. I think AI is only statistical methods.
Alerts and risk score of incidents in cybersecurity.
Other industrial automation AI research results
Other information includes:
A ranking of vendors considered AI leaders.
AI algorithms and technologies
Measuring AI implementation success
Top priority AI applications
AI implementation timing estimate
Positive impacts of AI
Optimize existing processes versus redesigning for AI
Impact of AI on industrial automation and other areas.
Methodology for industrial AI research
The sample for this study was selected from qualified subscribers of Control Engineering and/or Plant Engineering media with valid email addresses. The large sample was split into two in order to avoid list and respondent fatigue. Subscribers were sent an email asking them to participate in this study. The email included a URL linked to the questionnaire.
- Data collected: Aug. 4 through Aug. 17, 2023.
- Number of respondents:
First sample = 58 (+/-12.9% margin of error at a 95% confidence level)
Second sample = 71 (+/- 11.6% margin of error at a 95% confidence level)
Incentive was offered, a drawing for one $100 Amazon.com eGift Card.
Mark T. Hoske is content manager, Control Engineering, CFE Media and Technology, email@example.com. Amanda McLeman is the research director and project manager of awards programs for CFE Media and Technology.
KEYWORDS: Artificial intelligence for industrial automation
Have you integrated AI and automation, yet? Or maybe you have, and it’s so embedded, you didn’t even know.