Disclaimer: This article is published in partnership with Siemens. Siemens is paying for my engagement, not for promotional purpose. Opinions are my own.
In a short time, Artificial Intelligence (AI) has gone from a visionary concept to near-mainstream in many large companies. AI can be defined as broad spectrum of methods or technologies that perform tasks which would normally require functions of human intelligence such as learning, judging, and problem-solving. Boston Consulting Group’s Most Innovative Companies 2019 study has shown that top innovators occupy AI leadership and that also companies from traditionally non-digital industries are strongly capitalizing on AI’s potential. One of them: Siemens.
More than 30% of the BCG survey respondents have expected AI to be among the areas of innovation with the highest impact on their business in the next three to five years. Early AI programs tend to focus on improving processes and operational efficiencies – likely because early quick wins can be demonstrated in these areas.
The study also indicates that AI leaders rely on ecosystems. Digital natives, such as Amazon, Google, IBM or Microsoft offer AI capabilities through their cloud-based platforms. Ecosystems leverage collaborating partners that bundle the underlying technologies, applications, engineering tools and services needed to produce an integrated solution. Data ecosystems will play a determining role in shaping the future of competition in many B2B industries.
Industrial AI: key requirements
As can be seen from the BCG list, the highest ranked innovative companies deploying AI at scale reside in the B2C- and digital business (e.g. retail, e-commerce, music). But how does AI play out in the industrial sector? Here, the past decade has revealed AI’s potential and spawned for the most part predictive maintenance applications which have become integral part of Siemens’ product and service portfolio by now.
The next challenge will be about expanding also into other application fields and turning that potential into business value at scale. This will involve overcoming several challenges. Dr. Matthias Loskyll, Director Advanced Artificial Intelligence at Siemens Digital Industries names three being of particular importance:
- Robustness and trustworthiness: In many industrial organizations, bad decisions can have much more serious consequences than in other sectors – financially and safety-wise. So, if we hand something over to AI machines or replace entire hardware systems for quality assurance by them, we want to trust them to make the right decisions and function reliably in a dynamic, uncertain and volatile production environment. AI systems need to be robust enough to ensure continued operation within safe limits upon perturbations.
- Data availability: As opposed to many digital companies, who draw on abundant data from the Internet for training their AI models, industrial companies are confronted with imbalanced data, i.e. high availability of ‘good’ data, but only small datasets for the failure or defect cases the AI system is actually supposed to detect or predict. How to deal with this issue? Possible solutions involve Synthetic Data Generation or Few-Shot Learning.
- AI talent: AI experts and Data Scientists are in short supply these days. Deploying AI at scale requires contributions from virtually the entire workforce. Yet, there are still limited options to train for instance IT experts in the field of AI. This much-needed transformation calls for taking appropriate measures in Democratizing AI.
Application fields and use cases
At present, AI techniques (especially Machine Learning) are largely used to solve discrete problems that require patterns to be identified in large volumes of a single type of data or input (e.g. data from a specific machine). Current application fields of AI across various industries encompass (see image below):
- Anomaly / defect detection
- Quality inspection
- Lifespan prediction
- Quality prediction
- Object detection
- Predictive maintenance
- Autonomous systems
- Computer vision
- And various design- and simulation-related fields, such as document classification, physics simulation or generative design
I see another field with tremendous potential for AI applications: process optimization. – Dr. Matthias Loskyll
Intelligent AI solutions can analyze high volumes of data generated by a factory to identify trends and patterns which can then be used to make manufacturing processes more efficient and reduce their energy consumption. Employing Digital Twin-enabled representations of a product and the associated process, AI is able to recognize whether the workpiece being manufactured meets quality requirements. This is how plants are constantly adapting to new circumstances and undergoing optimization with no need for operator input. New technologies are emerging in this application area, such as Reinforcement Learning – a topic that has not been deployed on a broad scale up to now. It can be used to automatically ascertain correlations between production parameters, product quality and process performance by learning through ‘trial-and-error’ – and thereby dynamically tuning the parameter values to optimize the overall process.
A specific – and quite challenging – use case comes from the automotive industry. Quality inspection in the car production process (see image below) involves detecting scratches and dents in painted car parts, such as the hood. What may sound straightforward is in fact quite a tricky task: on a part’s shiny surface defects may look completely different which makes detection very difficult. However, Deep Learning enables automotive companies to significantly improve their quality inspection systems.
Industrial AI involves ecosystems
AI frontrunners typically leverage ecosystems – networks of companies, customers and other partners – that interact to create mutual value. This has been an enabler of some of the most profitable and valuable business models that exist today, such as Azure, Microsoft’s Cloud Computing Service. Ecosystems address two main aspects:
- Advantage of the data flywheel effect, i.e. enabling massive data accumulation, analysis and shared data pools, fueling improvements to products and business processes and thereby stimulating further growth and data access.
- Seamless and comprehensive domain competence by organizing business partners with complementary capabilities. This allows satisfying multiple customer needs, resulting in a lock-in effect.
Case in point: On top of being a technology leader, Siemens also positions itself as an ecosystem partner. One way the company does this is by orchestrating the Mindsphere World Organization. Mindsphere World is a global, interdisciplinary and independent association of companies as well as research and governmental institutions. Together, they aim at co-creating the future of the Internet of Things (IoT) and jointly nurturing vital innovations in this field.
Supporting tasks close to the shop floor, Siemens also offers an open Industrial Edge ecosystem, which can deploy AI training results in a digestible format. The deployment, assisted by an Industrial Edge management system, is less time-consuming due to scalable, orchestrated devices.
Outlook: The Future of Automation
In light of the current state of play in Industrial AI, what can we expect from the months and years ahead? How does the Future of Automation look like? Matthias Loskyll differentiates a short-term and a longer-term outlook:
- Short-term (ca. 12-18 months): This will likely be a period of disillusion after AI having been hyped extensively for many years. Many will increasingly realize that AI is not magic, but just intelligent algorithmics. The focus will be on optimizing the application fields and use cases mentioned above, e.g. by mitigating ‘model drifts’ – which denotes a degraded accuracy due to differences between actual production data and the model’s training data. This can lead to incorrect detections or predictions and significant risk exposure. AI-related offerings are supposed to target a comprehensive customer support including managed services, helping the customer increase efficiency, yield and, hence, financial return.
- Longer-term: Beyond that, AI will be deployed at a larger scale, enabling completely new production processes, leading to Autonomous Factories and Mass Customization, making use of highly agile, modular manufacturing systems that can be configured flexibly to meet the requirements of every product variant. By using AI, pieces can be picked and handled individually in those environments.
As companies become increasingly digitalized, AI is also coming more and more to the fore. Still, only a small minority of companies (e.g. 6% of German companies) use AI systems – despite their potential to generate around 25% more profit and increase innovative power. Specifically, in Industrial AI a variety of innovative technologies and applications have been developed to make production processes more efficient, reliable and flexible.
Bottom line: Companies that shoot for getting future-proof for Industry 4.0 better tap into AI’s potential by looking for the right partners to build up trust and AI capabilities now!
Do you have any thoughts to share on Industrial AI, Digital Innovation or the Future of Automation? Let’s start a discussion in the comments.