Industrial AI: an exciting now, a promising future

Siemens AI solutions understand industries better.

Integration of AI and industry

Many futuristic technologies that once appeared as mere figment of imagination in science fiction movies now exist in reality.


Artificial intelligence (AI), for example, has made its successful transition from reel to the real world.


Implementation of AI in products and services that people use daily, including smartphones, payment apps and automated cars, makes life easier and smarter. Since the outbreak of COVID-19, AI-based innovations such as thermal imaging cameras and intelligent disinfection robots have been playing a prominent role in keeping people safe.


In fact, the seeds of AI have been sown in a wide range of core industries including manufacturing, infrastructure, energy, mobility, healthcare, etc., and businesses are either reaping the benefits of AI right now or soon will be.


“Integrating AI into industrial scenarios will unleash its tremendous potential. We should focus on AI based innovation and applications in different industries to enhance performance of existing systems and bring digital transformation to a new level,” said Lothar Herrmann, President and CEO Siemens Greater China. “With more than 170 years of industrial experience, domain know-how and experts around the world, as well as the advanced AI technology, Siemens provides customers with secure and reliable AI solutions that understand industries better.”

Nipping potential problems in the bud

AI has become a game changer for traditional automotive manufacturing, opening up new ways to tackle exiting problems.


In the stamping process of automotive manufacturing, for example, vibration provides one of the earliest indicators of a machine’s health condition. Although vibration analysis acts as a powerful diagnostic tool, unless you are a vibration expert the information can often be difficult to decipher. And it is impossible for experts to track those activities 24×7, especially at production lines with advanced servo presses which involve constantly changing speeds, shifts and pressure that cannot be captured by the human eye. This is where AI can step in and make a difference.


Teams from Siemens Digital Industries Customer Services and from Siemens Corporate Technology China have set up a predictive maintenance system based on vibration analysis for Beijing Benz Automotive Co., Ltd. (BBAC), combining domain expertise with data science.


With its capability of recognizing vibration and other factors from more than 70 sensors installed into critical assets, the AI-powered system becomes an extension of experts’ minds able to evaluate equipment conditions in real time and predict possible faults. Each sensor collects more than 20,000 points of data per second. The system analyzes data with the support of machine learning models and provides a web-based user interface.


“One night shortly after its deployment, the system captured an abnormal vibration pattern, which triggered an early warning for maintenance professionals to proactively address the issue before equipment failure or downtime” said Zhou Linfei, Lead Research Scientist from Siemens Corporate Technology China. “This is how AI translates the intangible to the tangible for industrial customers.”

The “strongest brain” in a factory

Continuity of production is vital to process industries including petrochemical, chemical, cement, nonferrous metal and iron & steel. Unplanned shutdowns for a short time can cause huge revenue losses, and even trigger fires, explosions and other serious accidents.


With the support of Siemens Predictive Analytics (SiePA) system, Sinopec Qingdao Refining & Chemical Co., Ltd. (Qingdao Refining & Chemical) has enabled a closed-loop mechanism of predictive maintenance in its intelligent factory aiming to ensure continued safe production. SiePA consists of two major functionalities: risk prediction based on machine learning and advanced diagnosis based on natural language processing.


As the “strongest brain” in a factory, the comprehensive data intelligence system SiePA sends a warning to users whenever there is a potential risk of equipment failure and recommends maintenance solutions effective for similar historical cases. Users are able to report results from the ongoing maintenance activities back to the system through an intuitive interface, which in turn optimizes machine learning models and enables knowledge transfer over time.


In addition to equipment data, SiePA system also analyzes a vast amount of data available in the industrial environment and provides predictive insights into the overall production status, which helps ensure product quality throughout manufacturing processes. This could also be a great step forward for pharmaceuticals, food & beverages, fine chemicals and other industries that involve batch production.


*SiePA has been shortlisted for Super Artificial Intelligent Leaders (SAIL) Award by 2020 World Artificial Intelligence Conference.

Agile, accurate and stable

Thanks to photovoltaic (PV) technology, any sunny rooftop can be converted into a solar power generator. Manufacturing companies installed with eco-friendly PV systems not only save on electricity bills for themselves, but also help relieve the pressure of local power supply in peak hours.


Rooftop PV modules have a lifespan of about 25 years. Just like other systems at home, the components of a solar system need regular checkups to make sure they are operating stably over the long term. However, periodic maintenance - system inspecting, performance testing and PV module cleaning - is neither cost-effective nor able to detect potential problems in a timely and precise manner. A more intelligent maintenance solution over the lifetime of a PV system is needed.


Aware of these challenges, teams from Siemens Smart Infrastructure Regional Solutions & Services and from MindSphere Application Center for Smart Infrastructure in China have developed a highly intelligent life-cycle maintenance solution - Digital Solar Diagnosis Engine (DSDE) - for PV power stations. Coupled with AI and IoT technologies, the solution enables module-level performance monitoring and malfunction diagnosis.


Dashboard of DSDE clearly displays key performance indicators including module failure, degradation, cleanliness and the overall status of power station, helping maintenance staff pinpoint problems and optimize operation strategies.

Future-oriented industrial AI

As AI penetrates into the industrial sector, challenges never seen in the consumer market have emerged. To make AI more accessible to various businesses, Siemens AI team is addressing these challenges through innovative research and development.


“We provide our customers with affordable and satisfying industrial AI solutions,” said Tian Pengwei, Head of Research Group at Siemens Corporate Technology China. “By doing so, we help our customers unlock the hidden potential of industrial big data.”


Acquiring sample data at scale is a prominent challenge for industrial AI application. Labeling of industrial data largely relies on the knowledge of domain experts, but manual labeling is both expensive and time-consuming. By combining neural network with Bayesian methods, Siemens AI team incorporates industrial prior knowledge into machine learning process, which reduces the dependence on sample data scale. Moreover, the team cuts the cost of data labeling by adopting Active Learning which refers to a procedure to manually label just a subset of the available data and infer the remaining labels automatically using a machine learning model.


The lack of AI know-how may also hinder the adoption in many organizations. To overcome this challenge, Siemens AI team is exploring Meta-Learning techniques for automated machine learning (AutoML), which allows the system to automatically choose the best-fit algorithm and hyperparameter values for each user. With that, people from diverse backgrounds can easily use machine learning models to address complex scenarios.


“The real value of AI in industrial fields lies in its ability to identify the correlations and changes of parameters in an invisible world of the industrial systems, thus predicting and preventing potential problems,” said Prof. Jay Lee, Director of Artificial Intelligence Center at University of Cincinnati, author of Industrial Artificial Intelligence. “Once AI penetrates verticals and releases its full potential, the entire industrial world shall become worry-free.”


As a dream enabler of industrial AI, Siemens is turning this vision into reality.

July, 2020