Machine learning has already delivered remarkable results in certain niches where pattern recognition is obvious, but it’s making even bigger and longer lasting impacts on businesses that demand broad insights and efficiencies in their industries. The investments of tech giants in machine learning applications are drawing a lot of attention. Google’s largest collection of developers outside its US headquarters is a research group dedicated to machine learning. Microsoft open sourced CNTK, Baidu released PaddlePaddle, Amazon decided to support MXNet on AWS, and Facebook created two deep learning frameworks. The wave of machine learning applications in the consumer space will spill over into industry, which will help engineers and managers improve business operations with automated data analysis. In addition to driving innovation, machine learning offers practical, here-and-now business improvements such as operational uptime, production yield, and engineering efficiency.
Machine Learning Feeds on Data
The ability to network intelligent systems to improve data visibility is well-documented as both an Internet of Things (IoT) benefit and a Big Analog Data™ challenge. ABI Research (QTR 1 2017) says sensor and machine data from industrial equipment is expected to top 78 exabytes by 2020, and somewhere among all that data will be evidence of a machine failure, manufacturing defect, or critical validation test missed by today’s technology. Vast data sets will help train better models from machine learning algorithms and yield faster results, but only if they’re available. Today’s system designers need to view organized data collection as the first step to implementing machine learning technology and develop more comprehensive DAQ and management strategies for connected systems.
Dealing with design flaws during product development can be expensive, which is why design verification and validation test receives so much time, attention, and budget. Before machine learning can help focus costly engineering time on the product areas that need the most test and validation, historical test data needs to be organized and accessible.
Most manufacturers today screen for pass/fail conditions and save data for forensic analysis, calibration records, and genealogy. Some manufacturers use more advanced automated test methods, but machine learning models can help them screen for product defects regardless of root cause. Did the silicon-level components on the current build come from a new fab? Does the design include counterfeit components? Is the wave-soldering temperature off because of a faulty sensor? Endless anomalies can cause defects, so setting up test limits for all of them is not practical (or possible). Machine learning technology will alert manufacturing test engineers to defects missed in the design and test phases of product development.
Many companies in process manufacturing or other process industries have extensive databases of maintenance and operational data for their industrial assets. Maintenance engineers manually work with this data today, but future machine learning methods will process this data to classify operational states and detect anomalies. Properly trained systems will identify irregularities that need attention and alert maintenance personnel for troubleshooting.
Taking Advantage of the Edge
In many ways, the stage is already set for machine learning. The convergence of rugged processing and sensor fusion with machine learning will help engineers build better systems that can interpret data at the edge without needing to communicate with the enterprise stack. Some technology can already train and run models at the edge to give engineers the following system architecture options: model training and deployment in the cloud, at the edge, or both. Pushing intelligence to the edge with real-world signals reduces the latency of decisions and the need for costly infrastructure, which helps as billions of new devices come online and compete for limited bandwidth.
Platforms Will Harness the Power of Machine Learning
One key element to watch for is the incorporation of machine learning in technology platforms that help developers focus on new problems, save time stitching together adjacent technologies, and avoid getting lost in middleware. Engineers rarely want to spend time dealing with questions that have already been answered or deemed necessary only because of toolchains. What cloud analytics are supported? Whose cloud? Are there RTOS compatibility issues when deploying the models? Integrating machine learning into cloud, software, and hardware platforms will provide precurated technology stacks so engineers can focus on new challenges. “Huge opportunities exist to extract insights hidden in the disparate business information systems most businesses have today,” says Andy Timm, CTO of PTC. “A machine learning platform can quickly pull this data together along with newly acquired IoT sensor data. Let the platform do the non-value-add work so the business can focus on deriving actionable insight.”
Machine learning applications today can impressively find pictures of a dog in a photo album, but business leaders are looking to engineers, platforms, and the next wave of machine learning to help find uptime, yield, and efficiency in a sea of Big Analog Data.
Duncan Hudson, Chief Platform Officer, National Instruments