Sensors Will Drive AI Growth in Manufacturing


In the broadest definition, a sensor is a device whose purpose is to detect changes in the environment, and then generate a signal or data based on those changes. All living organisms contain biological sensors. Most of these are specialized cells that are sensitive to light, motion, temperature, magnetic fields, gravity, humidity, moisture, vibration, pressure, electrical fields, or sound, to name just a few.

Over the years, many thousands of mechanical sensors have been developed to detect changes in their environments.  A partial list of sensors used to measure just pressure includes the following:

  • Barograph
  • Barometer
  • Boost gauge
  • Bourdon tube
  • Convoluted diaphragm
  • Hot filament ionization gauge
  • Ionization gauge
  • McLeod gauge
  • Oscillating U-tube
  • Permanent downhole gauge
  • Piezometer
  • Pirani gauge
  • Pressure gauge
  • Tactile sensor
  • Time pressure gauge

Besides pressure, sensors have also been developed to measure changes in sound, vibration, chemical composition, electric current, electric potential, magnetic force, radio waves, flow, fluid velocity, ionizing radiation, subatomic particles, navigation instruments, position, angle, displacement, distance, speed, acceleration, optical, light, imaging, photon, force, density, level, thermal, heat, temperature, proximity and presence, again to name a few. The variety and sensitivity of mechanical sensors currently extends well beyond the range of biological sensors.

The significance of the Internet of Things (IoT) is that it combines sensors with software and network connectivity to enable objects to collect and exchange data. The amount of data the IoT will generate will be vast. Some experts have estimated that the IoT will consist of almost 50 billion objects by 2020. Even now, sensors in the Large Hadron Collider (LHC) at CERN produce about one petabyte of data every day – the equivalent of around 210,000 DVDs.

Artificial intelligence (AI) was not a part of the original concept of the Internet of Things, but there has been a recent effort to integrate AI and IoT. The sensors’ ability to detect changes in the environment, identify internal faults and biases in sensors, and determine suitable mitigation measures now constitutes a major IoT research trend. Machine learning and deep learning in particular are promising technologies because of the possibility of using unstructured learning to place sensor data into context. Without context-aware automation, the value of IoT products to be deployed in real environments is likely to be limited.

By the same token, most AI research has focused on computer vision, machine learning, and natural language processing (NLP), not on the myriad of sensations biological sensors produce to provide living organizations context within their world. Widespread adoption of the IoT is likely to spur much more research in this area.

In our Artificial Intelligence for Enterprise Applications report, we forecast that spending on AI software in the manufacturing sector will grow from a modest $13.5 million in 2015 to over $1 billion by 2024. Much of this growth is likely to come from the training and calibration of sensors in IoT systems.

AIE-15 Manufacturing chart

Comments are closed.