How Smart Sensors are Revolutionising Industrial Automation
Published on : Wednesday 01-09-2021
The progressive use of sensor technology has opened an avenue of opportunities for smart industries.
Conventionally, sensors were used to gather field data, which were then sent to I/O modules and controllers to process and provide meaningful outputs. With intelligence coming down to component level, smart sensors are not only able to gain field information for a range of critical processes, but are also capable of processing data, and making decisions using logic and machine learning algorithms.
High-speed and compact electronics are re-shaping the embedded world1 to provide cutting-edge technologies. Industry 4.0 is the vision every industrial enterprise is seeking to actualise and designing innovative technologies is the doorway to achieve it. There has been a significant uptick in the use of smart sensors in the industries to process and gather field device information in ways unheard of before.
Currently, low footprint electronics are becoming a trend. Sensors with embedded intelligence, reduced power requirement, and low memory footprint, are opening avenues of new technology innovations. Multiple application-specific sensor-based solutions are surfacing enabling improved data collection, therefore revolutionising the embedded world.
Need for sensors in Industry 4.0
Sensors2 are predominantly used in various embedded devices and equipment for a range of estimation and sensing processes. The recent influx of sensor systems has taken industrial engineering capabilities to a new height. Let’s see the ways sensors are helping industries:
Building Automated Systems
Technology is doing wonders to the production systems. However, there is still a gap that keeps shop floor data from reaching the cloud in real-time. With sensors able to gather critical field data, multiple automated systems can be built for monitoring, quality management, etc. This allows even the smallest embedded devices to become more efficient and less reliant on manual interventions. Environmental monitoring, smart energy management, smart grids, emotion recognition, etc., are some of the vast-ranging solutions developed using automated systems.
Improved Production and Quality Management
Sensors are allowing plant data to be more transparent making critical details visible at the IIoT end. This enables better analytics facilitating increased manufacturing precision. As a result, the production performance is improved. In the coming years, sensors will be capable of connecting the manufacturing plant digitally driving new efficiencies by integrating physical and cyber technologies.
Detecting Manufacturing Faults
Industrial sensors on the equipment are capable of collecting a lot of data about the performance at peak efficiency. Having logs of data enables monitoring the device’s health and avoiding the cost spent in repairing the damages. This information passed to the factory managers helps them keep track of the new assets ready for provisioning or fix the failures/damages to reduce delays.
Assets and equipment used in factories/plants require periodic maintenance throughout their life cycles. Failure to conduct timely maintenance may cost a lot. However, maintenance is a labour-intensive activity that requires periodic site inspections by the instrumentation experts. Embedding sensors into assets helps gather critical device parameters allowing monitoring, calibrating, and controlling the automatic functions/engines, therefore aiding in forward-looking planning and effective maintenance.
As we have seen, sensors are capable of performing a vast range of industrial activities. However, the field-level communication has been predominantly happening over wired networks. This confines the network components within the structural limits (wiring, connecting components, etc.), and restricts scalability. With wireless networks in place, the network components can be easily scaled regardless of the structural dependence. This is easy to maintain and a cost-effective solution.
The ability to transmit data over the wireless network will not only allow the devices to be less reliant on extra nodes for data processing, but will also offer greater flexibility for detection/estimation operations.
Wireless sensor networks
The recent advancements in Microelectromechanical Systems (MEMS) have allowed the embedded firms to engineer low-cost, low-power, and low-footprint sensor nodes that are capable of sensing, processing, storing data, and communicating the information to other sensor nodes. These small-scale sensor nodes can also form a wireless network over geographical areas, called Wireless Sensor Network (WSN).
The WSN formed among spatially distributed sensors consists of one or more Sink nodes, otherwise called base stations. The sink nodes collect data from other sensor nodes and communicate to the end-user using direct connections, internet, wireless networks, etc. A sensor node can act as a data originator as well as a data router.
Based on the applications, the WSNs are categorised into four types:
1. Wireless Multimedia Sensor Networks (WMSNs): Proliferated with the advent of Integrated Complementary Metal-Oxide Semiconductor (CMOS) camera sensors and integrated microphones, WMSNs are capable of capturing multimedia video and audio streams. A suitable application of WMSNs is real-time multimedia surveillance deployed in the areas prone to hazards or accidents like process control plants or traffic control systems.
2. Mobile and Robotic Wireless Sensor Networks (MRWSNs): Mobile and Robotic Wireless Networks consist of robots with sensor nodes, where the network has the ability to change the deployment and the coverage autonomously. MRWSNs can be apt for tracking assets’ movements in plants.
3. Underground and Underwater Wireless Sensor Networks (WUSNs): Besides the terrestrial environments, WSNs can also be used for underground areas such as caves and mines, dense soil or rock, etc., to monitor environmental conditions, such as detecting toxins in the soil, monitoring underground infrastructure activities, etc. Similarly, underwater WSNs can be used for subaquatic explorations, evaluating scarce resources, etc.
4. Space-Based Wireless Sensor Networks (SWSNs): The way WSNs are applied in terrestrial environments, they can also be used in space, i.e., Satellite Sensor Network. In this case, a large number of low-complexity satellites form a network to accomplish multiple goals, like continuous earth coverage for multipoint remote sensing, communications, chemical or physical sensing of soil, surface, etc., of other planets.
The applications of wireless sensor networks are ever-growing. Additionally, there are areas (remote location, extreme weather conditions, hazardous areas, etc.), where such smart networking can bring efficiency and flexibility. The embedded devices exposed to such conditions are expected to process data with limited processing ability, memory footprint, and power.
This has given birth to the whole new segment of low-footprint embedded devices. Modern power-optimised electronics have microcontrollers/sensors embedded into them, which make them suitable for deploying small machine learning algorithms. The ability to process the logics and draw inferences has made TinyML one of the most talked about subjects in the embedded world.
Intelligence in sensors – TinyML
As per the network architecture, sensors are responsible for processing distributed signals that help in two ways:
1. Eliminating noise, and
2. Making the useful object features clear.
The intelligence in sensors improves the signal selection capability of every sensor across deployed across the plant.
Conventionally, analysing, processing and making critical cognitive decisions have been the prime tasks of the IT layer. This requires collecting data from the field level and sending the data to the cloud to make sense out of it. However, critical operations may not always wait for the information from the cloud to take action. Also, transferring data from the field devices to the cloud also consumes a lot of power. Not only this, applying machine learning algorithms and drawing inferences require high processing power and a larger memory footprint.
Nevertheless, if intelligence is brought down to the field level, many of these challenges would be resolved. TinyML serves this purpose. TinyML is a concept that allows machine learning logic to be deployed on small microcontrollers/microprocessors, hence the name. By bringing intelligence to the field level, TinyML empowers small embedded devices like sensors to process data and make decisions.
This allows the embedded devices to respond faster enabling time-critical operations to be carried-out in a smoother fashion.
Industrial use cases of smart sensors
The progressive use of sensor technology has opened an avenue of opportunities for smart industries. Machine learning and deep learning applications have been thriving in recent years, and development in hardware has been the major driving force for it. Intelligent sensors are monitoring the complex industrial assets and analysing the signals for condition monitoring. Some of the prominent ways smart sensors are being used in an industrial context are:
Condition Based Monitoring
Condition-Based Monitoring (CBM) is an approach used for generating prognostics to avoid field device failures and reduce maintenance costs. Using smart sensors embedded into the integrated chips, processes like field data capture, data processing, local diagnosis, local feedback, especially State Detection (SD) can be managed efficiently. With complete visibility of the device data, data acquisition and initial data manipulation concentrate in the sensors, easing the maintenance, production, and logistics processes. This technique can prove to be effective in monitoring wind turbines, where the blades are crucial and expensive. Bi-Coherence Condition Monitoring can easily detect small physical changes in the machine from a noisy signal.
Image sensors are widely used in machine vision cameras. These image sensors are used for a wide range of applications in robotics, pattern recognition, information detection, etc. The need for high performance, low-footprint, and reliability has paved the way for integrated sensors that combine all the image sensing functions. For example, an Infrared Imaging sensor can operate irrespective of the light conditions. Due to this ability, IR imaging sensors are used in marine, military, air forces, etc., to capture thermal images of the objects.
Functional safety is an important aspect of the safety measures taken to maintain the industrial system. A system or equipment should respond to the potentially dangerous conditions by activating a protective/corrective mechanism. For example, a linear position transducer can be used in safety circuits. This type of sensor works on magnetostriction and determines the position, distance, and/or velocity of the objects in a plant.
Smart sensors are evolving as we speak. More and more intelligent systems are equipped with improved sensing capability and self-awareness that form the core for innovative intelligent systems. Bringing intelligence to the component level has a profound impact on industrial applications enabling sensing, monitoring, and responding to various transformations to optimise overall performance. With constant innovation and research, a whole new segment of intelligent sensors is soon to be seen built with micro and nanotechnologies. As machine learning is swiftly grabbing pace, a range of automated systems are en-routed with neural networks, machine learning, and deep learning taking the centre stage.