This Pump House Monitoring example demonstrates the capabilities of Edge Computing in small-scale or medium-sized applications; it helps these smaller applications stay up to speed with the quickly growing industrial world. Specifically, this example deals with Edge Computing applications in the water utility sector.
Water Pump Houses or Pumping stations are used to transport water between sites—often from supplier to consumer. For example, pump houses are used to transport water into canals, circulate water in treatment systems, and supply water to residential areas. Operation parameters which should be monitored in the pump house include the input power supply, power consumption, water flow, water pressure, and the pump on/off control. Often, the on/off control is based on a time schedule or based on demand. Monitoring can be set up remotely using wireless or wired network connectivity.
Conventionally, all the required sensors and peripheral devices are wired to a Remote Terminal Unit (RTU)/Programmable Logic Controller (PLC). Data is logged to the local memory, then sent to the utility control station via a cellular or ethernet connection for processing. Events which are logged in the data include power failures, pre-set threshold limit alerts, supply voltage fluctuations, and overload trips. While data is transmitted to a remote location, an operator must still visit the site to perform maintenance activities or fault recovery.
The following equipment is required for the traditional setup:
- Cellular Modem or Ethernet for remote connectivity.
- RS 232/RS 485 serial ports as well as analog inputs and outputs for energy meters, flow meters, and pressure sensors
- Digital inputs and outputs for pump on/off control and basic security e.g., panel door open/close, power failure, or overload trip indication
- Limited memory for various event log/history data storage
The traditional system has some limitations. The first of these is a high dependency on network connectivity to get real-time alerts. If there is a network issue, operators could miss an alert which requires immediate action, increasing system downtime. Additionally, if the system uses a wired network connection, there is limited interface with peripheral devices installed inside or outside the pump house. This may lead to reliability issues.
A second limitation is the need for operator visits in case any system-level diagnosis or maintenance activities are required.
If the data indicates an efficiency loss, there is no way to remotely determine the cause. Similarly, there is no mechanism for preventive maintenance. This leads to high costs for maintenance and failure recovery. It is difficult and costly to install a preventive maintenance feature in a traditional system.
Lastly, there is no security on-site. The pump house is left vulnerable to theft. Using a traditional solution, installing security requires IP cameras, a dedicated server, and integration into the user control room in real time.
Edge computing makes it simple and cost effective to remedy these problems. Edge computing is done at or near the data source rather than relying on the serve to do the complex analytics work. This makes real-time monitoring and control applications feasible.
The following components are required to install edge computing in a pump house:
- High speed (800 MHz~1 GHz) Processor
- Linux Operating System -2GB RAM, 4GB Flash
- Video Processing Unit (VPU), Image Processing Unit (IPU), 2D/3D Graphics Processing Unit (GPU), Asynchronous Sample Rate Converter (ASRC) hardware accelerators
- Pre-trained data analytics models for motion detection, human detection, and pump health status monitoring. The models will be trained with ML frameworks such as Tensorflow, Keras, Pytorch, or Mxnet on a powerful system such as a Deep Learning server. Then, the model that is trained will be exported to the edge device.
- Open-CV or Pillow library for IP Camera initialization and Image/Video Pre-Processing
- Cellular Modem, Ethernet Port, POE+ Port, RS 232/RS 485 Port
- Analog and digital inputs & outputs
- Zigbee/Bluetooth wireless interface options
- HART Interface option
The diagram below shows how Edge Computing can replace and improve upon any conventional system. First, it reads all the parameters from various field sensors, whether wired or wireless. In the pump house, this may include pump vibration sensors, water leak detectors, IP cameras, and HART sensors. Then, the platform processes the input parameters and passes the appropriate ones to the AI Data Analytics tool on the Edge Platform. Based on the data analytics, the Edge Platform controls the pump house systems in real time. In addition to taking pre-programmed corrective action, it triggers alerts for potential pump failure, water leakages, security concerns, and more. It logs all events and reports them to the control station. It also maintains a log of historical data for a user-defined period.
There are many advantages to installing an Edge Computing system in your pump house. The Edge Platform makes the pump house function without a critical dependence on network connectivity; it can log data and take corrective actions in real time. Additionally, systems processes can be remotely accessed and controlled—even remote debugging is possible. It also makes wireless connectivity to peripheral field sensors possible using Zigbee or Bluetooth, solving reliability issues in long wiring outside the pump house.
Additionally, operators will seldom have to visit a pump house equipped with Edge Computing. Built-in features such as HART monitor and control configuration based on smart field sensors, removing the pump house’s dependence on humans for configuration changes and field sensor calibration.
POE+ ports support IP cameras so that the premises can be remotely monitored for unauthorized access. The system also provides better security for your data since it remains on the edge device itself rather than being transmitted to the cloud.
Importantly, the Edge Computing system also predicts pump health status. This reduces costs and downtime associated with failure recovery by alerting the maintenance team of problems before a critical failure occurs.
After implementing the Edge Platform, data processing is faster, responsive actions can be taken in real time, system monitoring is faster, and data analysis is simpler. All of this is achieved using minimal system components which enhance system reliability at an optimal cost.