Evaluation of network conditions on the performance of an Industrial IoT control and monitoring system

Abstract

Advances in smart manufacturing and Industry 4.0 have drawn the interest of the industry. The rise in popularity of internet of things and cyber-physical systems in a manufacturing environment meant the broader connection of devices, servers, sensors and actuators within a closed-loop control system. This study aims to examine the impact of di erent network conditions on the performance of an industrial control process using both local controllers and a remote system controller. Network conditions examined are network node location of the remote controller, latency, packet delay variance (jitter) and packet loss. The study uniquely uses both industrial hardware and communication equipment from Siemens© for the emulated industrial plant, as well as consumer networking equipment for the connection en emulation of a cloud-connected remote controller. An industrial plant is emulated through a system of electric motors. The emulated system uses both local controllers and a remote controller. The local controllers are responsible for direct control and monitoring of a motor, including failsafe features for the motor. A remote controller is then responsible for control over the whole interconnected system. A considerable increase in system response time is observed when the remote controller is moved from a LAN connection to a WAN connection. A remote controller connected to a LAN connection is described as a fog node, while the remote controller connected with a WAN connection is described as a remote cloud node. Even with delay mitigation implemented on the remote controller connected via WAN, lower system performance is still observed than the remote controller connected via LAN. Applying a double exponential smoothing model as delay mitigation on theWAN connected remote controller improved system performance over not having any delay mitigations. However, performance is better when via LAN compared to WAN with delay mitigations implemented. Relationships between the average system response time and the di erent network conditions are made through curve tting of the measured data. As network conditions worsen, the performance of the system is degraded in a linear and sometimes quadratic fashion. The relationship between system response time and network latency is slightly quadratic, the relationship between jitter and system response time is linear, and the relationship between packet loss and system response time is quadratic. The performance impact of di erent network conditions are measured for a system with and without delay mitigations implemented on the remote controller: two di erent delay mitigation mechanisms where tested, exponential moving average and double exponential smoothing model. The exponential moving average proved ine ective for the emulated system by yielding lower system performance per test point, likely due to exponential moving average being more suited for small changes, and the implementation thereof increased processing time. Implementing a double exponential smoothing model as delay mitigation increased system performance for each test point. Double exponential smoothing model boasts a better prediction ability, making it more suited as for delay mitigation. Some future work could focus on testing the scalability of an IIoT system, as well as how implementing security protocols could impact the performance of an IIoT system. Tests should be done to determine how scaling the system or adding security protocols to the system with varying network conditions will impact the performance of an IIoT system, as the ffects of scalability and security will be compounded along with the performance impact of different networking conditions.