MEDICAL GAS SYSTEMS
10.0 9.9 9.8 9.7 9.6 9.5 9.4 9.3 9.2 9.1 9.0
◆ ■ ◆ ■ ■
◆ ■
■ ■◆ ■ ■
■ ◆ ■ ◆ ■ ◆
◆■■■◆◆
◆■ ◆ ■ ◆ ■ ◆ ■◆◆■ ◆
■ ◆ ■ ◆ ■ ◆ ■ ◆■■◆ ◆ ■■ ◆ ■ ◆ ■ ◆ ■ ◆ ■ ◆ ■ ◆ ■
n Predicted uMeasured
◆
■ 0
20
◆ ■
◆ ■◆◆■ ◆ ■ ■
◆ ■◆ ■
■ ◆ ■ 40
◆ ■
◆ 60 Time domain Figure 2. An example of control parameter prediction.
transmitter, and distributed cloud, both knowledge-based and data-driven AI techniques can be applied to correct errors and perform predictive control and maintenance to prevent failure without unnecessary interruptions, improving energy efficiency. Through the analytics of real-time data collected from IoT sensor networks, the AI system has enabled confident decision making and made it possible to optimise the controls and maintenance and to achieve maximum power efficiency and productivity. To evaluate the proposed real-time
AI system with DWNNs constructed for predictive control, an unmanned medical air plant was set up, which consists of four compressors and two dryers to compress and purify air for hospital users. To ensure all compressors and dryers provide optimum performance, predictive control strategy is employed with system parameters (such as pressure) predicted (an example is shown in Figure 2). After training the designed DWNN network by minimising the error between the actual outputs and desired outputs to determine the network parameters, it is used online to continuously predict the control parameters based on the past measured data to optimise the control strategy. The DWNNs training and processing is performed at the cloud server with optimized control parameters adjusted and sent back to the edge for optimal process control. Based on the choice of fixed speed compressor, with condition monitoring for medical compressed air plant, the real-time AI enables power efficiency to be optimised. From the tests performed, it is estimated that up to 30 per cent of energy costs can be saved through real-time data analytics by the DWNNs developed. As for faults detection, the presence
of faults in the MGSs was verified in an unsupervised manner. The system could diagnose each fault and perform cause analysis based on importance ranking
38
explained by the trained DWNNs model, where a fault can be related to a potential or unique type of component fault, and even to more than one fault, and then a knowledge-based system can be established from which the cause is analysed and solution is sought. For example, by data analysis, a fault detected causing oxygen generator shutdown was diagnosed and it can be attributed to very high demand of oxygen during the pandemic. When the flow rate sensor values increased at a very high level, the system could not generate enough output due to the limited capacity and that led to a system failure. Experimental results have
demonstrated that the real-time AI system is able to achieve optimum performance of medical air plant with power efficiency optimised by conducting qualitative and quantitative data analytics over IoT networks and performing predictive control and maintenance for the optimal operations of MGSs. By addressing key issues of deep learning, the newly developed DWNNs with Mexican hat wavelet as activation functions can achieve optimal network structure and parameters initialisation to improve accuracy and computational efficiency of the control system. With complex MGSs modelled and monitored by the trained DWNNs through multi-sensor measurements over the IoT networks, the real-time AI system designed can give valuable insights, track everything that is relevant to operations, and perform qualitative and quantitative data analytics to monitor machine conditions, detect faults, and predict and optimise system performance.
Conclusions To address the challenges and issues found, this paper presents a novel AI enabled system with technological innovations to make production processes and decision-making more efficient, autonomous, and adaptive using artificial
intelligence (AI), cyber-physical systems, big data and Internet of things (IoT). The real-time AI system can achieve optimum performance of medical air plant with power efficiency optimised by predictive control. By conducting qualitative and quantitative data analytics, the real-time AI system allows for valuable insights, tracking everything that is relevant to operations, monitoring machine conditions, detecting faults, and predicting and optimising system performance. Experimental results demonstrate that optimal operation of MGSs can thus be achieved by the real- time AI, where the sensing capabilities and the computational power are provided by the designed controller, transmitter and distributed cloud, so that by deep learning, both knowledge-based and data-driven AI techniques can be applied to correct errors and perform predictive control and maintenance to prevent failure without unnecessary interruptions, improving energy efficiency and maximising productivity – helping to digitise the NHS and delivering continuous innovations to better serve the healthcare industry.
References 1 Qiu T, Chen N, Li K, Atiquzzaman M, Zhao W. How can heterogeneous internet of things build our future: a survey. IEEE Commun Surv Tutor 2018; 20: 2011-27.
2 Luo G, Luo Y, Gan H. Predictive control with energy efficiency enabled by real-time machine learning. Proceedings, 1st IFSA Winter Conference on Automation, Robotics & Communications for Industry 4.0. 3-5 February 2021: Chamonix-Mont-Blanc, France, 007:34-38.
3 Luo G, Luo Y, Gan H. Real-time faults prediction by deep learning with multi- sensor measurements over IoT networks. Sensors & Transducers 2021; 251 (4): 1-10.
4 Dong B. Sparse representation on graphs by tight wavelet frames and applications Applied and Computational Harmonic Analysis 2017; 42: 452-79.
IFHE DIGEST 2023
◆ ■ ◆ ■ ◆
◆ ◆ ■■ ◆
◆ ■
◆ ■■ ◆ ■
◆ ■ ◆
◆ ■ ◆
◆ ■ ■
■
◆ ■
◆ ■ ◆
■
◆ ◆■■◆ ■ ■ ◆◆ ■
◆ ■
◆ ■ ◆
◆ ■
■ ◆
■ ■ ◆
◆ ■
◆ ■■■◆◆ ◆ ◆◆ ◆
■■ ◆
■ ■
◆ ■■ ◆
◆ ■ ◆ ■ 80 ◆ ◆ ■
◆ ■
■ ◆◆ ■
◆ ■
◆ ■ ◆ ■
◆ ■
◆
◆■■◆ ◆
■ ◆ ■ 100 120
◆ ■ ◆ ■◆ ■ ■
◆ ■ ■
◆◆ ◆ ◆
■ ■ ■ ◆ ■ ◆ ■ ◆ ■ ◆ ■ ◆ ■
10.0 9.9 9.8 9.7 9.6 9.5 9.4 9.3 9.2 9.1
◆ ■ ◆ ■
◆ ■
◆ ■
◆ ■ ◆ ■ ◆ ■ ◆ ■ ◆ ■ ◆ ■ ◆ 100 102 104 106 Time domain 108 110 112 ■ ◆ ◆ ■
n Predicted uMeasured
IFHE
Pressure (bar)
Pressure (bar)
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88 |
Page 89 |
Page 90 |
Page 91 |
Page 92 |
Page 93 |
Page 94 |
Page 95 |
Page 96 |
Page 97
