Page 46 - Market Analysis Report of Optical Communications Field in China & Global market 2025
P. 46
Table 1. Predictive model evaluation table
Production line Predictive index 7training set_R 9training set_R XR_ratio 7training set_MAPE 9training set_MAPE MX_ratio 7test set_MAPE 9test set_MAPE MC_ratio
54.59%
(0,0) Zero dispersion wavelength 0.491272 0.75944 15.18% 0.0010408 0.00084282 -19.03% 0.0013964 0.0026359 88.76%
14.31%
(0,0) λc mean 0.733162 0.84442 3.86% 0.0068304 0.00507987 -25.63% 0.0116615 0.0020858 78.87%
3.86%
(0,0) λc(in)mean 0.719932 0.82298 0.0068612 0.00536076 -21.87% 0.0125892 0.0179060 42.23%
(0,0) 1310MFD mean 0.929181 0.96504 0.0026713 0.00186093 -30.34% 0.0059144 0.0091062 53.96%
(0,0) 1310MFD(in) mean 0.917445 0.95306 0.0029256 0.00216591 -25.97% 0.0062456 0.0096742 54.90%
By drawing and collecting the measured data, the difference control in the field of optical fiber preform manufacturing,
between the prediction results and the actual measured and with the help of embedded big data management
values is compared online to dynamically evaluate the model, it has mined the potential risks in real time and
prediction effect of the model, as shown in Fig. 4. When improved the emergency response capability.
the model prediction deviation meets the requirements, it 3.2.1 Intelligent Equipment. FiberHome has independently
is connected to the production control system to realize the researched and developed a series of key intelligent
intelligent optimization of the production process. equipments in the field of optical fiber preform
manufacturing, including core rod automatic deposition
Figure 4. Real-time deviation monitoring of big data AI predictions for and sintering integrated equipment, automated core rod
optical fiber preform quality precision extension equipment and intelligent outside high-
speed deposition equipment, etc., which provide hardware
3.2 Automation of the Production Process support for the construction of intelligent manufacturing
Unmanned or less-manned operation of the production workshop.
process is one of the core objectives of intelligent Taking the optical fiber preform outside deposition process
manufacturing, which can significantly improve production as an example, the precise control of flame temperature
efficiency, reduce human errors and ensure product and position is a key factor to ensure product quality.
consistency. FiberHome has realized a high degree of The traditional manual observation and adjustment
automation in the production process by introducing methods have limitations such as slow response speed
advanced automation equipment and systems. and difficulty in tracking flame changes in real time.
FiberHome has fully carried out digital-enabled production FiberHome introduced a flame state detection system based
construction and independently developed the digital twin on YOLOv8 algorithm, as shown in Fig. 6. The system
3D intelligent optical fiber preform factory control system, can detect the flame direction and color in real time and
as shown in Fig. 5. 1:1 digital virtual simulation of all compare it with the samples during training. Once the
equipment, real-time monitoring of all key parameters of deviation is found, the system will automatically trigger an
the optical fiber preform production equipment, and trend alarm and prompt manual adjustment, which improves the
analysis and early warning of important core parameters, safety of the production process and significantly reduces
which enhances the enterprise's production control capacity the need for manual intervention.
and data asset utilization rate.
Figure 6. Layered architecture of flame monitoring system
Figure 5. Digital twin 3D plant panoramic view
Through production automation and the development of 3.2.2 Intelligent Logistics System.On the basis of
digital technology capabilities, FiberHome has realized the production automation, FiberHome has further developed
advantages of state visibility, quality visibility and remote an intelligent logistics system applicable to optical fiber
preform production, realizing fully automated material
44 distribution and management of the production process.

