Page 45 - Market Analysis Report of Optical Communications Field in China & Global market 2025

The core framework of optical fiber preform intelligent
factory mainly focuses on data collection and management
informatization, production automation and management
and control intelligence, which promote the development of
the factory in the direction of intelligence, high efficiency
and high quality from the three dimensions of management,
production and quality respectively. FiberHome is guided
by the lean production concept, integrating new-generation
information and telecommunication technologies, advanced
manufacturing processes and automation technologies to
build a whole-process visualization management system
for optical fiber preforms production, and to create a highly
automated intelligent manufacturing workshop. The overall
framework is shown in Fig. 1.
Figure 1. Core framework of optical fiber preform smart factory
3.1 Production Data Informatization and Intelligence
Data collection and management informatization is the
prerequisite for realizing big data application. Through
digital technology innovation and system integration,
FiberHome has realized the informatization and intelligence
of data collection in the whole process of optical fiber
preforms. No manual data entry is required, and the system
records the key elements of the production process online
through signal acquisition and control procedures, realizing
the dynamic monitoring of the whole manufacturing
process of the production and significantly improving the
efficiency and accuracy of data collection. The system is
also able to identify abnormal indicators in the production
process in real time, ensuring stable and controllable
production. In addition, the system is equipped with data
traceability function, and all the indexes of optical fiber
preform core rods and optical fiber preforms can be queried
online, which provides strong support for product quality
control.
Take the quality prediction of optical fiber preforms as an
example, because the quality characteristics of optical fiber
preforms can not be completely measured in the production
process, they must be inspected by measuring the quality
characteristics of the optical fiber after they are drawn into
optical fibers. In order to reduce the risk of defective optical
fibers and reduce the cost of optical fiber verification, a
multi-layer neural network regression analysis technique
is used to establish a quality prediction model for optical
fiber preforms, whose operation mode is shown in Fig. 2.
Based on the MES system, data acquisition system and
big data analysis system, the 4M1E elements of the optical
fiber preform production process are taken as the input
factors and the optical fiber quality characteristics are taken
as the output factors, and the dynamic prediction AI model
is established through the multi-layer neural network
algorithm to realize the real-time optimization of the AI
model. The big data platform monitors the accuracy of the
prediction in real time by comparing the prediction results
with the actual quality characteristics data. The specific
process is as follows.
Figure 2. Big data-based quality prediction model
First, feature selection experiments were conducted
through Python's SelectKBest method, traversing datasets
with dimensions ranging from 1 to 77, and comparing
the effectiveness of the filtered features with the model
prediction when using the full amount of features, and Fig.
3 shows the percentage of the effectiveness of the AI model
prediction graph line.
Figure 3. AI model prediction results
The differences between the model and the training
data were then reviewed to characterize the effect of the
predictive model through the linear regression coefficient
R2, as shown in Table 1.
43