Abstract: - Traditionally, food and drink have been used as a means of survival. In recent years, as technology

has advanced and our quality of life has improved, we have begun to seek out better food - not just for the sake

of the food itself, but to consume wisely, considering not only the ingredients but also the manufacturing

process. Companies that used to simply produce and deliver products now need to provide services such as

quality, promptness, and information, especially when it comes to quality control and safety, and quality

management, which are directly related to the lives and safety of consumers. Many companies are introducing

quality management systems to provide quality that meets the needs of customers and the market. With the

development of information technology, quality inspection methods are becoming more diverse and

sophisticated. However, the risk of defects occurring in the process still exists. In this paper, when a process

fails due to an unknown cause, the analog/digital data of that process is selected and compared with normal

1 Introduction

Providing quality and reliability has become a major

challenge for companies that supply food and

beverages. Given that potential quality defects can

occur in any process that produces a product for

direct human consumption, it is essential to establish

a system of control and management throughout the

process, [1]. This is where Hazard Analysis and

Critical Control Points (HACCP), a food safety

management system, comes in. It aims to prevent

technology, it is possible to perform advanced food

safety management through automation and

improvement activities by incorporating IT

analyzed using time series decomposition

(production instruction information during the

problem period, aseptic water flow statistics, heat

exchanger discharge temperature statistics, etc.

seasonal, and residual data with time series

characteristics. The de-composition model can be

broadly divided into additive and multiplicative

factor decomposition, [4]. Additionally, since the

residuals of the time series decomposition follow a

identified a list of tags that affect the sterility failure

point, and identified precursor symptoms for

selected tags, [6].

function, and identify possible causes. Analog data

such as temperature, pressure, and flow rate of

facilities, digital signals of pumps and valves, and

production instruction data by period from MES

(Manufacturing Execution System) are collected,

and simulations are performed based on the data of

Second, I briefly describe the process and

environment where the problem occurred, compare

the situation pre-symptoms and post-symptoms the

aseptic failure, and my hypothesis. Section 4

describes the analysis based on time series data and

the process of finding the cause. Finally, Section 5

concludes with a discussion of the result,

implications, and future work of this thesis.

including agriculture, processing, distribution, and

retail. It is also driven by a variety of trends,

they occur. In particular, the causes of the pre-

symptoms, which will be discussed in this paper,

can be deduced as follows. The hot steam from the

engine room boiler. Or, water entering the heat

exchanger from the buffer tank.

follows.

1st, drop in heat exchanger discharge temperature

(Blue line).

2nd, decrease in water flow rate(Orange line).

3rd, after some time, the water flow rate switches to

a maximum(Orange line).

4th, after another time, a sharp drop in the heat

exchanger discharge temperature(Blue line).

The heat exchanger discharge temperature, which is

a criterion for sterilization, has not reached the

Two PCs are used as shown in the specifications

in Table 4 and Table 5. They were used for

programming and simulation respectively. In Table

IV, the two machines have similar specifications

except for the graphics card.

servers from which the data was loaded. The

Table 6. View Basic Information: DB Server.

heat ex-changer discharge temperature, which is the

baseline information that affects aseptic failure,

varies within the baseline.

log/digital data, facility operating history, and

production history data. Analog data is temperature,

pressure, flow, etc. collected by SCADA. Log data

component of the smart factory system deployed in

the factory. It is a component of the smart factory

system built in the factory, utilizing data from MES

and SCADA. Unlike the conventional method,

which takes a long time to determine the

corresponding tag by simply comparing normal data

with sterilization failure-related data, it is possible to

determine the corresponding tag by simply

comparing sterilization failure data with sterilization

the previous task (Data selected by checking outliers

in the residuals). However, as mentioned earlier,

instantaneous changes in values caused by already-

known causes are not meaningful for sterilization

separately select only data lasting more than 30

seconds.

baseline, [15].

abnormalities in the boiler engine and heat