Refining India September 2025 Issue

knowledge of the plant and carries out the required analyses. HAZOP of acrylic resin production plant In this case study, to demonstrate the AI technology, a HAZOP study is carried out on an acrylic resin manufacturing plant. Taheri, et al. have described such a plant. For the purposes of this demo, the acrylic resin plant has been assumed to be part of a fictitious company called ChemTech Innovations, Inc. The process flow diagram of this acrylic resin manufacturing plant is shown in Figure 1 .² Building enterprise AI model of the plant Our understanding of this process was described using natural language in a document. This document was then fed to the hpad AI, which created an enterprise AI model of the process with the help of the hpad ontology. To create the enterprise AI model of the plant, the AI breaks down the process description into three parts:  Equipment and its functions: The AI first captures the information about the manufacturing equipment in the plant (see Figure 2 ). Equipment type and its functions form the first layer of the enterprise AI model. This becomes the base for further build-up.  Process streams: Process stream details are then laid onto this existing model. AI captures the direction of flow and materials for each process stream (see

CND01

HEX01

PSP01

REC01

MTK01

Figure 2 Capturing the equipment

sensors are measuring. In this model, valves have been used as the final controlling element, where each valve is connected to a sensor and receives signals from it. The connection between sensors and valves is also encoded in the enterprise AI model. Further, the relationship between each valve and the corresponding process streams is added to connect this new information with the existing model. Similarly, sensors are also related to the equipment or streams whose properties are being measured (see Figure 4 ). Thus, an enterprise AI model of the process is developed in these three steps. Further additions, such as detailed control system design, equipment design data, utility information, other process parameters, can also

Figure 3 ). It also relates the equipment to these process streams by assigning the flow direction from one equipment to another.  Sensors and controls: After adding the process streams, control system is fed to the enterprise AI model. This consists of information about the sensors present in the process and the properties that these an abstract level description of the

STRM10

STRM12

STRM07

STRM11

HEX01

STRM08

CND01

PSP01

STRM06

Dibenzoyl peroxide

STRM02

STRM09

STRM04

STRM01

Methyl methacrylate Butyl acrylate Styrene Acrylic acid 2-mercaptoethanol Solvent

Solvent

REC01

STRM05

MTK01

STRM03

Thick resin

Finished product

Figure 3 Adding process streams to the enterprise AI model

Refining India

55

Powered by