To vacuum pumps
CW
CW CW
Solvent
CW
C-1
PS-01
Solid raw material
Liquid raw material
R-1
B-1
Outlet thermal oil
Intlet thermal oil
Acrylic resin
F-01
P-01
Figure 1 Production process of acrylic resin
on its knowledge of a process, and performs the following tasks: Identify the relevant parts for the HAZOP study. Apply a given guide word to a property and identify the causes and consequences of a deviation. Identify the existing safeguards present in the process to prevent a deviation and its consequences. Suggest additional safeguards to help improve the process safety. Generate a HAZOP study report. HAZOP study HAZOP study is a detailed analysis carried out by a team of experts to identify risks and operability problems. It involves identifying potential deviations from the design intent, the possible causes of these deviations, and assessing the consequences. It is based on a ‘guide word examination’ in which a deliberate search for deviations from the design intent is made. For this examination, a system is broken down into parts (also called nodes) in such a way that it allows the design intent or function of each part to be adequately defined. Parts can be equipment or components in a process or electronic system, individual signals and equipment items in a control system, discrete steps, or stages in a procedure, clauses in a contract, among other things.
After selecting a part for study, the design intent of the part is specified in terms of discrete properties, which convey the essential characteristics of the part. Each relevant guide word is then applied to each property, thus systematically carrying out a thorough search for deviations. Having applied a guide word, possible causes and consequences of a given deviation are examined, and mechanisms for controlling the predicted consequences can also be investigated.1 Applying hpad technology for HAZOP The hpad AI companion for HAZOP studies uses the enterprise AI model of a plant to perform the HAZOP study. For this purpose, hpad’s AI is taught the process for the plant, and the AI encodes it in the enterprise AI model. To build this model, a natural language description of the plant is fed to the AI, which selects the relevant elements from the plant description and encodes this information in the enterprise AI model. hpad uses a meticulously defined ontology to guide the AI to iteratively capture information from the plant description. Any new information about the plant can be added to an existing enterprise AI model of the plant. This approach allows for the creation of deep models of any plant iteratively as required for a particular use case. AI companion reasons using the encoded
Refining India
54
Powered by FlippingBook