AI-powered O&M data preparation with bluebee® X LongAI Document Extractor

O&M (Operation and Maintenance) data preparation is a crucial phase in the implementation of a CMMS (Computerized Maintenance Management System), either a specialized system such as Coswin or bluebee® or the Plant Maintenance (PM) module of the corporate ERP (Enterprise Resource Planning) such as SAP.

O&M data preparation involves collecting, checking, and structuring maintenance-related data, usually working from a large number of suppliers’ technical documents, to fill the data into Excel templates for upload to the system.

In most projects, while the software is technically “Live”, most O&M data is missing, making it unusable. The IT department and external consultants expect the client’s maintenance team to collect O&M data. In practice, which may take months or even years. This leads to the commonly heard statement: “the maintenance team was not mature enough to use such system”: users are blamed for project failure.

O&M data mostly comprise:

Each data type poses specific challenges, but Maintenance instruction is the most difficult. The information is often buried inside O&M or technical manuals in PDF format. Suppliers use different format, structure, and vocabulary. The typical project may include anywhere between 100 to 10,000 such documents. Typically, a team of maintenance specialists need to collect and classify every O&M manual, read them, locate, and identify data, manually extract it and fill-in Excel templates. The massive workload, the tedious nature of the work, result in delays and errors. Many companies simply give up.

Over the past 20 years, Siveco China has worked alongside infrastructure and plants owners and EPC (Engineering, Procurement and Construction) companies to ensure smooth transfer of technical documentation from construction to operation through the CMMS. The company developed a methodology, ISO 9001-certified procedures and IT tools to streamline the O&M data preparation process. IT tools, part of the company’s bluebee® Smart O&M software use, have primarily focused on extracting asset data from Piping and Instrumentation Diagrams (P&IDs), BIM models, etc.

For many years already, Siveco China has utilized Artificial Intelligence (AI) for predictive maintenance and optimization purposes. With the rapid development of Large Language Models (LLM) in 2023, Siveco China teamed up with specialist firm LongAI to tackle the main bottleneck of O&M data preparation: Maintenance Instructions.

LongAI’s Document Extractor was optimized for O&M data preparation purposes based on Siveco China’s experts input and is now included in the bluebee® X suite of tools. While standard models have an accuracy of around 70% (unusable for industrial projects as the quality control process would essentially require doing the work all over again), the maintenance-specific model understands the language and terms, the different manuals structures, and relations between content elements, reaching an accuracy of 95% and above, comparable to human processing (the estimated accuracy of a junior engineer is 87%, that of an expert 98%).

Furthermore, the tool’s front-end and back-end have been adapted to the CMMS application. Starting from a raw input of all the O&M manuals in PDF format, the bluebee® X LongAI Document Extractor outputs data in a structured Excel format, ready for quality control within bluebee® X, a maintenance-friendly environment in which users can view the resulting data in context (for example work instruction, work frequency, linked to related equipment and spare parts, with attached reference documents).

The new LLM-based document extractor further enhanced the data preparation approach adopted by Siveco China for many years by increasing automation into the elusive Maintenance Instruction preparation process.

Thanks to the 95% workload reduction, expert’s time is reallocated to quality control and to work on optimizing the maintenance strategy. Accuracy is comparable to that of maintenance specialists performing the same work manually.

The tool is currently being applied to a major infrastructure project in Southeast Asia. Thanks to its nature as a LLM, it is continuously being optimized, improving accuracy and ease of use.

Contact Siveco China (info@sivecochina.com) to discuss how our O&M Data Preparation services can contribute to the success of your corporate ERP’s Plant Maintenance implementation or more generally for your CMMS needs.

Siveco China (www.sivecochina.com), founded in 2004, is a pioneer in Asset Management and CMMS consulting. Its R&D unit Bluebee Technologies designs innovative yet practical Smart O&M solutions known as bluebee®. Services include Asset Management consulting, Maintenance 4.0 & Risk Prevention solutions, Mobility, IoT integration, BIM solutions for O&M. Siveco China has over 1000 customer sites in public infrastructures and industrial plants all over Greater China, Southeast Asia and the Belt & Road.

LongAI (www.long-ai.com) is a leading generative AI startup in the APAC region with a vision to empower multinational companies to harness the efficiency gains of generative AI in a secure and privacy-compliant way. The founding team consists of experienced international startup founders from the fields of Data Analytics, Artificial Intelligence, and Natural Language Processing. Experience include LLM Agents, Process Automation, Data Integration, Local Deployment, Data Analytics, Information Retrieval.

Tags： AI、bluebee® X