Case Study: Intelligent Automation In Cost Estimation
Using computer vision we automated the process of interpreting schematic drawings of oil refineries for the purposes of extracting a bill of materials.
Outcome: Increase In Output By Factor Of 100.
Problem Description
When building new refineries oil and gas companies engage engineers to design the refinary. The design is typically captured in a schematic drawing. These schematic drawings contain components such as pipes, fittings, valves and instruments. Prior to building a refinary a bill of materials is estimated from the schematic drawing. This is a manual process called a "take-off", where a civil engineer aggregates relevant components. The extraction of relevant components is not just a matter of lookup, it involves the application of complex domain rules. A typical process is as follows:
- A collection of schematic drawings, some of which stretch across multiple PDF documents, is submitted for cost estimation
- A team of cost estimators manually reviews the drawings and extracts a summary of materials used
- The extracted bill of materials is uploaded into an engineering system for cost estimation
- The cost estimation system generates a final cost estimate
The entire process is costly since it involves a manual review. It is slow and also error-prone. When reviewing schematic drawings across multiple pdf documents, it is easy miss something.
Reliancy Solution
We used Intelligent Process Automation to semi-automate what cost estimators do. The intent was to provide an assistive technology rather than to replace cost estimators. The solution consisted of the following steps:
Study Of Cost Estimator Thought Processes
Our first step was to thoroughly study how cost estimators do their take-offs. We categorized scenarios into those that are simpler and those that are more complex in nature.
Implementation of Computer Vision Detection System
A computer vision system was built to interpret a schematic drawing and to convert it into a usable digital representation.
Implementation of Automation Logic
Given the inferred digital representation of a schematic drawing we automated about 90% of the cost estimation process. Our focus was on simpler more managable scenarios. The complex and rare scenarios we left for the cost estimator to address.
Flagging of Sub-Tasks That Require Human Review
A system was built to flag areas of the schematic drawing where our automation could not be done with a high degree of certaintly. Once the processing was done the cost estimator was expected to manually inspect the flagged areas.
Impact
Our solution significantly improved the process of cost-estimation:
Improved Speed of Processing
Previously it took a civil engineer an entire day to process approximately 8 drawings. With the Reliancy system 100 engineering drawings can be processed and reviewed for accuracy within 60 min.
Improved Quality
Providing an assistive technology helped cost estimators reduce and catch errors. Manually processing huge drawings which might not even fit on a single page is a non-trivial task for a human.
Improved Profitability
The increase in efficiency allowed cost estimators to significantly increase their output and cut costs. From a business standpoint this translates to improved competitiveness.