Nikolay Zakharov, a technical expert at COMETAL, shared the results of a two-month experiment integrating neural networks into everyday engineering workflows. The specialist succeeded in delegating about 20% of monotonous tasks to algorithms, freeing up approximately 8 hours per week to focus on complex analytical and manufacturing challenges.
What Functions AI Performs
The expert highlighted nine key areas where artificial intelligence is applied either fully or partially:
- Design documentation completeness check: verifying the availability of all drawings listed in the specifications.
- Product weight verification: identifying discrepancies in specified weight parameters.
- Search for non-standard steel grades and materials: automated detection of rare items that are difficult to spot manually.
- Routing sheet generation: preliminary layout of the parts processing workflow.
- Bottleneck identification: detecting structurally complex elements that may pose manufacturing challenges.
- Cost estimation based on design documentation: preparing technical and commercial proposals.
- Express estimation based on technical specifications: preliminary calculation at market rates in the absence of detailed drawings.
- Forecasting structural steel detailing design timelines: estimating lead times for metal structure engineering.
- Summary sheets: structuring procurement data for purchased materials.
The Impact of Automation: Multiplying Efficiency
According to the expert's evaluation, the execution speed of standard operations increased 2 to 3 times. The most impressive results were seen in drawing completeness auditing. While an initial review of a document package previously took around an hour, integrating AI reduced this time to just 5 minutes.
As an example, calculating a batch of 24 welded steel molds with intricate detailing would have taken up to two working days of manual verification and spreadsheet compilation, whereas the neural network handled the task in 30 minutes without the risk of typographical errors. This not only relieved the specialist's workload but also significantly accelerated the delivery of commercial proposals to clients.
Challenges and Technological Limitations
Language models do not handle all tasks equally well. While specification checks and initial estimations based on technical requirements achieve maximum reliability ratings, comprehensive cost calculation based on complex drawings remains the most challenging stage. Errors in determining the correct machining route directly impact the final price, so all calculations must be verified by a human expert.
The Claude model is predominantly used in the workflow due to its high speed and lower rate of false positives compared to ChatGPT, which sometimes requests drawings for standard purchased fasteners or bearings.
Security and Process Adaptation
To comply with personal data legislation (Federal Law No. 152-FZ) and safeguard trade secrets, all documentation undergoes preliminary depersonalization before being submitted to the neural network.
The expert emphasizes that implementing AI is not aimed at reducing staff. The primary goal is to free engineers from routine burdens so they can focus expert resources on bespoke engineering solutions and responsive customer collaboration.
Comments
to leave a comment.
No comments yet.