Mindtree, a Larsen & Toubro Group Company, was a Gold Sponsor at this year’s virtual ARC European Industry Forum. Business strategies and models are undergoing a sea change, and it is clear that flexibility, agility, and innovative approaches are required to succeed during these uncertain times. After giving a brief overview, Prabhu Venkatramanan, Head Digital Technology and Solutions, Larsen & Toubro, spoke about the applications and use cases of artificial intelligence (AI). His entire presentation can be viewed here.
Need for Artificial Intelligence
India headquartered Larsen & Toubro, with a revenue of $21 billion and presence in over 30 countries, specializes in technology, engineering, heavy manufacturing, and construction, which is one of their largest businesses. However, surveys and reports revealed that the construction industry was a laggard and there was huge potential to improve processes. Hence, it became imperative to understand how digital interventions can improve productivity and efficiency. Focus was on major areas, such as safety of construction equipment and the associated workers. Another aspect was to gauge project progress across multiple global locations with different quality requirements and restrictions. A wide variety of technologies were used to do all this, especially IoT for the connected equipment.
These applications have been running since the digital transformation initiative began four years ago. “So, there is a huge amount of data, which is a gold mine,” said Mr. Venkatramanan. Armed with this data, they wanted to optimize operations and organizational efficiency. And that is when artificial intelligence made an entry.
Applications and Use Cases of Artificial Intelligence
The company’s Alchemy Construction Intelligence Platform is an analytics engine to identify patterns and generate actionable intelligence.
Alchemy - AI in Vison:
- To keep track of workers for safety or other reasons, on a road construction site/tunnel, a Workman Management Application on the frontline supervisor’s mobile phone is used.
- For the pipe laying projects, about 20,000-30,000 pipe welds are done every month, and these have to be inspected for quality before client handover. This used to be a time-consuming process with many experts involved, but AI has simplified this process. The Quality Inspections Group AI captures the weld image through a contraption in the pipe, and this is sent to the cloud for analysis. The accuracy rate improved after the number of welds increased, annotated, and feedback was given to the system, and now the accuracy is an astounding 97 percent.
- Materials constitute 70 percent of construction projects, so any pilferage leads to severe losses. Image recognition is useful to check what material is brought to the site, weight of the loaded material, vehicle number plate, etc. As soon as a vehicle enters the site, it comes to an automated weighbridge with cameras to take pictures of the vehicle and the material inside it. This adds up to 15,000 transactions across the global project sites, which are automatically analyzed and verified, resulting in savings on material costs.
- CCTV cameras are located at appropriate locations to capture unsafe acts, such as a workman not wearing PPE, or using a mobile phone in a hazardous zone and are automatically flagged and alerts are raised. These are also useful to warn about unsafe conditions, such as open pits, barricades, etc. Considering the current pandemic and the need for social distancing, alerts are raised if the workers come too close.
Alchemy – AI for Speech Recognition and Comprehension
As most of the projects are distributed across multiple regions, it is tedious for the personnel to read and understand the quality and safety manuals, policies and so on. This can be done using speech; the user can ask the query to a bot in the mobile, and the bot responds to it from the SLP (speech language processing) document side. This bot is connected to all the different digital systems, the safety and quality platforms, and a system for digital stores for looking at material stock availability, material reorder status, minimum availability etc. For personnel on the move or handling many projects, the required information is just a question away. There’s also a screen that accompanies the speech, and this provides the contextual trend and information.
A lot of work is being done in the area of comprehension using natural language understanding (NLU). The company works on many tender documents, and some of the business units work on 20-30 tenders (each is about 1,000-2,000 pages) every week. An AI model has been developed to turn out a three-page summary in three minutes listing the risky processes. This is a tremendous improvement in how fast tenders can be assessed and awarded to contractors.
In conclusion, Mr. Venkatramanan said that the company’s focus is on predictive maintenance and managing resources intelligently by using AI. Moving ahead, they would like to be able to predict project delays, but this will involve analyzing thousands of parameters that impact the project.