BY KAVITA NAGPAL
Taking forward from our previous blog on ‘Artificial Intelligence (AI) in the Indian Defence Sector’ https://thedefencespace.com/2023/06/28/part-6-of-the-series-artificial-intelligence-ai-in-the-indian-defence-sector/ in this blog post we will be covering few more AI-Technologies under – Manufacturing and Maintenance, Operational Data Analytics and Perimeter Security Systems.
All of these mentioned AI-Technologies are being indigenously developed or developing by the India’s Defence Industrial Base (DIB).
Read on to know about these in detail. Also, refer image.
I. Manufacturing and Maintenance
1.Artificial Intelligence Based Predictive Maintenance Suit
Ships and other platforms are typically complex systems containing a large number of subsystems and components, operating in a harsh environment. These systems are presently maintained by performing a scheduled maintenance (hourly, daily, monthly and yearly) or, reactive maintenance.
This is not only a costly process, but it also limits the availability of the system. ‘Reduced downtime, high availability with reliability’ of the platforms is one of the prime requirements for the Armed Forces.
Predictive maintenance, also known as condition-based maintenance helps address this situation, in an optimum manner.
The AI-based Predictive Maintenance (PdM) Suite uses data from various sources like historical maintenance records, sensor data from machines, and weather data to determine when a system/equipment/machine will require maintenance.
Leveraging real-time asset data plus historical data, the AI system can predict potential failures in real-time thereby decreasing machine downtime.
The key benefits of AI-based PdM Suite are:
- Optimisation of available resources
- Reduction in equipment downtime
- Adaptive maintenance routine
- Early detection of potential faults
- Fleet level consolidated visibility
- Comprehensive Spare Part Management
2. Predictive Maintenance for Gun Fire Control Systems
Where AI scores is being able to predict breakdowns, downtime and speed up repairs or replacement if needed. Both time and speed are extremely crucial especially if it comes to gun control systems being operational at all times.
Lynx-U2 is a Gun Fire Control System which has been successfully deployed in more than 35 systems as the Ship’s Last Line Self Defence across Eastern and Western Naval Commands.It interfaces with various on-board sensors and Long/Short Range Guns.
The AI/ML-enabled, real-time predictive system has been developed and deployed to predict better maintenance of PCBs by using the concept of detect anomaly /failures/early warning and asset conditions parameters.
The software solution is useful in a wide range of systems, where the system’s health conditions need to be anticipated much prior to avoid severe damage / malfunctions.
Advantages of using Predictive Maintenance:
- Maintenance of system based on predictive rather than scheduled maintenance.
- Monitoring all critical sensor data of assets.
- Avoiding failure of LRU’s w.r.t to voltage spikes and temperature issues.
- Proper planning and replacement of spares.
- Minimise downtime by continuous monitoring and taking proper actions to prevent failures
3. AI Based Predictive Maintenance of Delhi Metro Rail Equipment
A smart (AI) system not only works smart but prevents downtime on its own.
In a system with many components working together and influencing each other’s lifetimes, it becomes a challenge to find the right moment when maintenance should be performed so that components are not prematurely replaced and the whole system stays functioning reliably. To achieve this outcome, a comprehensive solution is being developed at BEL.
Using Machine Learning (Regression and Classification algorithms) algorithms have been developed that are capable of anomaly detection and survival analysis that would help in achieving predictive maintenance schedules / metrics.
This product is unique for its functionality of predicting equipment failure based on equipment health, historical patterns of occurrence of failures and periodic maintenance schedules. These predictions are utilised for predictive maintenance of metro equipment, in order to enhance their operational duration. This product is being utilised for the Delhi Metro Rail Corporation (DMRC). Using this solution, downtime of equipment shall be reduced, which will further ensure user convenience.
Additionally, maintenance cost is reduced by proactively maintaining the module before occurrence of any major failure.
4. Alloy Development through Artificial Intelligence
The application of intelligent data analysis methods in materials science research can greatly assist in optimizing the virtual screening space for new materials and speed up the process manifold as compared to classical ways of finding/improving new materials.
The goal of the project is to design a framework for new alloy design using Artificial Intelligence. The preliminary studies were conducted around the composition of H13 tool steel. The focus of the targeted application is to achieve low thermal expansion coefficient with considerable strength of the steel using Machine Learning.
In future, this framework can be used for any other alloy systems like superalloys, titanium alloys, with respect to specific applications.
5. Predictive Maintenance of Mining Equipment Through Data Analytics and Telematics Enabled System
This AI-based system uses modern, analytical techniques to decrease cost of maintenance and reduce downtime. It does so by early identification of imminent equipment failure.
It employs Telematics, the remote monitoring of equipment and vehicles using a “black box” that collects, stores data and sends it via cellular or satellite to a cloud server and, ultimately, software accessed by the end user.
Telematic Solutions commonly provide detailed machine data, including parameters like engine oil, transmission and brake temperature, engine RPM, speed, tire pressure, fuel consumption and emission levels to name a few.
This data can be analyzed and combined with historical data and deep industry knowledge to predict an upcoming failure so that planned maintenance can be performed.
This avoids the cost and delays associated with a random failure and helps operators determine when maintenance is needed rather than schedule them based on time or miles spent.
Key features:
- External Device Interface Over Can 2.0, J1939, Obd2
- Immobilizer activation By SMS
- Over-the-air Firmware and Data Update (Two Way Communication)
- Easy-to-understand And Customisable Reports
- Location Report and Route Playback
- Alarms and errors Information
- AI based Predictive maintenance algorithms.
- Customers can get their own machines records automatically with telematics.
- Realize the operator’s working status.
- Proactively addressing problems smoother operations
- Decreasing the risk of unexpected events happier customers
- Detecting issues before failure= lower costs, improved safety
- Increasing uptime= more opportunity to generate revenue
Unique Features:
- The AI software analyses real-time information along with historical data to predict whether the equipment will fail.
- AI algorithms easily spot any changes, which may indicate failure, long before it could affect the vehicle’s performance.
- Predictive analytics for early warning.
- Artificial intelligence takes the guesswork out of preventative maintenance. Measurement of part depreciation saves costs and time.
6. Condition Monitoring System for Shipboard Equipment (Main Engine)
GSL and Infosys have jointly developed a Condition Monitoring System (CMS) for Main Engines On-Board OPV class ships using Artificial Intelligence (AI).
The system identifies anomalies and predicts the time of likely failure enabling predictive maintenance of main engines installed on board.
The system brings together the best of engineering and data science technologies to solve the complex problem of engine health prediction.
Unique Features:
The Condition Based Monitoring system for Main Engine is a predictive maintenance tool for health monitoring and failure prediction of Marine Engines which is unique in providing following capabilities to ship’s crew:
- Assess and monitor Engine Health continuously
- Predict failure
- Reduce maintenance and operational costs
- Improve inventory carrying levels
- Enhance performance and safety of ship
7. AI-Enabled Evaluation of Welding Defects in X-rays of NDT
This AI enabled intelligent system has replaced the manual evaluation/inspection system of weld defects in X-rays of NDT.
This intelligent system encompasses AI / ML technologies for carrying out the inspection of weld defects from X-rays of NDT. The software has an accuracy of 90%.
Unique Features
The use of this AI software would decrease the dependency on Human expert and eventually curtail the requirement of Human interface for evaluation of RT Films derived from X-rays from weld joints.
II. Operational Data Analytics
1.AI-Enabled Fake News Detector as Part of Social Media Analytics
Fake news is deliberate and intentional misinformation. It consists of fabricated news stories, with the intent to mislead.
To utilise the social media content in a constructive way, BEL has developed advanced computational models using Machine Learning (ML), Deep Learning and Natural Language Processing (NLP).
In order to identify fake news / content (text, image and videos) sourced from various social media and sound the alarm to users to curtail its spread before fake content causes irreparable damage to individuals / society.
By means of Fake News and Fake Media Detection, the product delivers the following capabilities to meet community expectations:
- Analysis and detection of fake or original news content.
- Fake News Detection with Title of News, News URLs and News Content with Confidence level indication.
- Analysis and detection of fake or original media content.
- Highlighting and displaying the tampered sections of image, if image is tampered
2. Operational Data Analytics for Naval Platform
A large amount of data, the ability to process it and gain valuable insights from it is critical.
The “Platform Sensor Grid” stores a large amount of sensor data from all the naval ships. It uses big data analytics techniques with advanced statistical and machine learning algorithms.
This provides an effective and efficient target analysis for various mission critical tasks. It finally predicts sensor performance and recommends best available naval ships for mission categorisation.
The sensor grid comes up with a dashboard visualisation that enables the user to effectively comprehend the vast amount of data at a glance.
The data insights provided by the Platform Sensor Grid aids the naval decision-makers in an effective and accelerated decision-making regarding mission categorization.
3. AI Enabled Automatic Information Extraction and Synthesis
In today’s digital world, the Internet, smart devices, IoT has caused an explosion of data. Human internet activity has become a tremendous source of potential information.
This information from diverse sources would be impossible for humans to manually process and understand.
Hence, BEL developed a comprehensive AI enabled Automatic Information Extraction and Synthesis.
Functionalities:
- Machine Comprehension
- Video Summarization
- Fake image detection
- Speaker Diarization
Some of the techniques that are employed for automated information extraction include Naïve Bayes classifier, Support Vector Machine, Decision Trees, Neural Network, etc.
Features:
- Automated Information extraction from a block of text either from text file or pdf. The system uses BERT NLP model to learn the context of the text from and provides the answer for the question within the context of the paragraph.
- The system summarises a video with the key frames, so that the user need not see the whole video.
- System is capable of finding the fake image
- Speaker Diarization is implemented, that indicates which speaker spoke at what interval of time
Potential users:
- All Law Enforcement Agencies
- Military/ Paramilitary / police forces
- Users of C4I / C5I applications
- Intelligence collection and Analysis agencies.
III. Perimeter Security Systems
1. Sarvatra Pehchaan – AI Based Intrusion Detection & Integrated Command Station
Intrusion and detection is a critical component of any defence strategy. Earlier, multiple Sensors deployed on the AIOS lacked common operating system and the feed was not integrated on a single dash board.
Video footage needed to be monitored manually. An AI integration system to monitor & process multiple feeds was sorely needed to improve efficiency and minimise errors.
SarvatraPechaan considers a fusion of sensory feeds from multiple imaging systems and sensors on a single dashboard.
AI-based video analytics techniques are used for anomaly detection, allowing rapid intervention.
Features:
- Integrated AI based Intrusion Detection System and Integrated Command Station.
- Fusion of multiple sensor feeds in single dashboard.
- Automated AI based anomaly detection.
- Indigenous Software.
- User interface application.
- Scalable for future upgrades.
- Time stamped Intrusion Detection and Identification.
2. AI-Enabled Forensic Search for Videos
AI-enabled technology is making a big contribution in the areas of surveillance. It’s a well-known fact that all surveillance cameras record video 24×7.
But to view any particular or isolated event of interest, the operator needs to view the complete data which can be an exhaustive and time-consuming exercise.
AI-enabled forensic search now makes it possible to view the events in the entire time interval according to a set of filters.
To perform a selective search, different search criteria options are provided to the user via forensic search user interface.
Using this technology, camera-stream recorded videos are processed using AI techniques to identify the objects in the scene. This process runs continuously in the background.
All the identified objects’ metadata and their properties are stored in the database for further analysis.
The link to the corresponding video file is also stored in the database.
Key features:
- Identification of intrusions hotspots,
- Visualise intrusion timelines
- Perform precise search for objects and activities
- One-click to export shortened summary video
3. AI-Enabled Gesture Recognition
Being able to positively identify enemy threats through movement detection of human forms is an indispensable tool in the hands of the armed forces.
The Artificial Intelligence (AI) enabled gesture identification system uses deep learning technology in identifying various human gestures like human walking with or without a gun, human crawling with or without a gun and human crouching with or without a gun.
The system can be easily integrated on a network of IP enabled cameras.
Applications:
- a) Securing the border with neighbouring countries which has to be monitored continuously for illegal and dangerous activities like terrorist intrusion, smugglers or illegal immigrants.
- b) Automatic identification of human gestures near the perimeter of defense establishments. Monitoring the activities effectively which allow the identification of dangerous threats for taking preventive steps.
4. Audio Doppler Based Object Classification
Artificial Intelligence makes it possible to identify and classify the nature of moving targets.
Automatic Doppler based object classification system determines the nature of a target moving in the radar’s field of view using its Doppler return signal.
The five classes considered for ground target recognition are: Single Moving Pedestrian, Group of Moving Pedestrians/Animals, Moving Wheeled Vehicle, Moving Tracked Vehicle and Unclassified.
The Automatic Doppler based system aims at replacing the human operator, thereby avoiding a very tedious task for the operator and in turn making the system independent of operator fatigue and health level.
This product is useful in classifying the behaviour of ground intrusions using radar, which can alert the ground forces.
