Application of AI in Predictive Maintenance of Vehicles
This article was published as a part of the Data Science Blogathon.
With the rapid expansion of new and advanced automotive features, the traditional methods of fault detection and control are becoming less effective, necessitating the development of smart solutions. New cars are laced with various sensors, instruments, and cameras that capture a wealth of data associated with the car and its parts. Why not leverage this multi-dimensional data along with past service records in conjunction with AI to provide insights into every aspect of vehicle operation and the timely remote diagnosis of faulty component(s)? Well, this can be a game-changer! It can not only suggest which component/system requires maintenance or replacement but recommend and enable control measures for various technical issues.
The ability to gather and analyze data about assets enables an organization to transition from corrective to predictive maintenance. Predictive maintenance is also a key goal for fleet operators, particularly transportation and logistics companies, where downtime is exceedingly expensive. Several major manufacturers have committed to leveraging the gathered data to inform about predictive maintenance of vehicles, allowing remote diagnosis of most vehicle problems before they arrive at the service bar. With the host of potential benefits of predictive maintenance for individual vehicle owners, fleet operators, and manufacturers, predictive maintenance is expected to be increasingly adopted in the automotive industry.
In this post, we will take a look at what predictive maintenance is, why it’s necessary, and some of the solutions provided by the automotive industry.
What is Predictive Maintenance?
Predictive maintenance is the data-driven approach to predicting the failure of operational equipment and implementing preventative maintenance to avoid unplanned downtime. It leverages a wealth of data gathered from in-vehicle sensors, past service records, and machine learning techniques to analyze various equipment conditions that could potentially lead to a potential system/equipment failure. AI system then notifies the user and the automobile manufacturer/maintainer that a certain component or system requires maintenance or replacement.
This approach minimizes the cost of unscheduled maintenance while maximizing the component’s lifespan, allowing us to get more value out of a part.
Furthermore, AI-powered quality control systems can be enabled to detect possible defects in components before they get installed.
Fig. 1 Diagram illustrating the crux of Predictive maintenance (Source: Intuceo)
Why is Predictive Maintenance important?
Predictive maintenance has various benefits, including spare part manufacturing optimization, stock management, lifecycle optimization, recycling management, etc. Predictive maintenance is also a key goal for fleet operators, particularly transportation and logistics companies for whom downtime is exceedingly expensive.
It has the potential to benefit car owners, dealers, and OEMs in the following ways:
Benefits to Car owners:
- Real-time alerts and early warnings cut down the cost of unscheduled maintenance while maximizing the component’s lifespan, allowing us to obtain more value out of a component
- Reduces breakdown scenarios through proactive communication by the dealer
Benefits to Dealers:
- Creates a personalized relationship with car owners through improved communication and trust
- Improve customer experience by real-time monitoring of potential faults
- Saves money on inventory and labor
- Boosts revenue
Benefits of OEMs:
- Boosts revenue from aftermarket sales and original parts/spares
- Reduces product recalls
Current Solutions provided by Automotive Industry
1. HMG: Sound-based Fault Diagnosis and Predictive Maintenance
The Engine NVH Research Lab at the Hyundai and Kia Motors Namyang R&D Center devised a novel solution that allows AI-learning of automotive sounds enabling the AI to detect faulty components.
The raw sound is extracted from various parts of fully functional and fault-induced engines to train the model. These gathered sounds are processed, analyzed, and categorized and become part of a growing database from which the model could learn.
Once the model has learned various nuances of sounds linked with vehicle components (faulty and non-faulty), it can be used to recognize similar patterns of sounds and provide inferred diagnostics. Based on that analysis of sounds, the model suggests the most likely cause of the abnormal sound.
However, the accuracy of sound analysis and inferred diagnostics depends on the quality of data the model is trained on. Currently, such a system’s claimed accuracy is around 88%.
Figure 2. Diagram depicting the sound-based fault diagnosis method for vehicle engine (Source: Hyundai)
2. Infosys: Vehicle Maintenance Workbench
When maintenance costs are high, it is tempting to delay scheduled maintenance, but doing so is often more expensive and increases downtime. Fleet maintenance managers use preventive maintenance or OEM guidelines. Though it avoids unscheduled repair, the high cost of Preventive maintenance (PM) and undesired vehicle downtime remains. Infosys devised a “Vehicle Maintenance Workbench (VMW)” to circumvent this. It is an integrated platform for fleet maintenance efficiency and safety based on AI and ML optimization algorithms.
The Infosys Vehicle Maintenance Workbench can predict the failure in advance. It helps maintenance managers in auto-scheduling vehicle preventive maintenance in their respective garages. Its built-in algorithm can schedule 5000+ maintenance jobs. In a few seconds, it shows the vehicles’ average garage utilization and downtime hours.
The failure prediction tab of Infosys VMW shows the complete status of the fleet. It indicates which vehicles require immediate service Vs the vehicle which are still safe to drive. It gives alerts on engine and transmission failures, brakes, battery life, tires, axles, and steering issues as well. Based on these alerts, maintenance activity can be undertaken.
Once the vehicle arrives at the garage for inspection based on a predictive maintenance alert, it undergoes thorough checks based on the alert types, and necessary repair works are carried out. In case of no issues with the vehicle, the technician updates the planner overrides the flag as “no” and with the relevant comments in the remarks section, and saves the detail. The machine learning model will further leverage these details to adjust and improve the accuracy of the future failure prediction.
- Improves vehicle availability by 10+ %
- Increases the life of the vehicle and parts by 15%
- Reduces TCO by more than 20%
Watch here| https://www.youtube.com/watch?v=A6ECsJhxYbk
3. Intuceo: Predictive Maintenance Solutions
Figure 3. Diagram depicting the method overview. (Source: Intuceo)
Intuceo leveraged a wealth of in-vehicle sensor data and machine learning algorithms to provide predictive maintenance solutions for OEMs and dealers. This solution claims to transform and analyze the in-vehicle sensor data (cars, trucks, EVs), allowing the customers to deploy this solution in the automotive manufacturing industry to reduce downtime and costs, as illustrated in Figure 3.
4. Questar: AI-based Predictive Analytics to keep fleets on the road
Questar bills itself as a “predictive vehicle health company” and provides a Vehicle Health Management (VHM) Platform that harnesses gathered in-vehicle data and AI techniques to generate early warning of potential malfunctions in vehicles.
This company claims that its VHM platform detected serious malfunctions in 10% of the buses in a fleet operated by Israel’s Kavim Public Transportation company in a pilot program.
Furthermore, its field data shows that by employing the VHM solution, fleet operators can realize a 30% reduction in costs on spare parts, a 10% reduction in fuel consumption, a 20% reduction in accidents, and up to 75% reduction in unnecessary downtime. Real-time health monitoring optimizes emission filtration and saves fuel, making fleets more environmentally sustainable.
5. IBM: Connected Vehicle Predictive Maintenance Solution
IBM developed a solution for monitoring connected vehicles for predictive analysis. Connected vehicle solutions typically include the vehicle, but with mobility as a service (MaaS), wherein the user can be in the vehicle as the driver or passenger or outside the vehicle.
Passengers can communicate with the vehicle via the user interface (human-machine interface) HMI or by using an external application or device.
Onboard sensors and cameras continuously monitor the health of the components in the connected car. Furthermore, the AI services analyze and act upon the data from the sensors and cameras by providing recommendations to the driver.
6. BMW: Predictive Maintenance Control Measures by BMW
BMW Group has devised cloud-based predictive maintenance solutions to increase its efficiency and sustainability. This technology leverages sensors to monitor the current status/health of the components and data analytics and other machine learning algorithms to forecast failures before they occur. Based on the detected status, the predictive maintenance system suggests when to replace components as a precautionary measure to avoid any unnecessary downtimes or unseen failures inflicted by the breakdown of component(s). This will reduce the maintenance cost and provide the company with fair insights into the design or functionality-related error.
Predictive maintenance is enabled on all production lines, allowing worn parts to be changed and faults to be identified before they cause operational issues. If stock levels of specific parts are too high in some locations and too low in others, or if specific transportation resources are under or over-resourced, all of this may be promptly adjusted.
7. Ford: Connected Vehicle Data Collaboration
Ford collaborated with CARUSO and HIGH MOBILITY to provide third-party businesses access to vehicle‑generated data safely and securely with drivers’ consent to create innovative, tailored products and services like usage-based insurance, predictive maintenance, and smart roadside recovery.
Notably, the vehicle owners retain control over their data, with the ability to control with whom and how the vehicle data is shared.
It does not apply to fleet customers.
AI-based predictive analytics is becoming a valuable technology across industries where predictive maintenance can save money and labor time, reducing overall stress levels.
To sum it up, in this blog post, we learned:
1. What is predictive maintenance?
2. Non-exhaustive list of current solutions provided by the automotive industry to leverage in-vehicle multi-dimensional data in conjunction with AI techniques to provide insights into every aspect of vehicle operation and the timely remote diagnosis of faulty component(s).
3. Benefits of predictive maintenance to car owners, dealers, OEMs, and the environment.
The media shown in this article is not owned by Analytics Vidhya and is used at the Author’s discretion.