The automotive sector has seen a major shift since the last decade, slowly but steadily the industry has embraced the disruptive digital technologies that are sweeping across multiple industries. Internet Things (IoT), one of the most prominent such technologies, has become ubiquitous in the supply chain, manufacturing, energy, electronics etc., and is changing the competition arena for both customer and commercial vehicles. Connected Vehicle Systems, which lies on the foundation of IoT, with a humble start has penetrated into nearly every, if not all, automotive OEMs. Subsequent associated use cases eg. Fleet Management, Autonomous Vehicles, and In-Vehicle Infotainment are also maturing gradually. The Automotive Maintenance and Repair area while being a direct benefactor of the IoT-based connected vehicle systems, has still remained elusive, very few implemented, less commercialised, or explored. There are multiple use cases however primarily the focus is on six key use cases for automotive repair being enabled by IoT.
1. Automated Damage Assessment.
2. Using IoT data to predict the maintenance and repair of a vehicle.
3. Dynamic visibility into repair and estimated time for completion.
4. Using IoT Connectivity between vehicles and workshops to estimate vehicle repair volumes.
5. Using IoT captured data and predictive maintenance to estimate parts availability and ordering.
6. Automated Car maintenance using Vehicle Diagnostics features.
We will examine the first two use cases in detail in this paper and discuss their feasibility, benefits, and probable solutions.
A. Automated Damage Assessment
IoT-enabled computer vision coupled with deep learning techniques is used to accurately classify vehicle damage to facilitate claims triage by training convolution neural networks. The rapidly growing automotive industry highly backs the equally fast-growing auto insurance market. Till recently this industry has been solely based on traditional ways to make repair claims. In case of an unfortunate accident, the claims for the car damage need to be filed manually through an inspector who is required to physically analyse the vehicles to assess the damage and obtain a cost estimate. In such a situation, there is also the possibility of inaccurate settlements due to human errors. Automating such a process with the help of IoT and remote usage would make the process a lot more convenient for both sides of the damage, increasing the productivity of the insurance carrier and the satisfaction of the customer.
To automate such a system, the easiest method would be to build a Convolution Neural Network model capable of accepting images that are captured through IoT-enabled sensors and cameras from the user and determining the location and severity of the damage. The model is required to pass through multiple checks that would first ensure that the given image is that of a car and then ensure that it is in fact damaged. These are the gate checks before the analysis begins. Once all the gate checks have been validated, the damage check will commence. The model will predict the location of the damage as in front, side, or rear, and the severity of such damage as minor, moderate, or severe.
Typical adoption/implementation of Automated Car damage assessment takes the standard MVP route. Once set it is then scaled.
MVP Process Flow:
a. Car Scan – Customer/Inspector scans through vehicle scan system through cameras.
Cameras capture all relevant parts of the car especially the damaged ones in detail. Images are pre-processed in the vehicle scan system.
b. Damage Prediction – Damages are assessed and predicted. An apt algorithm is selected. Images are sent to the central damage prediction system.
c. Review – Process experts review predictions in the web app. Experts provide feedback regarding positions and classes of predicted damages in the image and correct those to provide input for retraining and model improvement.
d. Damage Report – Further systems request damage prediction information via REST API calls to generate the final damage report.
e. Operations – AN operations concept ensures that the MVP solution is scalable and adaptable for long-term success in the market.
The model accepts an input image from the user and processes it across 4 stages:
1. Confirms that the given image is of a car.
2. Verifies that the car is damaged.
3. Finds the location of damage as the front, rear, or side.
4. Determines severity of damage as minor, moderate, or severe.
The model can also further be improved to:
5. Obtain a cost estimate.
6. Send assessment to the insurance carrier.
7. Print documentation.
B. Using IoT Data to Predict the Maintenance and Repair of a Vehicle
IoT-based predictive maintenance allows you to monitor, maintain, and optimise assets for more effective utilisation, availability, and performance. Gain clearer visibility into how your equipment is performing via real-time monitoring. Forecast machine failure and identify parts that need replacement. IoT failure prediction involves collecting sensor values and running algorithms to anticipate impending failures. Core building blocks include identifying the features or factors contributing to vehicle failures. Then you configure a predictive model to train the model, followed by scoring the test data to check the reliability of the predictive model.
Advanced predictive methods will enable you to switch from scheduled preventive maintenance to predictive maintenance. Sensors mounted on IoT-enabled vehicles collect and transmit data on a continuous basis.
The first step is to identify any substantial shift in vehicle performance using time-series data generated by a single IoT sensor.
Once a change point is detected in a key operating parameter of the vehicle, it makes sense to follow it up with a test to predict if this recent shift will result in the failure of a piece of the part. This pattern is an end-to-end walk-through of a prediction methodology that utilises multivariate IoT data to predict failures. A binomial prediction algorithm using logistic regression is implemented for this purpose.
Fixing something before it breaks is more efficient and cost-effective than fixing it after it breaks. It helps to:
● Avoid downtime and improve productivity.
● Extend the life of assets and defer new purchases.
● Reduce the cost and complexity of repairs.
● Mitigate additional or related damage.
● Meet regulatory standards and compliance.
● Manage spare parts, materials, and inventory.
● And ultimately, boost the bottom line.
Example – Battery Maintenance
A Japanese automobile manufacturer uses IoT to model the behaviour of their welding process. It wanted to identify causal factors of failures and faults and find top predictors of equipment failure. The system delivers 90% prediction of faults with no false positives; 50% of the faults are predicted over 2 hours in advance. The company saved 1.5 hours per fault thanks to advanced prediction.
A major aircraft manufacturer is using IoT to maintain calibration of precision assembly tools and improve manufacturing quality. Data from shop floor tools along with equipment failure data is used in predictive quality analytics to generate models that identify tools likely to need servicing. Faulty tools are proactively removed from the shop floor to be maintained and recalibrated, leading to significant improvements in manufacturing quality. The solution has enabled a 100% payback within one year — avoiding millions of dollars of rework and months of production delays by preventing out-of-alignment tools from remaining in the aircraft production workflow.
Effective predictive maintenance harnesses the convergence of data from instrumentation and IoT with advanced analytics and AI technologies through digitalised systems. An A T Kearney survey in Industry Week where 558 companies that used computerised maintenance management systems exhibited an average of:
● Approx 30% increase in the productivity of maintenance.
● Approx 20% reduction in equipment downtime.
● Approx 20% savings in the cost of materials.
● Approx 18% decrease in inventory maintenance and repair.
14.5 months pay-back time
What does organisation need to do to harness the benefit of IoT?
To use these systems successfully, organisations need to:
As part of asset management, organisations must track, assess and manage the reliability of a wide range of physical and technological assets. Adding to this challenge are technology infrastructures running applications and data in silos. Integrating “siloed” systems improve visibility and efficiency in locating and communicating about potential failures.
IoT data such as weather-related information, RFID-enabled data, traffic information, and information from other devices and sources can augment and strengthen predictive maintenance. For example, weather can affect external equipment in farming or oil and gas production or highly sensitive instruments in fields like healthcare and biotechnology. IoT can also consolidate information from potentially millions of pieces of equipment. Elevator and escalator maker, KONE Corp. for example, remotely monitors and optimises the management of more than 1.1 mln elevators and escalators in buildings worldwide.
Analyse Quality Data
The ability to gather and analyse data about assets allows an organisation to move from corrective to predictive maintenance. Predictive analytics and AI technologies such as machine learning can be applied to volumes of operational data to give organisations a more detailed and accurate understanding of equipment performance.
The quality or integrity of the data being analysed is important, too. Without completed fields or validated data, analysis is not possible. Analysis of the health of data fields in critical areas such as asset registries, item inventory and work completion is essential to supporting reliable analytic reports.
Focus on Reliability and Efficiency
Building on the strengths of predictive analytics, reliability engineers can create statistically valid models of equipment life based on operational data and other factors. These models enable them to focus on critical risks that affect operational reliability and availability.
This capability also enables the development of a maintenance strategy that can improve efficiency: analysis may indicate current equipment maintenance schedules and practices are ideal and no changes need to be made. Or, prescribe maintenance sooner to avoid failure. Or, postpone maintenance to avoid unnecessary cost and effort.
Executive-Automotive, Aerospace & Defense, CIC India IBM Consulting