The world of maintenance is a unique dynamic matrix of proactive and reactive maintenance. The practices and procedures might vary based on the contracts, trends and different geographical regions. The common questions among them being ’What’, ‘When’, and ‘How.’ What is the goal of the maintenance plan, What kind of maintenance activities do we need, What are the type of assets we have, What is the residual life; When is the timeline, When is the expected failure, When is the preventative maintenance due; How are we going to perform it, How many manhours do we need, How will the team be mobilized- and the questions can be added more and more. These are the basics for any maintenance activities. They were in the past, they are in the present and they will be in the future. We created, performed and executed as per the existing standards or let’s say standards or practices that were created in the past; the main question arising ‘Are we doing enough in the Present to be ready for the Future?’
The ASCE’s grade report of United States’ Infrastructures fluctuated between D and D+ in this last decade. A question arising among the Policymakers, Stakeholders, Businesses and Professionals: How to tackle the burgeoning problem of ageing and failing infrastructures. While funds and relevant policies are the front runner in rehabilitation and mitigation of any infrastructure problems or failures and we can relate it to the questions we asked above ‘What’, ‘When’ and ‘How’ ; the integration of technology, especially, recent development of Artificial Intelligence (AI) and machine learning is the future that can help us achieve A+ in a short period of time.
Drones to collect data
Drones, sensors, and cameras powered with AI infused data analytical capabilities will be the monitoring tools making decisions for infrastructure maintenance for almost all the companies in the near future. These practices have already been started by many private companies and public organizations. One of the best examples of such practice is the use of drones for monitoring oil pipelines, oil rigs over the expensive helicopter inspections and risky human inspections, and state transportation organizations like CDOT using drones for slope analysis and Geo tracking of assets. Also, another use of drone is railroad inspection. While monitoring alone won’t yield any desired results; the fusion of video/image content analysis using machine learning models will provide greater insight on the condition of assets. The data scientists from Avitas Systems, a venture of General Electric (GE) built a convolutional neural network for image classification and generative adversarial neural networks to minimize the amount of work involved in labelling captured images of the inspected assets. It helps in identification of defects automatically comparing with variety of models. Avitas Systems can target specific points of inspection and develop paths to collect data in the form of images and video. These paths driven by 3D models, with the capability of repeatability are used by their drones, robotic crawlers, and autonomous underwater vehicles (AUV) from the same angles and locations of previous paths. The paths’ repeatability with same angles and locations means a wide variety of images is captured over time with the same accuracy. This creates a cycle of image collection of a targeted asset over a course of time. Then, the advanced image analytics can detect changes and measure exact defects on an asset, such as cracks, corrosions and failures (Source: ge.com). In another example, Fujitsu deep learning technology successfully estimates degree of internal damage to bridge infrastructure through the vibration data collected from the attached sensors. Such technology gives a power of decision making and creates a system of inspection and asset management which in turn helps in planning, maintaining and executing the life cycle of infrastructures.
Challenges of Artificial Intelligence
These are the few examples on how AI, sensors and robotics will play an integral role on infrastructure management of the future. This might not match our imaginations of Robotic and Artificial Intelligence from movies like star war’s ‘R2-D2’ or ‘C-3PO’ helping human counterparts in the reality; we are still in the inception phase of such systems which is constantly evolving. There are still some challenges in the integration of already developed AI system in our workflow of present like: Policies in state level and federal level, lack of skilled manpower for the operation of such systems, awareness and education among public ( we still hear news of people shooting down the drones for privacy), increasing overhead cost of cutting edge technology, and most importantly bringing changes in the already established systems. While it might be a few decades before the full integration of AI and Machine learning in our maintenance activities or planning; we are already in the path of achieving such feet from the small steps we are embracing. For instances, we are already using data analysis tools which is also a part of AI in our day to day activities. The data collected from simple ESRI applications on field through the use of location based asset tracking on determining the maintenance activities or conditional assessments of assets, development of dynamic decision making dashboards in ArcGIS through data collection, analysing the trends of historical data using software like ‘R’ and making decisions or predictions through correlation, use of ATS system by HR for potential candidate selection, use of chatbots in the websites and many others has already set us in the path of data analysis, decision making and artificial intelligence in its infancy.
We are already in the path of AI and machine learning. And there are challenges and hurdles along the way which can solved through good policy, management and awareness among public; the same question arises,’ Are we doing enough in the Present to be ready for the Future?’