The Economics of Digital Twins and Predictive Maintenance

What if you could have a model of a valuable asset, such as a piece of equipment, to use to determine when, how often, and what type of maintenance and repairs would be needed on the asset?  This can be accomplished through a digital twin of a system, allowing tests to be run without jeopardizing the safety and integrity of the real system.  For a website, a digital twin can reveal how much Internet traffic would overwhelm the site.  A digital twin can also reveal what type of user actions would result in errors, affecting customer satisfaction.

Digital twins can also be designed to mimic physical capital goods, such as factory equipment or vehicles.  Typically, this involves networks of sensors placed in one piece of equipment, like an airliner, that feed data in real time to computers that crunch that data and look for anomalies or signs of problems.  This helps the computer system learn the physical strengths and limits of the piece of capital.  After enough data is collected to create a baseline, technicians can run digital tests to see when failures would likely occur.

Economics of Digital Twins

Reduces Underutilization of Capital

Digital twins can save large companies billions of dollars by making capital maintenance and replacement more streamlined.  Knowing when and how to repair a piece of expensive equipment can save money in the long run.  Without the use of digital twins to allow for predictive maintenance, companies may over-repair and replace capital too quickly, losing money due to underutilized capacity.  Replacing an eighteen-wheeler truck a year before it is necessary could cost the company thousands of dollars; they could have delayed the higher payments on the new vehicle.

Reduces Failure (Overutilization) of Capital

AI can be used to simulate thousands of scenarios to test the limits of physical equipment, potentially determining failure points and revealing fixes to those failures.  Although this would require companies to spend more money to bolster the strength of their capital goods and infrastructure, this additional up-front spending can be seen as an investment in reducing the need for repair, replacement, and potential legal fees from equipment failure affecting consumers.  Thus, if properly utilized, digital twins and predictive maintenance can save companies lots of money.

Increases Efficiency Through Planned Maintenance

Knowing how much use and stress a piece of capital can take before it needs maintenance, especially when the digital twin can accurately track and predict rates of use, allow the company to more efficiently plan for maintenance periods.  This can save time and money through fewer unexpected repairs due to damage or failure.  For example, if the digital twin reveals that the tractor engine needs an overhaul every six months of heavy use, the farmer can plan to remove that tractor from the rotation at the correct time and have it delivered for overhaul.  Without this predictive maintenance, the tractor may develop unexpected problems during harvest time and cost more time and resources to repair on an emergency basis.

Broader Implications of Digital Twins, Predictive Maintenance, and AI

Increased Productivity

Overall, digital twins and predictive maintenance, especially when coupled with AI, significantly increase productivity by allowing for fuller use of capital goods.  Companies spend less resources on unneeded repairs or emergency repairs.  By reducing costs and increasing output, companies can pass these productivity gains along to consumers in the form of lower prices.  Thus, a broader implication of digital twins is increased consumer surplus and producer surplus, meaning a gain in total surplus.  How much this occurs may depend on the market model and level of competition; monopolies and concentrated oligopolies may keep most of the productivity gains as profit.

Insurance Industry

Insurers may begin requiring companies to use digital twins and predictive maintenance software, or charge those companies that do not do so higher monthly premiums.  This could disadvantage small companies, which may not easily be able to afford such software for their few capital goods.  A software subscription for predictive maintenance on an eighteen-wheeler truck or a tractor may be cost-effective for a large agricultural firm that owns dozens of each, but prohibitively expensive for individual farmers.  

Copyright Fights

Eventually, AI will likely allow small businesses to access the benefits of digital twins and predictive maintenance, as the digital mapping and testing of common models of equipment will become part of the Internet.  This could result in legal battles as companies that paid for such digital services demand that they not be given away for free to smaller firms.  Similar to downloading digital entertainment without paying for it, will small businesses risk lawsuits for downloading predictive maintenance schedules for their equipment?  This could become part of the right to repair movement, where purchasers of equipment demand increased legal rights to repair and modify such equipment without paying additional fees to the producer.  If you buy a tractor, should you also automatically have the rights to AI predictive maintenance guides for said tractor?

Liability for Predictive Maintenance Errors?

Digital twins and predictive maintenance AI programs may be highly accurate, but is there liability in the event of error?  For example, if a company pays a tech firm for predictive maintenance services for its array of capital goods and the maintenance guide turns out to be incorrect, does the tech firm have liability?  What is the acceptable degree of error, for both underutilization and overutilization, of capital goods when an outside firm is keeping digital watch?  How many variables, ranging from atmospheric to road conditions, should digital models be required to manage?

There will almost certainly be legal battles over this, with the tech firms arguing that unpredictable variables, perhaps including severe weather patterns due to climate change, causing unexpected equipment failures that should be exempt from modeling.  To limit the chance of liability, tech firms may adjust their predictive maintenance models to overpredict the maintenance needed or schedule for capital replacement, preventing equipment failure but costing the client companies extra money.  Therefore, fear of expensive liability may reduce the effectiveness of digital twins and predictive maintenance as tech companies tailor their models to be very safe, resulting in unneeded maintenance.