The manufacturing world faces a significant challenge: a staggering 82% of companies experience unplanned downtime due to human error or inadequate knowledge of equipment conditions. Depending on the industry, downtime costs range from $39,000 to over $2 million per hour. This underscores the uphill battle that process engineers confront daily: Tasked with managing intricate production processes, they grapple with elusive disruptions, such as breaks, jams, or overcooking. These disruptions confound engineers and remain difficult to predict and prevent. When the root causes lie beyond the engineers' comprehension, the consequences can be dire, contributing to the alarmingly high rate of unplanned downtime.
However, sufficient information often is hidden beneath telemetry and other data. While most companies still struggle to use their process data for optimizing production processes, AI can produce valuable insights to ensure timely prediction and effective prevention of process disruptions.
This article illustrates how process engineers and production managers can substantially benefit from leveraging process data with AI and highlights:
In the end, you will understand how AI can prevent process disruptions in manufacturing and positively influence overall equipment effectiveness (OEE).
Data generated by a vast array of machines presents a significant challenge for human analysis. It becomes increasingly difficult for process engineers to identify interdependencies between events and uncover the root causes of disruptions.
As a result, they often resort to superficial one-time fixes that fail to address the underlying issues, leaving the core problem unresolved. Additionally, even when attempting to analyze the data, humans are prone to overlooking crucial details, causalities, and interdependencies.
In this complex environment, finding the relevant signals is akin to searching for a needle in a haystack: it's difficult to locate, but once found, the correlation becomes evident and easy to verify. The challenge lies in determining which signals are essential and how their unusual behavior might be interconnected to paint a complete picture of the developing problem.
Where human engineers may fail, the aivis® engine, a powerful industrial AI engine designed to work with large amounts of untidy raw data, can consume all available historical data from the shop floor to autonomously determine which signals are crucial for understanding and predicting the disruption.
The AI engine aivis® was developed by Vernaio and consists of six different engines that can be used for different use cases. Among these engines is the aivis® State Detection engine, which analyzes telemetry and real-time data to identify root causes and proactively prevent disruptions in production processes.
By analyzing average behavior of these signals right before a disruption occurs, aivis® helps to find and understand the root causes of the disruption.
Typically, process disruptions can manifest themselves in multiple problematic signals. These signals form a distinctive pattern or fingerprint that reflects the unhealthy behavior of the process right before the disruption occurs. Each signal in the fingerprint is relevant and conveys a story of the transition from a healthy to an unhealthy state. The fingerprint helps identify the hidden contributing factors that can influence the root cause of the disruption.
Our industrial AI engine aivis® can even identify several fingerprints for a single disruption scenario and find the underlying root causes that may not be self-evident from the available data. Identifying a fingerprint is a quick and easy process that doesn't require any specialized knowledge from an engineer, as aivis® takes care of the heavy lifting automatically. This method requires little initial investment and yields results within minutes or hours, providing value within a few days to weeks. Furthermore, the AI can generate straightforward visuals, making it easy for the engineers to understand the nature of the critical incident.
This enables the engineer to interpret the fingerprints and to decide if either a one-time fix or a series of ongoing countermeasures are suitable to address the issue.
It is crucial to know when to apply ongoing countermeasures, since they can minimize the impact of the disruption and prevent it from happening again in the future. It is generally worth applying ongoing countermeasures when the disruption is recurring, and its root cause is tied to an inherent aspect of the production process that cannot be permanently resolved. Importantly, the expenses and work required to apply ongoing countermeasures are typically lower than the potential losses that could result from production disruptions.
To prevent these disruptions and improve key KPIs, fingerprint recognition is useful. aivis® can proactively monitor and evaluate the risk levels in real time. When the risk associated with a specific fingerprint surpasses a predefined threshold, the AI triggers a warning, indicating that it's time to apply the ongoing countermeasure. This warning enables the process engineer to evaluate the critical process situation and to implement a series of suitable countermeasures.
On top of that, aivis® automates the implementation of countermeasures. A ticketing system pushes the prepared countermeasure to the operator as soon as a disruption warning is issued. This enables the operator to apply effective countermeasures in time to reduce the risk of impending disruptions before the actual disruption occurs, leading to an improvement in KPIs such as OEE.
One of the key performance indicators in manufacturing is Overall Equipment Effectiveness (OEE). OEE measures the efficiency of a manufacturing process by considering three main factors: machine availability, production performance, and output quality. Implementation of an AI solution results in an improvement of OEE by positively influencing all these factors:
By utilizing aivis®’ State Detection functionality, process engineers can significantly lower the occurrences of disruptions in any process and improve crucial KPIs, such as OEE, downtime, waste, energy and CO2 emissions. With the ability to analyze vast amounts of industrial raw data, AI-driven solutions such as the aivis® engine can identify the root causes of disruptions through problematic signal behavior and the disruption fingerprint (generic pattern of unhealthy behavior).
This data-driven approach enables a more complete understanding of the problem and the development of effective countermeasures, whether a one-time fix or an ongoing procedure at the right time. By harnessing the power of AI, manufacturers can achieve their OEE goals and reduce the costs associated with unplanned downtime, rework, and returns.
