Smart Trac: AI-Assisted for Predictive/Condition-Based Vibration Monitoring
How factories are deploying AI on production lines
Augury’s sensors used in PepsiCo factories have been trained on huge volumes of audio data, to be able to detect faults such as wearing on conveyor belts and bearings, while analysing machine vibrations. By also collecting information and insights into equipment health on the whole, such as identifying when a machine might fail again in future, the technology lets workers schedule maintenance in advance, and avoid having to react to machine errors as they occur.
Prof Brintrup, professor of digital manufacturing at the University of Cambridge’s Institute for Manufacturing, leads the Institute for Manufacturing’s Supply Chain AI Lab, which has developed its own predictive mechanism to identify where ingredients such as palm oil may have been used in a product, but disguised under a different name on its label. The lab’s recent research suggested that palm oil can go by 200 different names in the US - and these might not stand out to eco-conscious consumers.
🪱🤖 GE Develops Worm-Inspired Robot For On-Wing Engine Inspections
Resembling an inchworm, the Sensiworm (Soft ElectroNics Skin-Innervated Robotic Worm) uses untethered soft robotics technology to move easily through the nooks, crannies and curves of jet engine parts to detect defects and corrosion. The robot is also able to measure the thickness of an engine’s thermal barrier coatings.
Developed in partnership with SEMI Flex Tech, Binghamton University and UES, Inc., Sensiworm is controlled by an operator using a device that GE says is similar to a gaming controller and can be programmed to follow specific paths. “It has a sticky, suction-like bottom that enables it to climb and adhere to steep surfaces. Also, because the robot is very soft and compliant, it won’t harm any surfaces or cause any damage during an inspection,” says a spokesperson for GE.
According to GE, Sensiworm could reduce unnecessary engine removals and downtime, enabling faster turnarounds. Although Sensiworm is currently focused on engine inspections, Trivedi says the OEM is developing new capabilities that would enable the robot to execute repairs once it finds a defect.
Long–short-term memory encoder–decoder with regularized hidden dynamics for fault detection in industrial processes
The ability of recurrent neural networks (RNN) to model nonlinear dynamics of high dimensional process data has enabled data-driven RNN-based fault detection algorithms. Previous studies have focused on detecting faults by identifying the discrepancies in data distribution between the faulty and normal data, as reflected in prediction errors generated by RNN models. However, in industrial processes, variations in data distribution can also result from changes in normal control setpoints and compensatory control adjustments in response to disturbances, making it hard to differentiate between normal and faulty conditions. This paper proposes a fault detection method utilizing a long short-term memory (LSTM) encoder–decoder structure with regularized hidden dynamics and reversible instance normalization (RevIN) to compactly represent high-dimensional measurements for effective monitoring. During training, the hidden states of the model are regularized to form a low-dimensional latent space representation of the original multivariate time series data. As a result, the prediction errors of the latent states can be used to monitor the abnormal dynamic variations, while the reconstruction errors of the measured variables are used to monitor the abnormal static variations. Furthermore, the proposed indices can reflect operating conditions, even when the distribution of test data changes, which helps distinguish faults from normal adjustments and disturbances that controllers can settle. Data from numerical simulation and the Tennessee Eastman process are used to illustrate the effectiveness of the proposed fault detection method.
A new intelligent fault diagnosis framework for rotating machinery based on deep transfer reinforcement learning
The advancement of artificial intelligence algorithms has gained growing interest in identifying the fault types in rotary machines, which is a high-efficiency but not a human-like module. Hence, in order to build a human-like fault identification module that could learn knowledge from the environment, in this paper, a deep reinforcement learning framework is proposed to provide an end-to-end training mode and a human-like learning process based on an improved Double Deep Q Network. In addition, to improve the convergence properties of the Deep Reinforcement Learning algorithm, the parameters of the former layers of the convolutional neural networks are transferred from a convolutional auto-encoder under an unsupervised learning process. The experiment results show that the proposed framework could efficiently extract the fault features from raw time-domain data and have higher accuracy than other deep learning models with balanced samples and better performance with imbalanced samples.
📦 How AWS used ML to help Amazon fulfillment centers reduce downtime by 70%
The retail leader has announced it uses Amazon Monitron, an end-to-end machine learning (ML) system to detect abnormal behavior in industrial machinery — that launched in December 2020 — to provide predictive maintenance. As a result, Amazon has reduced unplanned downtime at the fulfillment centers by nearly 70%, which helps deliver more customer orders on time.
Monitron receives automatic temperature and vibration measurements every hour, detecting potential failures within hours, compared with 4 weeks for the previous manual techniques. In the year and a half since the fulfillment centers began using it, they have helped avoid about 7,300 confirmed issues across 88 fulfillment center sites across the world.
DuPont + Augury: Driving Innovation with Predictive Maintenance
Detecting dangerous gases to improve safety and reduce emissions
The primary advantage of differential optical absorption spectroscopy is its scalability. Two elements are required: a calibrated light source tuned to emit a specific wavelength, and a receiver able to read the same wavelength. In some cases, the receiver must also read a reference source for comparison. The two elements can be within the same housing to function as a point detector, but the source and receiver can also be separated, sending a beam across an open path, looking for a cloud of the target gas to move into its field of view.
Yokogawa and Mitsubishi Heavy Industries to Undertake AI-enabled Robot System Project for the Nippon Foundation - DeepStar Joint Research & Development Program
The aim of this project is to develop an automatic inspection system that utilizes robots to identify and predict hazards in offshore facilities. The use of a wide variety of robots to enable unmanned operations and thereby reduce the risk of performing inspections on offshore platforms has long been considered; however, the centralized coordination of individual robots is complex as it requires the management of multiple systems and the data that they acquire. Yokogawa has already been engaged in the research and development of a robot service platform that centralizes the management of multiple robots and seamlessly links them with existing control systems. Leveraging the findings of this R&D, this project will build a communications infrastructure and robot system that is well suited for the environment found on offshore platforms, and utilize an AI application to convert for use in offshore platform operations the image and sound data acquired by robots.
Detecting low-flow cavitation using predictive maintenance system SAM4
Introducing new Google Cloud manufacturing solutions: smart factories, smarter workers
The new manufacturing solutions from Google Cloud give manufacturing engineers and plant managers access to unified and contextualized data from across their disparate assets and processes.
Manufacturing Data Engine is the foundational cloud solution to process, contextualize and store factory data. The cloud platform can acquire data from any type of machine, supporting a wide range of data, from telemetry to image data, via a private, secure, and low cost connection between edge and cloud. With built-in data normalization and context-enrichment capabilities, it provides a common data model, with a factory-optimized data lakehouse for storage.
Manufacturing Connect is the factory edge platform co-developed with Litmus that quickly connects with nearly any manufacturing asset via an extensive library of 250-plus machine protocols. It translates machine data into a digestible dataset and sends it to the Manufacturing Data Engine for processing, contextualization and storage. By supporting containerized workloads, it allows manufacturers to run low-latency data visualization, analytics and ML capabilities directly on the edge.
Common challenges to machine health and ways to overcome them
Dual Linear Phased Array Corrosion Mapping
Asset health is paramount to the efficient and safe operation of facilities producing energy and manufactured goods. Ultrasonic corrosion mapping is a non-destructive testing (NDT) technique that uses data from ultrasonic measurements to map material thickness across a piece of equipment, such as tanks, pipes, and pressure vessels. The data is used to graph corrosion on the equipment for easy visual interpretation. Currently, there are a number of tools available to complete corrosion mapping inspections. However, one automated dual linear phased array technique offers increased productivity, accuracy, and data density over other methods.
Detecting Corrosion and Erosion in Horizontal Boiler Tube Assemblies
Boilers play an essential role in improving the efficiency of thermal power generation. Three boiler sections, economizer, superheater, and reheater, are tightly bundled tube assemblies inherent to the process by maintaining high temperature feedwater and steam that drives the steam turbine and generator. Tube assemblies can be vertical or horizontal, but the focus of this article are assemblies in the horizontal configuration. Because of the curved design, depth of tubing, location, and contents they are subject to a variety of corrosion and erosion mechanisms that can result in failure and unplanned outages.
The susceptibility for failure in a tube assembly is further exacerbated by inadequate inspection methods for detecting or predicting corrosion and erosion damage. However, specialized robot-based NDT techniques, such as Rapid Ultrasonic Gridding (RUG), offer unparalleled coverage and data compared to traditional methods, giving owner/operators the confidence that their equipment can operate optimally.
Using digital twin for cost-efficient wind turbines
CBM of the wind turbine is usually conducted by monitoring vibration at many points on each component with dedicated sensors. Simply increasing the number of monitored points and components leads to an increase in monitoring cost. In our approach, the digital twin acts as virtual sensors for monitoring any component whose behavior can be simulated from a smaller number of sensors as input to the digital twin. Thus, CBM with the digital twin contributes to identifying critical turbines, components, and positions that need maintenance.
MSWR-LRCN: A new deep learning approach to remaining useful life estimation of bearings
Rolling bearings are important components of industrial rotating machinery and equipment. The prediction of the remaining useful life (RUL) of rolling bearings is of great significance for improving the safety of the machine, reducing the economic and property losses caused by the failure of the bearings. However, for the task of predicting the RUL of rolling bearings, the information of the past time and the future time are as important as the information of the current time. In order to make better use of the extracted features for RUL prediction of rolling bearings, this paper has proposed a novel deep learning framework of multi-scale long-term recurrent convolutional network with wide first layer kernels and residual shrinkage building unit (MSWR-LRCN). The major difference from the previous deep neural network is that our new network organically combines the attention mechanism with multi-scale feature fusion strategy, and improves the anti-noise ability of the entire network. In addition, moving average (MA) method and a polynomial fitting model are also used, which help predict the RUL of rolling bearings effectively. The results show that this method has improved the prediction accuracy compared with the existing methods.
Augury Raises $180M To Become One of the First Industry 4.0 Unicorns
The new investment and valuation is a validation of the emerging Machine Health category, of which Augury is the pioneer and leader. Machine Health uses the Internet of Things and Artificial Intelligence to predict and prevent industrial machine failures and improve machine performance. Machine Health allows manufacturers to reduce downtime, increase production capacity and productivity, optimize the cost of industrial asset care and accelerate their digital transformation.
Augury’s customers include some of the world’s top manufacturers, including Colgate-Palmolive, PepsiCo, Hershey’s, ICL and Roseburg. The company’s Machine Health solutions deliver an ROI of 3x-10x for customers, with programs paying for themselves within months.
Augury Becomes a Unicorn But Machine Health is Just Getting Started
Augury went into overdrive in 2021. Our revenue grew 150% and our team doubled as we made our 100-millionth machine recording. We saved one customer a million pounds of snacks and another 2.8 million tubes of toothpaste. We are helping our customers make medicines, produce clean water, and deliver so many products that make our life better, from diapers to construction materials, snack foods to vaccines. With this new funding we can continue to expand globally, innovate in Augury’s core manufacturing market and step into new ones.
Machine Monitoring Becomes Simpler And More Affordable Than Ever
What makes all this possible is a new application of a simple technology—the current transformer, essentially an amperage meter. As Dunford explains, maintenance engineers have used these small, inexpensive devices for decades to detect, for example, when a machine starts drawing excess power, possibly indicating a need for maintenance or even an impending malfunction.
Guidewheel uses the same information to detect when a machine is running or stopped, how long it has been running or not, and the number and period of cyclical operations. In the case of continuous operations such as extrusion, the level of current draw can be correlated with production rate.
COBRA: COntinuum roBot for Remote Applications
A pressing case for predictive analytics at MacLean-Fogg
Metform chose to focus specifically on the AMP50XL’s drive train because “that was the area where we saw the biggest opportunity for improvement.” While they’d previously been gathering data from the machine for predictive-maintenance use, the old process was neither efficient nor of adequate detail, they realized. “From a data collection standpoint, there was a lot of spreadsheets, a lot of handwritten notes, a lot of tribal knowledge,” Delk said. “We wanted to make sure we could gather that information and put it into context as we were analyzing the equipment.”
“We’re able to monitor the machine health, see in real time how the machine is doing and see a signal of a problem before it becomes a major problem. We have a long way to go in terms of learning how to better use the system and gain further confidence in the system, but at this point, I’m really pleased with the progress we made. I’m anxious to expand this to the other nine Hatebur presses.”
Graph-based semi-supervised random forest for rotating machinery gearbox fault diagnosis
Random forest (RF) is an effective method for diagnosing faults of rotating machinery. However, the diagnosis accuracy enhancement under insufficient labeled samples is still one of the main challenges. Motivated by this problem, an improved RF algorithm based on graph-based semi-supervised learning (GSSL) and decision tree is proposed in this paper to improve the classification accuracy in the absence of labeled samples. The unlabeled samples are annotated by the GSSL and verified by the decision tree. The trained improved RF model is applied to the fault diagnosis for the rotating machinery gearbox. The effectiveness of the proposed algorithm is verified via hardware experiments using a wind turbine drivetrain diagnostics simulator (WTDDS). The results show that the proposed algorithm achieves better accuracy of classification than conventional methods in gearbox fault diagnosis. This study leads to further progress in the improvement of machine learning methods with insufficient and unlabeled samples.
Efficient federated convolutional neural network with information fusion for rolling bearing fault diagnosis
In the past year, various deep learning-based fault diagnosis methods have been designed to guarantee the stable, safe, and efficient operation of electromechanical systems. To achieve excellent diagnostic performance, the conventional centralized learning (CL) approach is adopted to collect as much data as possible from multiple industrial participants for deep model training. Due to privacy concerns and potential conflicts, industrial participants are unwilling to share their data resources. To solve the issues, this study proposes a fault diagnosis method based on federated learning (FL) and convolutional neural network (CNN), which allows different industrial participants to collaboratively train a global fault diagnosis model without sharing their local data. Model training is locally executed within each industrial participant, and the cloud server updates the global model by aggregating the local models of the participants. Specifically, an adaptive method is designed to adjust the model aggregation interval according to the feedback information of the industrial participants in order to reduce the communication cost while ensuring model accuracy. In addition, momentum gradient descent (MGD) and dropout layer are used to accelerate convergence rate and avoid model overfitting, respectively. The effectiveness of the proposed method is verified on a non-independent and identically distributed (non-iid) rolling bearing fault dataset. The experiment results indicate that the proposed method has higher accuracy than traditional fault diagnosis methods. Moreover, this study provides a promising collaborative training approach to the fault diagnosis field.
The Cost of Unplanned Downtime for Refineries
According to the American Institute of Chemical Engineers (AlChE), the cost of missed production for a U.S. refinery with an average-sized fluid catalytic cracking unit of 80,000 barrels per day will range from $340,000 a day at profit margins of $5 per barrel, to $1.7 million a day at profit margins of $25 per barrel, based on a conservative estimate. A single, unplanned shutdown that lasts hours can lead to the release of a year’s worth of emissions into the atmosphere, according to John Hague, Aspen Technology Inc.
One type of innovative inspection process is Rapid Ultrasonic Gridding (aka RUG), which creates data-rich visual grid maps that identify areas of corrosion and other damage mechanisms. It is 10 times faster than traditional gridding and competing methods. In most situations, the operator can quickly make the decision of whether to proceed with maintenance measures to resolve the issue, or to return the inspected asset to operation.
Sensor-based leakage detection in vacuum bagging
A majority of aircraft components are nowadays manufactured using autoclave processing. Essential for the quality of the component is the realization of an airtight vacuum bag on top of the component to be cured. Several ways of leakage detection methods are actually used in industrial processes. They will be dealt with in this paper. A special focus is put on a new approach using flow meters for monitoring the air flow during evacuation and curing. This approach has been successfully validated in different trials, which are presented and discussed. The main benefit of the method is that in case of a leakage, a defined limit is exceeded by the volumetric flow rate whose magnitude can be directly correlated to the leakage’s size and position. In addition, the potential of this method for the localization of leakages has been investigated and is discussed.
Haschenburger, A., Menke, N. & Stüve, J. Sensor-based leakage detection in vacuum bagging. Int J Adv Manuf Technol (2021).
How SparkCognition Improved Production Efficiency for a Beverage Manufacturer
We developed seven new deep learning models to detect anomalies in resource consumption, machine status/health, and overall efficiency. (As always with a Total Plant solution, these models were tailored to the specific data, technical context, and business goals and strategies of the client.)
Once developed, the models were deployed into our AI platform for execution and KPI-driven reporting. Another key new function we delivered: predictive analysis, to anticipate problems before they occur, based on patterns detected in current and historical data, and notify the beverage manufacturer in time to take preventative action.
Finally, the results of the AI-powered analysis were delivered via a configurable dashboard that provides at-a-glance insight into the plant’s efficiency, including new KPIs reflecting water usage, water balance, power consumption, heat generation, and waste levels. This information can also now be streamed whenever, wherever, and to whomever the manufacturer requires, now or in the future.
Colgate-Palmolive Focuses on Machine Health to Improve Supply Chain Operations
Colgate-Palmolive is feeding this wireless sensor data into Augury’s machine health software platform. Pruitt pointed out that this enables Colgate-Palmolive’s machine data to be compared with machine data from more than 80,000 other machines connected to the Augury platform around the world.
“That massive analytical scale brings us insights on how to optimize the performance of equipment and make ever-smarter choices on how and where we deploy it,” Pruitt said. “What’s possible only gets more compelling as this AI solution harnesses more data to create better health outcomes for our machines and our business.”
Providing a specific example of how Augury’s Machine Health system has helped Colgate-Palmolive, Pruitt noted that the system’s AI detected rising temperatures in the drive of a tube maker and alerted the plant team. “Upon inspection, they discovered a problem with the motor’s water cooling system,” he said. “By getting it quickly resolved, we prevented the drive from failing due to overheating, which would’ve stopped the tube production line and incurred replacement costs. We figure the savings at 192 hours of downtime and an output of 2.8 million tubes of toothpaste, plus $12,000 for a new motor and $27,000 in variable conversion costs.”
Predictive Maintenance ROI: A $432 Billion Boost To The World’s Leading Manufacturers
By extrapolating our findings across Fortune Global 500 (FG500) industrial companies, we’ve calculated that these companies are losing 3.3 million hours in production time annually to unscheduled downtime and taking a near $1 trillion financial hit - equivalent to 8% of their annual revenues.
From the returns seen from our clients, we estimate that the widespread use of advanced, AI-driven machine-health monitoring and Predictive Maintenance could save FG500 manufacturers 1.7 million production hours a year and deliver a 4% productivity boost worth $432 billion.
Equipment Health Indicator Learning using Deep Reinforcement Learning
We propose a machine learning based method to solve health indicator learning problem. Our key insight is that HIL can be modeled as a credit assignment problem which can then be solved using Deep Reinforcement Learning (DRL). The life of equipment can be thought as a series of state transitions from a state that is healthy at the beginning to a state that is completely unhealthy when it fails. Reinforcement learning learns from failures by naturally backpropagating the credit of failures into intermediate states.