Exapro, an online marketplace that specializes in connecting buyers and sellers of used machinery, has recently announced a partnership with Neuron Soundware, a company specializing in detecting anomalies in each individual component of any given machine accurately.
Exapro management is really excited about this partnership as they believe it will add tremendous value for their customers. Dealing with used machinery carries an inherent risk in that the exact condition of the machines is not always known, nor can it be determined easily. Customers will now be able to use Neuron Soundware’s innovative product to detect anomalies accurately. This will allow them to quickly and easily determine which repairs are required. The new partnership represents a symbiotic relationship that will benefit not only Exapro and Neuron Soundware, but also their customers.
Who is Neuron Soundware?
Neuron Soundware is a relatively new company that was founded in 2016 with the help of StartupYard, a CEE program that acts as a seed accelerator for technology startups, helps companies build viable products, launch fast and raise capital successfully.
StartupYard recognized Neuron Soundware with the Startup and Idea of the Year award in 2016. This was followed in 2017 by Neuron Soundware achieving third place in Forbes’ TheNextBigThing and being funded €5.75 million by VC fund JT Ventures. Neuron Soundware was also recognized as Cool Vendor by Gartner in 2018.
Despite its short history, Neuron Soundware has an impressive list of customers, including Siemens, Airbus, Volkswagen, BMW, Mercedes Benz, LG and Schneider Electric. Their diagnostic solution can also be used on a huge range of different and divergent machines, including escalators, wind turbines, cranes, CNC, print machines, turbines and pumps.
In a nutshell, Neuron Soundware recognizes broken machines using sound. This is achieved through a three-step process:
- Sound Recording – sounds and other parameters of the machine are monitored and recorded.
- Neural Network Training – Data is passed through a tailored deep neural network to determine the risk of an outage of each individual device.
- Problem Detection – Unusual or anomalous behaviours are detected in real-time. These are then visualized and a notification sent to the responsible person.
How does it work?
For as long as machines have been around, humans have listened to the noises they make to try and identify any problems. This method, however, relies heavily on experience and intuition and is not always accurate, especially as machines are increasingly becoming more complex. Neuron Soundware has taken this basic concept and, by using cutting edge technology, have turned it into an innovative diagnostic tool that provides automatic data collection, AI prediction and continuous results.
By using a combination of software and hardware, Neuron Soundware provides predictive maintenance, asset monitoring, and acoustic quality control with an end-to-end solution including an advanced edge computing device.
With this service, you can get the most out of the acoustic emissions of your assets to know the roots of an issue as soon as physically possible, 24/7, with analysis of every second of every minute. The solution can also be deployed in fully offline mode.
The Role of Software
The data is visualized on an online dashboard, available on all devices from mobile, tablet, laptop, to desktop. This analytics and monitoring platform allows customers to listen to, visualize, and label recorded data, collect feedback on alerts, and search through the audio database.
Customers not only have the online dashboard, but there is also a possibility to download a mobile app (compatible with their nCard,) which visualizes recording results in real-time.
An AI That Gets Smarter by the Day
Neuron Soundware leverages complex data augmentation techniques and validate models on independent datasets.
Their research is inspired by latest work in the field of AI, including neural networks. Millions of recordings have been collected of all types of machines in order to provide the best possible accuracy.
An AI That Learns Continuously
Every new machine in Neuron Soundware’s fleet allows them to improve the accuracy of their system. Language such as TensorFlow and Keras are used, and they have mastered technologies such as Docker, Kubernetes and Kubeflow.
Neuron Soundware’s infrastructure is designed to process massive volumes of data. Their advanced edge computing not only allows them to process and analyze data directly within the IoT device, but it can be used in a fully autonomous offline mode.
This solution can provide alerts and updates in real-time at any time of the day and year. A variety of sensors are used and every second can be analyzed.
The Hardware that Forms the Foundation of the System
Neuron Soundware uses a suite of propriety hardware products at the very core of their system.
This standalone IoT recording and computing device can be installed to enable audio monitoring of any asset. It contains computing power that will give customers an edge by running complex machine learning models without the need to transfer data to a cloud storage. Up to 12 channels of simultaneous and synchronous recording are available. nBox is available in an indoor, outdoor and also ATEX certified version.
Structure-Borne Sound Sensor
The durable piezo-microphones are the standard sensors used in Neuron Soundware’s installations unless the specific use case requires specialized sensors or microphones. Due to the inertia of the seismic mass, the sensor moves in correlation to the machine motion. This motion results in a compressive force which is converted into a voltage signal via a piezoceramic sensor element.
In addition to the cable version, Neuron Soundware also provide a wireless modification designed for temporary installation. The battery supports up to 6 hours of continuous audio recording and streaming. It is secured in place by strong magnets, so it can be installed and removed quickly and easily.
This custom-designed sound card connects to a mobile phone and allows users to make high-quality audio recordings anytime, anywhere. All data is simply uploaded directly to Neuron Soundware’s central repository.
Recording of broken machines has never been easier. nCard is pocket-sized and lightweight with a 2-channel recording capability (2 microphones/piezo-sensors).
The potential impact of Neuron Soundware’s technologies on your Business
Implementing Neuron Soundware’s technologies can have a huge impact on your business, including the following:
- Lower maintenance labour costs
- Reduced equipment and energy costs
- Reduced lost production time
- Increased efficiency
- Improved reputation and liability management
- Higher efficiency of employee time
All of these points will ultimately lead to savings and improve your profitability.
To put things in perspective, Neuron Soundware presents the following Return on Investment (ROI) example:
Artificial intelligence is not something that is adopted to look good. Neuron Soundware focuses on adding value for their customers.
An economical loss of more than 100k EUR annually can for example be reduced by eliminating unplanned downtime. In addition, more than 75k EUR can be earned by decreasing time spent on maintenance. Adding other savings mentioned in the list above will ultimately result in a total saving of more than 200k EUR per year. This translates to an ROI of more than 20 times.
The Neuron Soundware Solution Helps Solve Quality Problems
At first glance, it seems logical that technological growth should automatically improve manufacturing quality, but in many cases, the opposite is true. There are several factors that make it more difficult than ever before for manufacturers to maintain high production quality.
This has resulted in quality becoming a top priority for many organizations. As Industry 4.makes a massive impact on manufacturing, it’s logical that the same technologies are being leveraged to meet the new demands on quality. This has resulted in the birth of Quality 4.0.
Solving quality control problems will in itself bring about huge savings and improved efficiencies. The Neuron Soundware solution assists in solving quality issues in two main areas:
1. Manufacturing Process Monitoring
Some steps in manufacturing and assembly are done by workers or robots that are not able to perform quality checks on the fly. The quality of these steps or events can, however, be heard. The Neuron Soundware solution can evaluate whether these steps have been completed successfully or not.
Examples of sound events monitored:
Manual and robotic steps in car assembly, such as POPP clamp closing, electrical connector plugging, seal insertion into car body, robotic drilling and chip extraction, and robotic hammering are all good examples of where sound monitoring can be used successfully to improve quality.
Micro case study: Chip extraction
Acoustic checks are used to make sure the drilling stops when the chips are not extracted properly. A system alert was set up to ensure production is notified when chip extraction is interrupted.
2. End-of-line (EOL) Testing Solutions
EOL Testing is usually done by human operators or a combination of sound recording systems and human supervision. In this context, ‘human’ often unfortunately means errors and consequently, unpredictable cost. A method driven by AI can enhance the accuracy of these checks.
Examples of EOL testing monitored:
Good examples of EOL testing monitoring in the automotive industry includes combustion engines, automotive sub-assemblies such as roof systems, door systems and windshield wipers. In the home appliance industry, refrigerators, microwave ovens, washing machines and tumble dryers are all prime candidates for EOL testing monitoring.
Micro case study: Windshield wiper
The Neuron Soundware solution was used to monitor electric motors which were manufactured and assembled together with the gear mechanism. The human detection of faulty ones was normalized and enhanced by using two sets of acoustic checks.
Industry Tailored Solutions
The term “Industrie 4.0” was first used by the German government in 2011 and digital disruption in manufacturing has moved fast from there. Industrial IoT is progressing at a rapid rate as manufacturers start to understand the huge potential this technology has for revolutionizing their operations. Industry 4.0’s leading use cases include fleet management, data-driven R&D, digital twin, predictive maintenance and condition monitoring.
Neuron Soundware has implemented and tailored its solution to a huge range of industries, which can be divided broadly into 2 categories – Heavy Duty and Manufacturing.
The specific industries and machines for which Neuron Soundware has tailored their solution is listed below:
1. Heavy Duty
- Energy & Electricity
- CHP units
- Wind turbines
- Engines & Turbines
- Diesel engines
- Electric engines
- Pumps & Compressors
- Progressing cavity
- Piston and others
- Industrial robots
- Brakes & engines
- Arms and tools
- Point machines
- Conveyor belts
- Shrink-wrap machines
- Filling machines
- Labelling machines
- Press and print
- Food processing
- Portioning machines
- Cutting machines
- Conveyor belts
- Spinning machines
- Weaving machines
- Cutting machines
- Plastic and rubber
- Cutting machines
- Metal and wood
- CNC machinery
Custom Projects of Neuron Soundware
If your type of machine has not yet been monitored by Neuron Soundware, or your project is unique or different in some way, the company will embark on a custom project with you.
Neuron Soundware is a flexible, agile company, and their technology can be adapted to a wide range of use cases. Their team consists of dedicated experts in IoT & machine learning, while their researchers have been developing a unique way of processing sound. They have spent years of work and hours of computational power to develop their technology.
Neuron Soundware is no stranger to a range of industries and they enjoy the challenge of new projects. Some of the industries in which they can make this happen to include:
- Glass processing machines
- And many more
A custom project was undertaken by Neuron Soundware typically consists of 3 phases:
- Installation, Data collection and Data labelling – this phase provides the data that can be used to design appropriate algorithms.
- Algorithm development and testing – during this phase, algorithm details are fleshed out and tested with precision measurement on test data. Neuron Soundware plans algorithm deployment together with the customer.
- Monitoring, deployment and validation – the customer is given access to the Analytics & Monitoring Portal. A master deployment plan is also finalized and presented
If you are intested in the Neuron Soundware and its nBox, do not hesitate to visit Exapro’s dedicated page to learn more.