The evolution in optimization and automation brings efficiency beyond traditional human capabilities.
A company specializing in the integration of artificial intelligence in industry, offering advanced AI solutions for optimizing production processes, enhancing efficiency, and reducing costs. These areas encompass automated inspection and quality control products, as well as services for predictive maintenance, optimization and intelligent control, all aimed at improving performance and competitiveness in the industrial sector.
We are not yet another industrial automation company trying to jump on the AI bandwagon before it is too late. We are a group of originally AI-rooted researchers turned businessmen, with longstanding history in AI that predates its current popularity, delivering complete AI-based products.
Dynamic intelligent control systems are capable of responding to changes in the production process, shifting decision-making from human operators to an automated computer system. This speeds up and streamlines processes, even in operations that traditional management methods can no longer optimize.
Benefit: Labor savings and maximum utilization of nozzle costs by avoiding preventive replacements.
A laser cutting machine uses nozzles that wear out due to plasma flow. We trained a classifier using a database of annotated photos of worn and functional nozzles. The system detects nozzle wear and alerts operators for replacement only when necessary.
Benefit: Replacing human labor and maximizing production capacity by eliminating machine downtime during plan creation.
When cutting thin sheet metal, the resulting parts may tilt and cause damage to the cutting head. We developed an optimization algorithm running on GPUs directly on the machine, which creates cutting plans within 20 seconds—during sheet changes—eliminating downtime.
Benefit: Increased production capacity by preventing unexpected machine stoppages and reducing manual intervention.
After cutting, a robot removes products using suction cups. Sometimes, the cut shape causes parts to jam. We created a "removability" model to predict how easily a product can be extracted, signaling operators to intervene manually when necessary.
Benefit: Automating production processes, eliminating manual stations, and increasing efficiency.
For an automotive seat cover manufacturer, we developed a prototype orientation detector that identifies the main direction of the fabric, ensuring proper alignment during production.
Benefit: Automation, elimination of human error, cost savings, and reduced dependency on key personnel.
Glass bubbles form during production due to configuration parameter variations. We built a factory model to predict glass quality (specifically bubble count), enabling real-time parameter adjustments for optimal quality without human intervention.
Benefit: Increased production efficiency, reduced costs, eliminated delays, and directly boosted profitability.
Semiconductor materials move through multiple production stations, with station usage dictated by the product's recipe. Our agent-based solution dynamically responds to current manufacturing conditions, optimizing workflow to maximize throughput.
Benefit: Automating operations, eliminating human error and delays, and saving energy and steam costs.
Using a simulation of steam management in a chemical plant, we implemented a control system that ensures sufficient steam supply while minimizing unnecessary transfers, optimizing energy use and reducing costs.
A system where artificial intelligence independently inspects defined defects or general deviations from normal conditions in your products. It operates consistently, tirelessly, and quickly — unlike human capabilities.
More about our quality inspection serial products
Benefit: Minimizing complaints caused by insulation defects that result in significant damage to buildings.
On a 2.6m-wide mineral wool production line, we installed a camera system that uses a specially trained neural network to identify material defects, categorize them, and flag products that don't meet quality standards for removal from the line.
Benefit: Reducing complaints; previously, steel defect detection was beyond human capability.
Using a specially trained neural network running on 4 GPUs (circa 2012), we inspected surface defects on a 3m-wide steel strip moving at 10 m/s. This technology enables automated, precise, fast, and tireless surface inspections.
Benefit: Improved quality control, reduced waste, and potential for fully automated glass plant operations.
Bubbles are a common defect in the mass production of sheet glass. A trained detector identifies these bubbles in camera images, similar to surface defect detection.
Artificial intelligence enables groundbreaking changes in work and production methods, increasing efficiency, expanding markets, and creating entirely new products.
Benefit: Enabled production of a flowmeter with larger diameter, which was previously infeasible due to high signal noise.
The reflected signal in an ultrasonic flowmeter is processed to determine the flow rate and material type. When there is too much noise, the conventional methods fail. Using modeled signal propagation and a neural network-based detector, the system operates reliably even with high noise levels.
Benefit: Significantly reduced quotation costs, opening up new market opportunities.
A model estimates mold preparation costs based on CAD drawings, allowing rapid and cost-effective quotations.
Benefit: Increased efficiency and scalability by automating previously manual design tasks.
The customer produces metal castings for the automotive industry. There is a database of CAD drawings of castings and an estimate of the mould making price for already completed projects. When a new CAD model is uploaded, a trained AI model estimates the price automatically.
Benefit: We laid the foundation for the in-house development of an AI parking assistant at a renowned automotive company
A car finds free parking spot through local camera and executes parking maneuver using the camera and ultrasonic sensors only, steered by recurrent neural network controller that was optimized with evolutionary deep reinforcement learning.
Benefit: Launching a new product—tools that prevent operator injuries.
The customer, a manufacturer of hand-held power tools, installs accelerometers in individual tools. From the measured data, it can be determined when the hammer makes the last strike when hammering a nail. This needs to be detected and the tool stopped, as the next blow could cause injury to the operator. We use neural networks to find the last strike in the accelerometer data, according to the shape of the accelerometer signal. Based on this information, the tool is stopped automatically after the last strike.
Let’s explore how machine learning can make your manufacturing processes more autonomous, efficient, and reliable.