Pattern Recognition

Defining Conditions – A pattern is a reliable sample of traits, acts, tendencies, or other observable characteristics. Data streams from the sensors on and around equipment produce precise patterns that are exact sequences and data combinations; so small that humans cannot see them. Mtell Previse agents learn those minuscule, complex patterns using powerful auxiliary processing of the input signals. The agents can then distinguish between normal behavior and impending failure and can learn and adapt when they sense unknown conditions. Pattern recognition based or real data is essential for improving asset integrity because it detects the earliest point of degradation and offers more accuracy than any other monitoring technology solutions.

Machine Learning

Fortifying Assets – The way all equipment operates leaves digital shadows in the data streams that betray its activities. By interpreting them, Mtell agents discriminate between normal operating conditions and the signatures of previous fault conditions using data compiled from machine history and run-time operations. But, when anomalies occur, agents classify them as a new normal state, or a previously undetected fault condition. New faults are converted immediately into much more accurate precise signatures that start far earlier than any anomaly detection can recognize. The agents automatically retrain to adapt and capture the learning from the newly classified anomalies. Mtell agents with Machine Learning exponentially increase knowledge and intelligence over time ever increasing protection from failure and boosting production output.

Detection Guidance

Uncovering the Actual Problem – Whether symptoms of deterioration can be recognized early depends on the monitoring technology. Mtell machine learning detects issues more accurately and earlier than all prevailing methods, recognizing signatures that no human or other solutions can detect. Locating the Actual Problem – Unlike other products, Mtell indicates exactly where the problem occurs. A color chart shows which sensors contribute most to the problem: darkest to lightest. Now, the technician knows exactly where to investigate the problem.

Prescriptive Maintenance

Taking Necessary Action – Prescriptive Maintenance requires early and accurate recognition that an event is looming and then taking automatic action to change what would happen. Mtell agents predict known degradation and failure issues with pinpoint accuracy and schedule action. Service or repair occurs with simple maintenance, well before extended shutdowns and complex repairs. Agents also sense minor deviations from normal operating conditions and request action to inspect to reveal the root causes.

Transfer Learning (M2M)

Sharing Protection – Mtell agents transfer learned behaviors of normal and degradation/failure conditions to similar machines across both local and global networks. After a very short retraining period every machine has the same safety and breakdown protection. Using M2M, agents improve throughput and provide state-of-the-art safety and risk management to multiple locations.

Process Optimization

Improving Net Output – Pioneering customers recognize even greater potential in Mtell’s machine learning technology. Besides detecting machine degradation, they now monitor to detect deterioration in manufacturing processes to improve operational performance and profitability. Product quality, output, and efficiency suffer when operations deviate from normal. Using machine learning, one customer now monitors and improves drilling operations. Another detects very slight changes in operating conditions, which when corrected early prevent spoilage of multi-million dollar product batches.

Process Affects

Preventing Collateral Damage – Detecting symptoms is not the same as identifying the cause. Mtell Smart Machines use machine learning to cast a wide net around the machine to determine the actual cause of damage.

For example: a vibration sensor may indicate damage already done. However, tiny changes in process flows, temperatures, and pressures are messengers of impending failure that show the root cause lies somewhere else. The collective of process sensors for a cooler and separator (that dries wet gas before it enters a compressor) indicate increased vapor velocity and ​
consequent liquid carry-over. This will inevitably cause damage to the compressor impeller, leading to bearing damage, etc.

Mtell Smart Machines move the actual cause to the top of the sensor ranking list. Early warning allows for a small process change that can prevent damage occurring at all.

I want to know more about

Smart Machines