Speaker

Gary Kent Ramsey

Gary Kent Ramsey

Sr. Engagement Manager,

Softweb Solutions

About Webinar

Underutilization, poor working condition, aging or maintenance-related challenges, unexpected downtime and no local suppliers for spares, are some of the biggest challenges that businesses face today. Delays in timely repairs to malfunctioning machines lead to obsolescence. Furthermore, inaccuracy in production can make it tough for companies to ensure compliance and maintain quality.

Today’s technology such as machine learning and AI can help you identify problem signals early so that you can effectively avoid unplanned downtime and cost impact due to decreased yield or quality. In this webinar, we will show you how automatic detection of anomalies in business processes can empower companies to detect problems in minutes and helps them overcomes problems with a system that relies on traditional thresholding. Our speaker will also discuss the roadmap to deploy smart solutions powered by AI to take the first leap in the realm of digital transformation.

Webinar agenda

  • Introduction
  • Remote monitoring to improve transparency
  • Why traditional methods to find production issues are inefficient
  • Cognitive anomaly detection
  • Common myths vs reality
  • How to implement cognitive anomaly detection
  • Use cases
  • Benefits
  • Smart asset monitoring powered by AI
  • Q&A

Questions & Answers

The following are the answers to the questions that were asked during the live webinar.

Question 1: In order to properly do training of a model, how much data would I need?

Answer 1: How much data you need will depend on your business use case and the desirable outcome. Nonetheless, in the webinar, we showed that we helped our client to detect anomaly and solve their problems with an unsupervised machine learning approach - which was more of an ongoing training model. You can refer to our blog to learn more.

Question 2: What platform/system do you use?

Answer 2: We have built smart solutions leveraging Azure technology stack extensively. Besides, we have developed custom models using R & Python. Our smart solutions are built over our IoT platform – IoTConnect.

Question 3: What types of algorithms do you use for anomaly detection?

Answer 3: It completely depends on the type of data you have. Some of the most common machine learning-based approaches are -

  • Density-Based Anomaly Detection
  • Clustering-Based Anomaly Detection
  • Support Vector Machine-Based Anomaly Detection

If you are not sure about what fits best for your application, get in touch with our experts

Question 4: Can you explain Anomaly Detection use case in renewable energy production?

Answer 4: When it comes to renewable energy production, the first things that come to mind is wind-mill and nowadays solar panels. Since the number of solar panels deployed worldwide is increasing rapidly, I am taking that as an example.

Solar panels are usually placed in areas, which are not easily accessible. Hence, to manage them remotely, panel owners use IoT-enabled solution. In that, they get solar panel’s entity data continuously via IoT sensors. The data get classified by anomaly detection algorithm or other classification machine learning techniques like time series and based on algorithms’ results BRE (business rule engine) created, and get alerts.

We have done a similar project with an energy provider in TX. They have some generators and wanted to reduce downtime by detecting anomalies. Annually, they had to spend a month in maintenance. With anomaly detection, we helped them to reduce that downtime from a month to a week.

Question 5: How does unsupervised machine learning work if the asset is deemed as a bad actor?

Answer 5: As you know, in unsupervised learning, the labels are not available. Hence, the task of an AI agent is not clear-cut. However, if handled well, the unsupervised learning approach can prove more robust because of the following reasons.

  • It can attempt to understand the fundamental structure of assets via clustering.
  • It can separate the asset problems into different categories through feature separation techniques.

Overall, the unsupervised learning approach can find interesting patterns above what we are targeting to look for. In fact, most creatures learn through unsupervised learning. To learn more about the difference between supervised and unsupervised learning and their benefits and limitations, contact our data science team.