Manufacturing Analytics

Manufacturing analytics is the use of operations and events data and technologies to ensure quality, increase performance and yield, and reduce costs. Manufacturing analytics is particularly important for the process manufacturing industry including semiconductor, chemistry, energy production, and biopharmaceutics.

Manufacturing analytics is a part of industry 4.0, and is closely related to AI and machine learning, IoT, digitization, AR/VR, electrification, clean energy, and other new technologies. MATLAB® enables data engineers and process engineers to develop defect detection and advanced process control algorithms, and deploy them into applications for industrial systems.

Using MATLAB for Manufacturing Analytics

Detect defaults in manufacturing analytics using visual inspection system based on deep learning technologies and private data from industrial camera, SEM, X-ray images, and other sources. MATLAB can help engineers with the whole workflow including data preparation, AI modeling, and deployment.

9 images of semiconductor wafers, each marked with a class of defect such as scratch and donut.

Classifying defects on wafer maps using deep learning in manufacturing analytics.

Access operation and test data via databases (SQL, NoSQL) in manufacturing analytics, specific file format (STDF), or industrial IoT communication system (OPC) from manufacturing equipment. You can also connect to cloud data using MATLAB Cloud interfaces to popular services like Amazon S3, Azure Data Lake, and Google Storage. 

Table of databases, cloud storage data sources, and message-based streaming, connected to MATLAB Production Server for analytics.

Access big data in manufacturing analytics.

Apply machine learning or multiobjective optimization technology in manufacturing analytics to multiple-variate data to implement advanced process control, monitor the process, predict drift and default, identify root causes, and optimize manufacturing recipes. You can choose from the most popular classification, clustering, and regression algorithms using interactive apps, such as Classification and Regression Learner apps. Automate the process of building optimized machine learning models using AutoML technology includes feature selection, model selection, and hyperparameter tuning.

Classification Learn app showing scatter plot in center pane and options for training models listed in the left pane.

Classification Learner app, in Statistics and Machine Learning Toolbox, guides you through the classification process while minimizing the coding required to create models.

Deploy data analytics functions to manufacturing production systems on embedded edge hardware or enterprise IT systems. MathWorks helps IT and engineering work together to deliver tangible business results by using your chosen IT infrastructure without recoding into other languages.

Deployment diagramshowing MATLAB and Simulink connected to desktop, web, and enterprise systems through Simulink Compiler, MATLAB Compiler, and MATLAB Compiler SDK.

Cloud (enterprise), edge, and embedded manufacturing analytics applications.

A digital twin model helps overcome typical manufacturing analytics difficulties: expensive hardware test costs, difficulty to obtain failure data, time alignment between many sensors, and complex design space. Digital twin models can include physics-based approaches using Simscape™, statistical data-driven approaches, or AI-based approaches. The models reflect the operating asset’s current environment, age, and configuration, which typically involves direct streaming of asset data into tuning algorithms.

Representations of a physics-based model, statistical model, and a CNN as a AI model.

Modeling methods for manufacturing analytics digital twins: physicals-based, data-driven, AI-based.

For more information about machine learning with MATLAB, see Statistics and Machine Learning Toolbox™.


Software Reference

  • pca - Principal component analysis of raw data - Function
  • controlchart - display measurements of process samples over time - Function

See also: Statistics and Machine Learning Toolbox, Deep Learning Toolbox, Optimization Toolbox, Predictive Maintenance Toolbox, MATLAB Compiler SDK, SimEvents, Simscape, Industrial Communication Toolbox

“By partnering with MathWorks Consulting, we developed a robust platform for supervisory control with MATLAB and transitioned our pilot plant to a modern automation control system. This enabled our researchers to rapidly take algorithms from idea to implementation, simulation, and deployment.”

Dr. Ryan Hamilton, Genentech