Easy to use multivariate process monitoring solution for on-line or at-line applications

Transcription

1 Bring data to life Easy to use multivariate process monitoring solution for on-line or at-line applications Predict and correct process deviations before they become problems Customizable, easy to set up and easy to use for process operators Affordable, flexible and scalable for almost any process environment

2 02 CAMO Unscrambler X Process Pulse Process Pulse has proven to be versatile and configurable to suit our unique management system for monitoring and controlling our production processes. Process Pulse has been very useful for PAT (Process Analytical Technology) implementation in our manufacturing plant. Josep Ma. González, Director of Quality, Menarini Group, Spain Process Pulse can help organizations: Reduce process failures with Early Event Detection Improve yields through better process understanding Reduce variability and improve end product quality Reduce raw material, scrap, energy and rework costs Accelerate scale up from R&D to production scale Speed up cycle time and run processes closer to limits Enable knowledge and technology transfer Bring data to life

3 Software for every industry Unscrambler X Process Pulse can be applied across many industries and research fields to improve product development, manufacturing and quality control by near real-time monitoring of processes with powerful multivariate models. Pharmaceuticals & Biotech Research & Academia Chemicals & Petrochemicals Energy & Renewables Agriculture Automotive Food & Beverage Medical Devices Pulp & Paper Electronics Mining & Metals Engineering Oil & Gas Semi-conductors Manufacturing Aerospace & Defense Unscrambler X Process Pulse 03 CAMO

4 04 CAMO Unscrambler X Process Pulse Process Pulse can be used for end-point modeling, with the plot above showing a pharmaceutical drying process monitored using NIR spectroscopy. The model is trained to identify when the process has reached its end state i.e. sufficiently dry. New batches can be monitored as they move toward the end point, shown by the area in the ellipse (top left Scores plot). This enables users to make sure they don t stop the process too early nor continue it longer than necessary. Process Pulse has extensive drill down options. The figure above shows parameters from an injection molding process, with the predicted quality clearly deviating as the process evolves. It is possible to drill down into individual predicted samples to look at the raw data for all the variables and compare them with calibration data. Operators can drill down further into individual variables to see what the raw data look like for neighboring samples. Improve process performance with real-time multivariate process monitoring If you are involved with a manufacturing process, you ll understand the value of being able to identify and correct deviations before they become problems. Yet most solutions are either limited in their ability to analyze complex process data, too difficult to deploy, or too expensive for small or mid-size process environments. Now you can monitor at-line processes using powerful multivariate models, without the large price tag or complex installations of other solutions. Process Pulse is a plug n play solution that is easy to set up and can be used by process operators. It allows you to apply the power of multivariate models to near real-time data (at-line) for small systems and portable units, or scaled up and integrated into large control systems. It offers predictions, classifications and projections. Results are graphically displayed in an intuitive interface. Importantly, Process Pulse visualizes the process based on the output of multivariate models, rather than being limited to traditional Statistical Process Control (SPC). It is ideal for applications where diagnostics are required for critical quality decisions. The solution has the flexibility to handle almost any type of data. It lets you take the pulse of a process as it progresses, giving you full visibility and control, with the ability to address issues as they arise, not when it s too late.

5 Unscrambler X Process Pulse 05 CAMO The figure above shows the predicted concentration of three different components in a pharmaceutical blending process. The upper right Scores plot indicates that the overall product is within specification, with all data within the ellipse. Projection is a new feature in Process Pulse 1.1. Projection enables the user to apply a multivariate model to new samples to see how the new samples compare with the calibration data, even in cases where there are no responses to predict. Simultaneous use of multiple prediction models is a new feature in Process Pulse 1.1. This provides the flexibility to monitor different responses in a process, which is very helpful for understanding complex processes. In a process with many variables, it can be difficult to monitor all process parameters and detect underlying relationships with univariate charts. The top half of the screen above shows the Scores plot, which is a multivariate trend chart that captures the most relevant underlying variability in data, condensed into one plot. Example applications of Process Pulse Process Pulse is suitable for continuous manufacturing to provide feed forward and feedback information, or as an alternative to standard vendor packages in spectroscopy and process. It is designed for applications including: Unit operations in pharmaceutical or biotech manufacturing Test beds in pilot laboratories Proof of concept for scaling up to manufacture scale Run-time applications for spectroscopic analyzers e.g. Gasoline blending Fluid bed drying Grain blending Online food quality testing Ingredient mixture monitoring Measuring ore or mineral quality on conveyor belts New in Process Pulse version 1.1 Projection: Non-quantitative process trend analysis. Indicates that a process is on target and not deviating out of specification. Shows how well new samples fit calibration samples from existing or known data Run multiple models: Have 3 or 4 predictions or classifications simultaneously. For example, predict protein, moisture and starch in one measurement Improved Multivariate Statistical Process Control charting: Includes influence plots with drill down, ability to display data in its raw and transform state, predict responses with confidence intervals Additional import options: OPC compatibility, connection to OSI PI, TCP IP, PE etc. (customization may be required)

6 06 CAMO Unscrambler X Process Pulse During prediction and process monitoring, outlier statistics are powerful tools for identifying if samples differ to those used in the calibration set. Such outlier statistics are used for detecting process deviations, potential process upsets or measurement issues. The plot to top right shows the Hotelling T 2 statistic, which is the multivariate counterpart of the Student s T and is a powerful tool for outlier detection in many applications. Outlier statistics are used in identifying if samples are different to those used in developing a calibration. Q-residuals (bottom right plot in figure to left) is another powerful outlier diagnostic in Process Pulse. It detects if completely new variations are observed in new samples compared to the calibration data. For example this may indicate if equipment failure is about to occur or if new constituents are found in a chemical reaction. Contribution plots (figure above) drill down into Hotelling s T 2 or Q-residual plots to show the contribution of individual process parameters or spectral peaks. This helps process operators determine which variables require adjustment to bring a process within specification. How Process Pulse works 1 Develop multivariate Classification and/or Prediction models in 4 The Unscrambler X software based on historic or known data, then simply import the models into Process Pulse. Alarms can be set if the sample is reading outside a pre-set limit, allowing the operator to shut down the process and investigate the root cause of the issue. 2 Process Pulse provides a simple configuration wizard so users can be up and running without hardware links or complicated 5 Plots of Hotelling T² and/or Q residuals at user-defined significance levels allows for diagnostic monitoring of data. data transfers. 6 Outlying samples can be easily detected, and using drill-down 3 Process Pulse reads data collected on instruments such as FTIR, NIR etc or process data via OPC directly and analyzes in real-time using the pre-defined multivariate prediction and options in the plots, the variables that are different from others can easily be identified and such information can be used to set operating conditions in the process. classification models.

7 Unscrambler X Process Pulse 07 CAMO Features of Process Pulse version 1.1 General Real Time Analysis Wide Range of Data Imports Accepts ASCII and Excel formats Supports most spectroscopic data formats including JcampDx, Grams (*.spc), Bruker OPUS, Brimrose (*.BFF4), ASD Indico, Omnic (*.spa), Varian (*.bsw),guided Wave, NSAS (Foss NIR Systems), Perkin Elmer (*.sp) OPC ready Easy to Set Up Plug n Play style setup Uses The Unscrambler X predictive and classification models Set up wizard guides user through the process Simple product tree and data view formats Utilises full capabilities of The Unscrambler X Models Prediction with regression models, Classification and Projection Automatic pre-treatment of new data before application of The Unscrambler X models Set the number of components to use for each model in the configuration setup Utilises all pre-treatments available in The Unscrambler X Apply PLSR/PCR and MLR predictive models Apply PCA Classification (SIMCA) models Plotting and Data Visualization Visualization Options Prediction values as line plot over a period of time with deviation and without deviation Visualization of scores as scatter and line plot Outliers detection using Hoteling s T 2 & Q-Residuals Contribution drill down plots Enhanced Plotting Features Plots can be customised using a number of pre-defined themes Interactive drill down to raw data Numerous formats for displaying plots Interactive drop-down boxes to switch between plots Copy and save option to capture data at specific time points Fully Customisable Plots can be added based on specific user requirements Scalable standalone application Can incorporate new methods and preprocessing algorithms as required Batch and Hierarchical Models available with customization Data Security and Integrity Simple Security System Windows domain authentication Audit trailing including time stamping and time zone logging Password protection option for configuration files Ability to see previous history of configuration Simple Database System All result files are time stamped and retained in a simple archive with easy file searching capabilities Results stored in Excel format including all predicted results and diagnostics Separate alarms folder for post analysis of any faults detected Editions and System Requirements 32 bit operating system Windows 2003, Windows 7, Vista, XP (SP2 or higher) compatible Minimum 1 GB RAM recommended Minimum 150Mb HHD recommended Internet Explorer 7.0. or higher Editions and System Requirements Can be configured to enable 21 CFR Part II compliance Key features of Process Pulse Advantages of Process Pulse Process Pulse is the perfect tool to improve process monitoring for small and medium sized processes. It offers: Powerful real-time quality predictions with visualization of results in a choice of plots Outlier diagnostics calculated from PLS models such as Hotelling T², Q residuals, Y predicted with deviation Easy to use, configurable dashboards for process operators with security options Alarms at pre-set limits that alert process operators of potential issues Drilldown plots to investigate variables that contribute (or cause) a process to deviate Flexible and easy to use with existing systems and file formats Export in.csv format, with ability to view historical data easily within the system Process Pulse is ideal for Laboratory Managers, Technical Services, Engineering, R&D or Production departments. It is: Easy to set up and use Scalable and flexible Portable Affordable Independent from instrument vendor software Extensible with modules such as Hierarchical and Batch Models* Able to have Expert versus Operator views* Able to integrate with instruments and control systems* *May require customization

8 About CAMO Software Founded in 1984, CAMO Software is a leader in Multivariate Data Analysis and Design of Experiments solutions with offices in Norway, USA, Japan, India and Australia. Multivariate analysis is a powerful set of data mining techniques that help identify patterns and understand the relationships between variables in complex data, allowing organizations to: Efficiently analyze big data sets Build better predictive models Accelerate R&D timescales Efficiently scale up to production Reduce process cycle time Reduce process or equipment failures Reduce final product variability Improve product quality and safety Supports PAT and QbD programs Additional CAMO Software products & services The Unscrambler X Leading multivariate analysis software used by thousands of data analysts around the world every day. Includes powerful regression, classification and exploratory data analysis tools. Enterprise solutions Customized solutions which can be integrated into automation and control systems to enhance their analytical capabilities. Available for client-server and web-based architectures. Analytical Engines Software integrated directly into analytical or scientific instruments for on-line predictions, classifications or hierarchical models directly from the instrument. Consultancy and Data Analysis Services Do you have a lot of data and information but don t have resources in house or time to analyze it? Our consultants offer world-leading data analysis combined with hands-on industry expertise. Training Our experienced trainers can help you use multivariate analysis to get more value from your data. Classroom, online or tailored in-house training courses from beginner to expert levels available. Our partners CAMO Software works with a wide range of instrument and system vendors. For more information please contact your regional CAMO Software office or visit Find out more For more information please contact your regional CAMO office or sales@camo.com Connect with CAMO NORWAY USA INDIA JAPAN AUSTRALIA Nedre Vollgate 8, One Woodbridge Center 14 & 15, Krishna Reddy Shibuya 3-chome Square Bldg 2F PO Box 97 N-0158 Suite 319, Woodbridge Colony, Domlur Layout Shibuya Shibuya-ku St Peters Oslo NJ Bangalore Tokyo, NSW, 2044 Tel: (+47) Tel: (+1) Tel: (+91) Tel: (+81) Tel: (+61) Fax: (+47) Fax: (+1) Fax: (+91) Fax: (+81) CAMO Software AS