Manufacturing

Contents

Preface
Statistical Process Control (SPC)
Measurement Systems Analysis (MSA)
Process and Product Optimisation
Process Prediction and Modelling
Complex Processes and Batch Processes
Reliability
Six Sigma
ISO 9000, TS 16949 and other standards
Data Access,Analysis and Reporting Systems
Customer Satisfaction
Data Analysis in the Pharmaceutical Industry
PROCEED
Reviews and References

Preface

Every day, today's organizations face new challenges resulting from changing customer requirements, stronger competition and rapid technological development. As consumers our expectations are continually increasing, but as producers we have to face the issues involved in satisfying these demands.

In order to achieve optimum product and service quality while simultaneously minimizing costs, many organizations have introduced quality management strategies and standards, such as Six Sigma, ISO 9000 and ISO/TS 16949. For these systems to be effective, a suitable IT infrastructure must be introduced for acquiring data, transforming it into actionable information and then disseminating this knowledge.

Many companies have found through long experience, that effective data acquisition and analysis greatly increases productivity, improves quality and results in more efficient and profitable manufacturing operations. Using data to inform decision-making is one of the most important premises of Six Sigma and the foundation of success for companies using this strategy. Reliable data acquisition, storage and analysis is required by law in pharmaceuticals and many other industries. The software and services we offer will help you to use your data more effectively every day.

This brochure provides examples of some useful and common applications of data analysis, which bring the best return on investment. These include Statistical Process Control (SPC), quality prediction, process improvement, product optimisation and others. For each of the areas described StatSoft Ltd. has a suitable solution, and possesses the necessary knowledge and practice from successful past implementations. Our skills and experience will help you to introduce analytical methods to your organization rapidly and effectively and make maximum use of the knowledge hidden in your data, while meeting appropriate standards and regulations.

"Our cooperation with StatSoft is productive and enables us to carry out engineering tasks much more efficiently than we could on our own. The StatSoft team provides competent technical service, as well as cooperation on solving our own specific problems and tasks. They implemented constant monitoring of our main production processes, creating statistical process monitors using the STATISTICA Enterprise system. These monitors are fully customised to our particular analysis and presentation needs, even though our source data is stored in various external databases, not typical in terms of size and structure. We have an excellent relationship with the StatSoft team, and this has provided new skills and experience to our engineers and increased the professional abilities of our own employees."

Adam Dolinski, SPC programme co-ordinator
Thomson Displays, Piaseczno Plant
One of the biggest colour tube producers in the world

Statistical Process Control (SPC)

Quality Monitoring

Pareto Analysis

Capability Analysis

Back in the 1920s, Shewhart demonstrated conclusively that industrial processes can be effectively controlled using appropriate statistical techniques. Almost a century later his methods are still relevant and being applied to new areas. Quality Control Charts, along with Capability Analysis, are central to SPC and enable manufacturers to monitor and control process variability in order to improve product quality and reduce costs. Again and again, the introduction of this technology has proved to be a key characteristic of the world's most successful, customer-focussed manufacturing organisations.

Measurement Systems Analysis (MSA)

R&R Analysis

Gauge linearity

Gauge stability

Whether you are employing traditional SPC or a fullscale corporate Six Sigma strategy, you cannot avoid making measurements. But before you can start to base your quality decisions on data, it is vital to demonstrate the operational quality of the measurement systems and devices themselves. Ignoring this issue can result in unforeseen additional costs and problems caused by measurement errors. Measurement Systems Analysis is a vital part of any process monitoring operation because it allows you to quantify the variation introduced by the measurement process and separate it from that originating in the production process itself.

Process and Product Optimisation

Design of Experiments (DoE)

ANOVA - Variance Components

Response Profiling and Desirability Functions

Constant improvement of processes and products is vital for the existence and development of modern manufacturing organisations. One of the prerequisites for improving products is to acquire accurate and reliable information about the way manufacturing processes work, and one proven method of getting this knowledge is via designed experiments. This type of research is sometimes dismissed as too expensive and time-consuming, but by using appropriate Design of Experiments (DoE) methods, manufacturers can find answers to vital questions at minimal cost. The data gathered from designed experiments can then be analysed in order to find the optimal combination of process settings. By using advanced tools such as multivariate response profiling, several competing quality parameters can be optimised simultaneously.

Process Prediction and Modelling

Prediction

Root Cause Detection

QC Data Mining

Predictive Quality Control

Modern production processes are often extremely complex, sometimes involving hundreds of process parameters, environmental factors and quality measures. In addition to "classical" mathematical and statistical modelling methods, recent developments in data mining technology such as neural networks and tree-based methods have proven to be very effective in addressing difficult manufacturing problems. A statistical process model based on historical data can be used to detect the causes of defects, identify critical process parameters and even predict events such as machine failures or fault occurrences, without costly and time-consuming experimentation. It is then possible to correct quality problems before they appear, or to terminate the manufacturing process for batches that will not meet customer requirements.

Complex Processes and Batch Processes

Batch Processes

Multivariate Exploratory Techniques

Multivariate Statistical Process Control (MSPC)

In modern industry, processes are very often controlled by automatic systems that can collect and store large amounts of data. However, this information is often left sitting in huge databases or simply discarded. With the application of multivariate methods, these data stores can be a rich source of knowledge about products and processes that can be used to solve difficult quality problems, build statistical process models and apply predictive quality control.

Batch processing industries such as chemical, pharmaceutical, petrochemical and food, require special methods to handle the very large, complex data sets that are generated. Multiway PCA, multiway PLS and multivariate SPC or, where a linear approach is not adequate, predictive trees, neural networks and MARSplines algorithms can be used to ensure critical product quality and give manufacturers a vital competitive edge.

Reliability

Reliability and hazard functions

Failure time analysis

Expected operating life

In many industries (particularly electrical and mechanical products), long-term reliability is a vital element of product quality. It is important to be able to determine the reliability of a product and its expected lifetime. An accurate estimate of reliability, along with a properly calculated confidence range, allows both manufacturers and customers to safe-guard their investments with appropriate servicing and replacement intervals.

There are many practical approaches to reliability management, such as Fault Tree Analysis, Failure Mode, Effects and Criticality Analysis (FMEA/FMECA), Reliability Centered Maintenance, Risk Analysis, etc. All of these implement probabilistic and statistical concepts and require the flexible analytical tools available in STATISTICA.

Six Sigma

Continuous Quality Improvement

Data-Driven Decision Making

Cause and Effect Analysis

Six Sigma was first introduced by Motorola when the company realised that traditional methods of quality control could not bring sufficient benefit when used in isolation. Applying statistical methods within a well-structured corporate culture of constant quality improvement proved to be a powerful and successful strategy that has flourished into a global phenomenon. It is worth emphasizing that the driving force behind this movement is its central focus on statistical data analysis, which had often not been allowed to achieve its full potential in earlier quality movements (e.g. TQM).

ISO 9000, TS 16949 and other standards

Quality Control Charts

MSA

Document Management

Validation and Compliance

One of the ways to meet your customers' demands and concerns is to apply for ISO 9000 series certification. Due to the fact that the original ISO standards did not embrace all of the intricacies of industries such as automotive engineering, other standards have subsequently been introduced. During the implementation of a quality system, 90% of the work is in establishing and documenting Standard Operating Procedures for SPC. A crucial time-saving element is the choice of appropriately integrated analytical and document management software. STATISTICA is unique in combining an FDA-compliant analytical platform with the necessary validation, security and document management features in a single, off-theshelf system.

Data Access,Analysis and Reporting Systems

Integration with Virtually any Data Repositories

Collaboration, Security and Permissions

Centralized Administration

Automated Analysis and Reporting

Web-enablement

Given the massive amounts of data that are generated in most manufacturing processes and the huge cost of data acquisition, it is perhaps surprising that many organisations make so little use of the data they collect. An automated or semi-automated data access, analysis and reporting application can offer significant commercial advantages. Examples include solutions to collect on-line production process measurements and combine these with other types of production, testing, R&D and quality data stored in different, non-integrated, databases to generate reports. Online and offline analysis results and report templates may be made available to specific user groups or published more widely via Internet or Intranet servers.

Customer Satisfaction

Warranty Claims Analysis

Understanding Customers� Needs

Customer Satisfaction Monitoring

Surveys

Complaints Analysis

One of the key challenges for newly established companies is to gain an understanding of their customers' requirements in order to refine and improve products to meet their needs. Quality management strategies, such as ISO 9000, Six Sigma and TQM are focused on customer satisfaction. But customer requirements can only be satisfied by first finding out what those needs are. Equally, it is important in customer relationship management to know the causes of customers' dissatisfaction and the most common reasons for complaints.

Data Analysis in the Pharmaceutical Industry

Computer Systems Validation

FDA, 21 CFR Part 11 Requirements

Manufacturing / Process Analytical Technology (PAT)

Drug Discovery and Development

Quality Control and LIMS

The pharmaceutical industry is set apart from other industries by the strict regulatory framework in which it operates. Maintaining the quality and safety of pharmaceutical products is a critical issue. For this reason, producers are required to validate and certify production processes and demonstrate compliance with regulations. Statistical data analysis methods are vital in this process.

One of the latest, most technologically advanced movements in this area is Process Analytical Technology (PAT), which can sometimes replace traditional process validation methods. In contrast with many other 'niche' applications that offer a limited range of analytical options, STATISTICA provides a comprehensive solution that is ultimately much better equipped to meet the diverse challenges of modern pharmaceutical manufacturing. It provides a wealth of techniques packaged within a single software environment, including Multivariate Statistical Process Control (MSPC) using PCA/PLS, on-line monitoring of batch processes, general linear and non-linear predictive modelling, tree-based models, clustering, multivariate adaptive regression splines, neural networks, support vector machines and other algorithms. STATISTICA integrates directly with process information repositories, LIMS databases and MRP systems, provides tools for access control, configuration management, audit trails and integrated documents versioning.

The STATISTICA product range is fully compliant with all the latest regulatory requirements and even custom-built systems can be rapidly and easily validated. This is also important for other FDA controlled industries such as biotechnology, medical devices and cosmetics.

PROCEED

Predictive Models

Model Exploration

Simulation

The use of traditional data analysis methods in the manufacturing of complex, high-value products can be problematic, and often fails to produce optimum results. This problem has been addressed by a unique collaboration between StatSoft and Caterpillar in the development of a new manufacturing software solution coupling a patent-pending approach proven at Caterpillar Inc. with STATISTICA Enterprise analytical technology.

PROCEED is designed to allow the rapid development of predictive quality control models. It combines both novel and traditional knowledge extraction methods to derive and validate casual relationships between manufacturing processes and product quality outcomes. Data selection, cleaning, modelling, simulation, optimisation and final model implementation, monitoring and approval steps are joined in a single "recipe-based" work-flow interface.

The PROCEED approach achieves results in any complex manufacturing environment, including aerospace, automotive, heavy equipment, power equipment, transportation and industrial manufacturing.

Prediction of product quality from manufacturing process parameters leads to a reduction or sometimes the elimination of expensive and time-consuming finished product testing. Optimising the manufacturing process results in a reduction in necessary post manufacturing product adjustments and the loosening of process or raw material tolerances where they do not contribute to finished product quality. ROI is derived from cost savings in test equipment and labour and from increased product throughput from the reduction of the time spent testing and adjusting. ROI is also achieved by reduction of expensive re-work and scrap, by reducing the rate of product failures and by the use of less expensive process steps or materials.

Reviews and References

COMPUTERWORLD

ALSTOM Power Elblag is a top manufacturer of turbine castings. Cited in a COMPUTERWORLD article, Jerry Slawinski, ALSTOM Power's Quality Manager relates the use of STATISTICA Enterprise (SEWSS) system to optimise production processes and ensure quality:

"The SEWSS system is not just a package that runs on the computer on the desk of an analyst. It is an integrated solution which links production, data acquisition, database and analytical parts� About 30 operators and engineers and a wide group of managers use the system for data collection, data management, and statistical analysis. SEWSS continuously retrieves data: first, components and raw materials characteristics are measured (this additionally controls the reliability of the suppliers), then characteristics of moulding sand (tensile strength and permeability), furnace charge, fire-clay moulders, melted metal composition and pouring process are monitored, parameters of heat treatment recorded, casting quality is measured (repair welding factor) and dimensional compatibility checked. Each cast generates several hundred records of data. Measurement data are retrieved by the system automatically or sent to the system by operators with measurement devices or keyboard data entry. SEWSS stores measurement data in an external database (in this case it is an Oracle database) and runs on a network server... The global market in our sector is very competitive. Turbine castings dropped in price by 50% in the last 5 years. In this situation, quality is not a negotiable option, we talk only about costs and time to delivery. Producers who will not manage to change in time will not stay in this market.
The comprehensive analytical tools, flexible alarm notification, interactive query facilities, and many other innovative features in SEWSS help ALSTOM Power meet these essential quality standards."

Jerry Slawinski, ALSTOM Power's Quality Manager

DM Review

STATISTICA Enterprise (SEWSS) received a glowing review in DM Review magazine from Photronics, a leading worldwide supplier of photomasks (high precision quartz plates used in the manufacture of semiconductors). A team from Photronics conducted a thorough investigation of several SPC packages. "We selected SEWSS because of its ease of integration with our existing manufacturing execution system and because it provides flexibility beyond our existing systems. The scalability and flexibility of SEWSS greatly complements the manufacturing environment at Photronics. SEWSS is being run on an Oracle database at six sites internationally with plans of having six more sites." In the future, Photronics plans to standardize their approach to Design of Experiment (DoE) and Failure Mode and Effect Analysis (FMEA).

Barbara Manville, Photronics
Director of Corporate Software Development

Quality Digest

"Quality Digest magazine published a review of version 6 of STATISTICA QC Charts, Data Miner, and Neural Networks. Author Felix Grant remarks on STATISTICA QC Chart's ability to handle large data sets, multiple databases, and real-time auto-updating saying "There must be limits to the density and variability of data flow, but I've not yet discovered them - despite some very demanding work that would make most software packages cry." Grant calls STATISTICA Data Miner "the best tool I've seen yet for actually applying what you learn at the sharp end of industry" and "the easiest to use data mining control I've encountered." He goes on to praise its friendly user interface and strong set of exploratory analytic routines. "These new additions to the STATISTICA product range take traditional quality further , make it more accessible, and introduce a new level of integration that yield considerable synergy payoffs. If you have unsolved analytical problems, try STATISTICA".

Felix Grant, Quality Digest Software Review

QUALITE REFERENCES

VISTEON Automotive, one of the world's largest suppliers of automotive components and systems, gave rave reviews to their implementation of STATISTICA Enterprise system in popular French magazine QUALITE REFERENCES. In the process of redesigning their SPC system and preparing for QS 9000 certification, they chose the STATISTICA Enterprise (SEWSS) system for a variety of reasons discussed in the article: "SEWSS is really flexible, and we could perfectly match the software to our current system in a very easy and quick way. Each entry is customisable (operator's name, lot number) which allows for very detailed analyses of the process variations. The extensive tools that are offered in SEWSS (DoE, multivariate analyses) were also an important aspect of our choice." VISTEON Automotive is clearly more than pleased with their choice, saying "As of today, our system grows very quickly, and is not limited by our software."

QUALITE REFERENCES
(July 2000, page 45)

DM Review

John Deere is a global leader in providing equipment for use in agriculture, construction, forestry, and lawn and turf care. John Deere's Quality Engineer Navaid Ahmed gave STATISTICA rave reviews published in DM Review magazine.

"Excellent service, support, and broad based statistical analysis capability were the major advantages that STATISTICA had over others considered. Also, the versatility and flexibility of STATISTICA was one of the main reasons for its selection... STATISTICA works well with our CMM systems... [however] the interactive ability of the software is its main strength. The different ways that data analysis can be performed and its ease of use with easy-to-read output are also some of the main strengths of STATISTICA... The decision to select STATISTICA at John Deere was unanimous".

Navaid Ahmed, John Deere Quality Engineer