Data Analytics for Manufacturing Excellence

Application of Data Analytics for Manufacturing Excellence

Today most of the data analytics is in the form of monitoring (what is happening) the current state of the system/ process or historical data investigation (What has happened); trying to understand the events. The basic application of statistics (such as covariance/ correlation) helps in doing deterministic analysis such as root cause analysis of the past events and identify ‘Why’ such events happened. There are different applications of data science and analytics, which can help in performing various levels of analytics from diagnostics to probabilistic to augmented data modeling. The Machine learning techniques and application of right methods can unlock the data and give meaningful insights. However, to realize the benefits of this technology, the organizations need to have right ‘Analytics Strategy’, skills, methods and right solutions.

Following are the levels of analytics and possible Key performance indicators (KPIs):

Data Analytics Excellence (L5)

  • Analytics templates & vertical application
  • Roll-out of Analytics Solution across all departments
  • Realization of Return of Analytics & Data

Prescriptive Analysis (L4)

  • Converting trends into future scenarios to make decisions
  • Complex algorithms using multiple source of data to convert into action
  • Exhaustive statistical analysis of process development
  • Statistical process control of quality parameters.
  • Continuous process refinement & enhancement
  • Risk assessment and mitigation

Predictive Analysis (L3)

  • Start modeling future outcomes with past data
  • Modeling, prediction & traceability of quality parameters
  • Performance monitoring and Predictive maintenance
  • Fault/anomaly detection and prediction for Batch and continuous process
  • Causation and correlation investigation in process control parameters
  • Energy & Production cost monitoring and alerts

Descriptive Statistics (L2)

  • Understand State of the System
  • Deterministic Analytics - Performance Monitoring, Process Variability & Correlation Analysis
  • Diagnostic (Why?) — Root Cause Analysis, regression analyses to understand past trends

Basic Analysis (L1)

  • Time-linked Deviation
  • Asset / Productivity Monitoring and OEE -Reliability Engg. (Failure detection)
  • Firefighting to deal with problems

By applying right methods and advanced analytics solution - Seeq, one can achieve following KPI’s

Operational effectiveness / Productivity Analytics

Manufacturing companies need to track and categorize operational effectiveness and performance losses to identify bad actors, justify improvement projects, and do historical and global benchmarking. By employing data analytics solution one can identify times of “at-risk” operation, and save tremendous losses in production, which otherwise would result in unplanned shutdowns, lead to many safety issues and result in millions of dollars of lost production opportunities.

With an advanced process data analytics solution (Seeq), it is feasible to identify, categorize, and summarize performance losses. The summarized analysis is placed in dynamic dashboard for automatically updating periodic reports, which can save 1-5 days of valuable Process Engineer' time per month. Easily exportable historical loss data enables engineers to spend more time adding value to improvement projects and less time developing cost justifications.

Process Variability & Correlation Analytics

Typically, in a manufacturing process, it is often difficult to aggregate data to perform analytics across multiple assets. In addition, continuous monitoring of key performance indicators (KPIs) in near-real-time is necessary to maintain the process within the desired set limits.  KPI monitoring is also required to adjust the process prior to any variables trending outside of the set control limits. A continuously updating, operational dashboard is required to monitor KPIs for the entire continuous process.  This enabled operators to have all process relevant data in an easily viewed format on the production floor. The dashboard identified deviations from the design space parameter ranges, tracked downtime deviations, idle time, and calculated when manual operations such as reagent loading were required.

For historical data analytics and identifying root cause of the events, the descriptive statistics and regression methods such as principal component analysis can be applied to have understanding on the correlation/ covariance and identify synthetic variables, its impact on outputs, etc.

Quality & Traceability Analytics

It is difficult to measure batch quality in near real time, to enable the operator to adjust process inputs effectively.  Currently batch quality is determined by taking samples to a lab and waiting for results from the equipment readings.  This takes some time after the batch has completed – which doesn’t leave an opportunity to save a batch that is going bad and instead it has to be discarded.  Or the energy and raw materials are being wasted on a quality that is already good. Maintaining specific boundary conditions within a process is crucial to product quality and traceability.

Historical batches with good quality are selected as inputs to create operating (good) boundaries for the critical process parameters. A predictive model for output (for e.g. yield) is then generated based on the identified critical parameters and its impact on the quality outputs. The model is then deployed online to detect abnormal batches. Online predictive model allows rapid identification and root cause analysis of abnormal batches to minimize the amount of production downtime. Using the model to adjust process parameters to avoid out-of-specification batches could result in significant financial implications by reducing the quantity of out-of-specification batches, which could save millions.

Energy & Utilities

In a process plant, there are hundreds of items of equipment that contribute to the total energy use. Sensor data from this equipment—such as flow, level, pressure, temperature, and other parameters—is stored in time series data historians. Subject matter experts (SMEs), typically process engineers, can use this data to build predictive models of total energy consumption using each individual item’s sensor data as inputs. With this method, they can determine if each piece of equipment is operating efficiently based on the coefficients in the model. With advanced visualization and dynamic dashboards (such as Tree map View), engineers can monitor their assets and energy consumption rates.

Seeq enables manufacturers organizations to drive sustainability improvements with advanced analytics applications, monitoring visualization for assets at scale, and simple connectivity to many different types of data sources.

Preventive Maintenance and Asset Monitoring

Right Data Analytics Solution (like Seeq) can help to predict future failure events, allowing for proper planning and less cost to fix equipment. Implement a condition-based monitoring analysis to monitor asset health across hundreds of assets. Utilize the historical data to create a predictive maintenance forecast to preemptively detect failures before they occur. The state of overall asset health is created by monitoring its performance with various analytics which leads to: identifying bad actors, risk identification, and prioritization of maintenance activities.  Seeq Tools can also help to predict future failure events, allowing for proper planning and less cost to fix equipment. Reduce unplanned maintenance requirements by thousands by predicting and planning for one failure event.