ALL POSTS BY

Blog

Industrial Performance Optimization (IPO)

Let's talk about IPO: Industrial Performance Optimization. 

 

In Industry, the general goal of IPO is to achieve optimal performance on the factory floor. Metrics are gathered and then analyzed by planning software to identify potential improvements to the factory line layout, materials handling steps, staffing, operator procedures, etc.

 

There are three primary metrics to measure IPO: availability, performance, and quality.

 

These are all time-based metrics, further defined as:

 

  • Performance = actual output / expected output
  • Availability = (running time - planned stops) / elapsed time
  • Quality = (items made - those needing rework - scrap ) / (total items made)

 

These three metrics can be broken down further into specific key areas such as:

 

  • Efficiency - output / resources, translates into costs.
  • Machine downtime - affects availability.
  • Energy use - translates into costs.
  • Operators waiting - for material or instructions.

 

Find Out About  Predictive vs Preventive Maintenance

 

Analytics Dashboard 

The IPO approach needs these tools to facilitate this planning and analysis:

 

  1. A customized dashboard
  2. Graphs and charts with filters
  3. Rules engines configured to detecting alerts
  4. Off-the-shelf planning software
  5. Advanced analytics to spot trends, highlight variances, and make predictive models
  6. Real time notifications of performance and quality

 

While real time notification is necessary for daily operations, these metrics are also studied offline for after-the-fact analysis.

 

Chalmers Manufacturing Case Study

At Chalmers, in one of their bus factories, an IPO project was kicked off using metrics to determine where to move existing stations and which stations to break into smaller, separate stations.  The goal was to optimize station location and improve performance, while bearing in mind the constraint of the cost of relocation.

 

In order to make this calculation, it was crucial to have quality input data.  In a paper on the assembly line optimization at Chalmers, Artun Törwnli, explains the problem with relying on people to input data into end of shift reports, “However, encountered unconformities between different sources indicated problems with updating of the databases, which reduced the reliability of the collected data for a line balancing procedure.” 

 

To try and address this, the team automated the collection of data, interviewed line workers, and then plugged the data into ProBalance line balancing software.

 

Data on machine uptime, materials handling time, etc., comes from PLC controllers and other devices connected to the ethernet network, which uses routers that translate industrial protocols to IP ones. This data is then transmitted to the cloud for analysis; some of the data is fed into a constant flowing data process, called a “machine learning pipeline,” while other data is manually entered into off-the-shelf planning software.  Using Jupyter Notebooks, programmers can then walk through that data to build up their models and do their data transformation in stages.

 

 

Setting Benchmarks 

In order to flag variances and achieve IPO, we must first establish norms and baselines, which can be based on observations of the standard operating procedures. A model is built to predict performance, availability, and quality.  The data collected about these three variables is analyzed against actual results to highlight variances.  In a big data environment, these metrics can be fed into a machine learning pipeline, with dashboards showing results and alerts flagging out-of-bounds conditions and spotting trends. 

 

Then, as the factory makes improvements, the norms can and will be adjusted again as operational metrics improve.

 

 

blog post manufacturing industry 4.0.jpeg

 

One potential measurement is the flow of material through the assembly and subassembly stations. The goal at Chalmers was to correlate quality with number of stations, number of operators at each station, target take time, precedence of operations, and velocity of input. This is the simplest of machine learning problems: simple regression.  

 

So, how did it go?

 

Chalmers input their data into Probalance software and used the Simplified Layout Planning method, with the restraint set to minimize reorganization costs, to make optimal changes to the line.  The resulting IPO: keep stations with high relocation cost at their current location, move other machines, and add subassembly stations as needed.

 

In addition to making changes to the line, interviews with operations and further analysis showed that:

  • Material needs to be positioned at the operator’s station at the correct time.
  • The operator needs to be able to use both hands.
  • Work instructions need to be available and clear.

 

As you can see, Chalmers was able to pinpoint exact areas that could be optimized to improve profit and performance. IPO can empower any company or industry to identify key areas where small changes can result in big gains. If your business needs a tune-up, but you’re not sure where to start, contact relayr. Together, we’ll discuss how IPO will work for you.

 

Find Out About  Predictive vs Preventive Maintenance