Digital darkness in manufacturing floor leaves plant management blind

By | 10 minute read | December 9, 2016

leadspace image of green padlock for digitally dark

Most manufacturing plants are digitally dark.  They collect limited set of data from plant floors, essentially only that which pertains to key metrics or key processes.  But for those, there are hardly any digital visibility that drive decision making or predictive insights that illuminates production risks.

Plant managers start the day with little visibility on what is going to happen during the day; neither have idea on what the day’s productivity will be, nor have insight on potential constrains they will meet in the course of the day.  Consequently the plant management is blind to potential risks and mostly reacting to situations.

The maintenance department is flooded with corrective and emergency maintenance work orders even as they religiously keep recommended preventative maintenance schedules because they lack the ability to foresee the condition of machines and adopt the maintenance approach accordingly.

In a digitally dark plant, the quality engineers are in the dark to impending quality issues therefore can only deal with an issue when a quality rejection happens. The manufacturing engineer suffers from the lack of information to improve the process. And of course, the facility manager is also at a disadvantage lacking insight to enable effective the energy and utility consumption.
Above all, the General Manager is handicapped as well due to lack of digital insights from shop floor as he takes vital decisions resulting in sub-optimal decisions, higher costs, lost opportunities and low return on investment and assets.

Today manufacturers have alternatives for managing the plant better.  Transforming the shop floor through digital visibility has potential to improve productivity, lower cost and improve return on assets.

Gain sight through data insight

Today’s production machines and automation systems generate a lot of useful data.  In a discrete manufacturing plant, most automation system have end-to-end cycle time data for each process steps.  Most controllers are programmed to raise faults and warning events in case of exceptions.  Some controllers even capture parametric data like torque, current etc.  Positional information on robot arms are another set of data that robot controllers generate.  Metrology equipment capture very useful quality measurements.  Most companies keep track of productivity figures, maintenance activities and quality inspection reports.

Therefore the challenge most times is not unavailability of data but the ability to collect, organize and analyse the data in order to create insight.  If the data mentioned above can be aggregated, analysed for patterns and correlations and understood, then it reveals opportunities, warns of future pitfalls and show light on optimal decisions.

Aspects of digital visibility: Descriptive, Predictive, Prescriptive and Cognitive

There are three aspects to digital visibility. The first is descriptive insight where simple visualization of current and past data opens up opportunities for improvement. The second aspect is predictive insight – where the past and current data is used to predict the future behavior of the plant process and machines.  The third aspect is prescriptive insight, which is the ability to process the past, current and predicted information to prescribe actionable recommendations to the plant staff.  The fourth and final aspect is cognitive insight where the intelligence derived out of the raw and derived information is used to advice the plant staff for decision support.  These four aspects of the visibility are explained in the 4 sample use cases below.

Increase target productivity through process visibility

Digital visibility of cycle time data for end-to-end process offers invaluable insight into the opportunities to optimize the process, squeeze more out of the assets and increase productivity.   There are instances where discrete manufacturing plants have been able to get 20% increased product count from the same line by optimizing the process; the benefit is not only increased productivity but increased return on assets that can saves capital for added capacity.  Another benefit of cycle time visibility is to identify opportunities for energy optimization.  Intelligently staggering motors and drives starting optimizes energy consumption.

Improve Overall Equipment Effectiveness (OEE) through predictive analytics

fig 1. graph showing how oee is impacted by 3 big losses Figure 1 – OEE impacted by 3 big losses

Overall Equipment Effectiveness (OEE) is a key metric that plant management uses to assess productivity of the plant and plant equipment.  OEE is impacted by 3 key losses; (a) downtime loss – when machine unexpectedly goes down and stops production (b) quality loss – when bad parts are being produced reducing the total product count (yield) and increasing the scrap (c) speed loss – when the production line or equipment is not producing targeted number of parts due to bottlenecks, defective machines or even slower manual feeds.  When such losses occur, the total number of parts produced is less than the target and OEE degrades.  One of the objectives of the plant management is to reduce these 3 losses.

Bringing predictive visibility to these losses enables the plant management to take early action that will avoid occurrence of these losses.  Imagine a situation where the manager can see the probability of meeting target OEE in a dash board even before the shift begins. If the dashboard says there is 70% probability that the OEE is going to be 60% he or she can find out why the projection is so low, and what potential actions could mitigate the risks the lead to poor performance.
Let’s say the plant manager asks the shift supervisor why the overall equipment effectiveness is trending down from 88% to 60%.  In reply, the shift supervisor is able to respond with precise information – line number two, station number 14 has a good probably of going down in the middle of the shift; station number 18 has a good probability of producing bad parts; or line number one is going to be slower than expected because of an inefficient operator. Equipped with the ability to see what’s happening in real time and what is yet to come, the plant manager can work with the shift supervisor to take pro-active actions and keep up the OEE.

Adopt Targeted Maintenance Approach through prescriptive advice and digital validation

fig 2 illustration showing transformation to targeted maintenance approachFigure 2 – Transforming to targeted maintenance approach

With ability to predict equipment health, downtimes and defects, it is now possible for a maintenance engineer to adopt a targeted maintenance approach rather than simple preventive maintenance approach.  The maintenance engineer see the list of equipment that are recommended for maintenance along with criticality and priority in a dashboard.  The prescriptive analytics capabilities are used to convert the probability numbers into specific maintenance recommendations. Even the recommended maintenance plan can be generated with as part of prescriptive analytics.  Now Maintenance engineer can ask the supervisor to create appropriate maintenance work orders.  Once the maintenance work is completed, the digital visibility enables the automated validated of maintenance tasks performed before automatically closing the work order there by offering a true closed loop targeted maintenance system for plant maintenance.   Thus the plant can drive towards zero downtime and save on spare parts and maintenance overheads.

Support decision making through cognitive insights

An automotive assembly plant is served by hundreds of tier 1s producing parts which then supply those parts in time to be used within the plant’s broader manufacturing assembly process. The reality is there are often hundreds of manufacturing plants working together in collusion to produce that one truck which is being produced at 60 – 76 units per hour.
In this situation, when a plant line stops because of a problem, the plant manager hopes whatever has caused the disruption will be resolved quickly because when one line stops, the other 99 lines which are supplying “just-in-time” also stop.  Suddenly there could be hundreds of people sitting around and waiting for decision.

Imagine a capability where the plant manager is advised by the cognitive advisor on the optimal time to stop the line vs. slow down, the duration of stoppage based on the underlying conditions well in advance.  This can be effectively used by the plant and 99 Tier 1s to plan their labor force; may be announce an early lunch rather than waiting for an hour without decision.  The cost savings from such timely decisions can be vast. It’s operational decisions like that, made daily, which help to reduce the total cost of operations.

IBM Plant Performance Analytics Bring Digital Visibility to Plant Floor

fig 3 illustration showing ibm plant performance analytics overview Figure 3 – IBM Plant Performance Analytics Overview

The IBM Watson IoT Platform brings digital visibility to the plant floor. IBM Plant Performance Analytics (PPA) is a new solution offering for manufacturing which was released in October 2016. The solution is designed to predict constraints that impact overall equipment effectiveness and prescribe remedies and bring cognitive insights for decision making.  This tool helps the plant managers, manufacturing engineers, maintenance engineers, and quality engineers be proactive and take early action to avoid losses to OEE.

IBM Plant Performance Analytics turns plant floor data into actionable insights that help increase productivity and reduce operational cost.  IBM PPA collects data from production machines like cycle time, positional information, parametric data and fault and warning events.  Equipment master data, and transactional data from Enterprise Asset Management system, Quality Management system are also integrated. PPA performs advanced analytics on these data to bring out predictive insights, descriptive and prescriptive insights on the production stations and line.

Predictive insight

PPA dashboards offer simple way to visualize predicted Availability, Performance and Quality metrics for the forthcoming shifts; this may be for a line, specific plant area, station or entire plant.  You can drill down to up to a specific asset to find out the root cause for the prediction.  Once you drill down to specific station or an asset, you’ll be able to find out detailed predictions – what’s the probability of the event occurring, what is the probable time to the occurrence, what’s the probable cause or defect or failure.  Historically how much time it has taken to recover should the even occur.  This shines light on most critical potential dangers that impact plant productivity

Descriptive insight

This capability pertains to the capability to analyze the predictions.  What parameters lead to the prediction?  How often has it happened earlier?  What is the impact on downstream operations?  Is the prediction right or is it a statistical anomaly?

Prescriptive insight

Once a critical or “hot” prediction is made, PPA has an optimization model that can recommend best course of action.  What is the best time to repair the machine considering the current production schedule and maintenance plan?;  What if there are not parts, what’s the impact on OE?  What is the next best time to schedule maintenance?  What should be the optimum process set-points to mitigate critical quality predictions?  What’s the impact of line speed on overall OE?
PPA delivers these value as Software as a Service.  We have seen Plant OE improvement upwards of 5% and operational costs reduce by over 7% with advanced analytics.

Analytical models and Information to represent plant operations

PPA incorporates information model based on ISA-95 standards and hence able to represent any manufacturing operations.  PPA can capture the plant organization, understand relationship between stations, devices within stations and pay points thus able to realistically predict performance at station, line and shop floor level.

Also included are Industry Analytical Models built specifically for production operations.  For instance the weld station model captures the behavior of robots, weld attachments and clamps that work together in a weld stations in order to deliver station performance predictions.   These pre-built models substantially bring down the project start-up time and delivers ready to use content.

The solution comes with mobile friendly applications that are focused on plant persona.  For example the plant ME application focuses on downtime predictions, analysis and maintenance advising.

How are cognitive capabilities used within the optimization process?

In many organizations the primary focus is on optimization of operations and predictive capabilities that can help to predict risk. But what of cognitive? Why is it so important? The reason it’s important is most of the predictive capability available is tactical. It is numbers – alerts based on percentages of probability, directing attention to a potential error that is going to happen, within a certain period of time. These are tactical numbers for plant managers or staff who may not have the capability to interpret or assimilate them as raw data. It is the cognitive capability of Watson which processes these predictions and their descriptions, converting them into actionable steps which a field person or a plant manager can easily assimilate, understand, and act on.

Where to find additional information:

Learn more about IBM’s IoT for manufacturing solutions.