43
Complexities of wayside detection equipment data management Nick Aschberger Software Development Manager

Nick Aschberger, Trackside Intelligence: Complexities of wayside detection equipment data management

Embed Size (px)

Citation preview

Complexities of wayside detection equipment data

management Nick Aschberger

Software Development Manager

Commercial in Confidence

This is my first time speaking at a conference

So.

This will go one of two ways.

Commercial in Confidence

Commercial in Confidence

Niiiiiiiiiiiice

Commercial in Confidence

Here’s what we’ll talk about

1. A bit of background - introduction to wayside condition monitoring devices.

2. What the data is used for and why is this complex?

3. Simple software engineering approaches to handling this data and reducing error.

Commercial in Confidence

Wayside devices - background

A train goes by a wayside con-mon system.

Data for each component passing by (depending on the device type) is recorded.

Many different devices in heavy haul.

Data is used by maintenance planners.

Commercial in Confidence

Device timeline

1950/60s

HBDs & DEDs

1980s

WILD & WID

Hot/Cold Wheels

1990s

Bogie Geometry

2000s

RailBAM

Wheel Profile

Commercial in Confidence

HBD and DED

Commercial in Confidence

WILD

Commercial in Confidence

RailBAM

Commercial in Confidence

Wheel Profile

Commercial in Confidence

But Wait! There’s more! Bogie Geometry

Hi-Speed imaging and recognition (brake pads & shoes, couplers, wedges, etc)

Noise/Squeal (EPA)

Steak knives!

Commercial in Confidence

When planning maintenance?

Prioritization

Condition monitoring data

Existing work requests

Wagons that are maintained

Commercial in Confidence

So what’s so hard about that?

So you just incorporate some con-mon data into your maintenance

process.

Easy Peasy.

Commercial in Confidence

It’s not easy.

If it were easy, I would be unemployed and my children would

starve.

Commercial in Confidence

Commercial in Confidence

There are a number of reasons it’s not easy.

So let’s discuss.

Commercial in Confidence

#1 - Devices are all different

Each vendor supplies a system. Each vendor has different ways of retrieving and interpreting data.

The planner needs expertise in multiple systems. It takes hours (days) to interrogate and compile that data.

Commercial in Confidence

#2 – The data contains errors.

The “dirty secret” of condition monitoring devices.

Devices can make errors – resulting in errors in maintenance planning, or time is

spent compensating/analysing.

Commercial in Confidence

Errors come from many situations

HEAT

DUST

CALIBRATION

WEATHER

SUNLIGHT

(Birds, Snails, Plastic Bags)

Commercial in Confidence

The maintenance system too!

If you are doing the job of planning for rolling stock maintenance, then work

requests in the maintenance system are an input to this decision as well.

As are maintenance campaigns.

But the same problem exists! Data can be entered late, incorrectly or not at all.

Commercial in Confidence

#3 - Data association problems

All modern systems use AEI readers (vehicle AVI tags) to associate readings with

vehicles, axles and components.

But, this is not 100% fool-proof.

(Bad reads, Failed tags, Intermittent tags, Wrong programming)

Commercial in Confidence

AVI Tag Statistics

(Sample data aggregated over our heavy haul customers)

Total AVI tag read attempts 25,603,252

Total missed AVI tag reads 1,296,702 or 5.06 %

Total missed due to

failed/missing tags on vehicles

1,117,062 or 4.36 %

Total missed due to other

reasons (tag reader failure,

environment)

179,640 or 0.7%.

Commercial in Confidence

#4 – How to choose one job over another?

Which is worse?

(Example)

1.A medium level wheel flat alert, or

2.Two low level acoustic bearing faults alerts?

Unfortunately we can’t tell you this!

And it changes over time anyway!

Commercial in Confidence

The NET effect?

If you’re not making optimal decisions, then you’re either:

• Over-maintaining: Extra cost

• Under-maintaining: Increasing production delay risk

Commercial in Confidence

OK, then.

Commercial in Confidence

It’s not so bad

For 80% of decisions, things work pretty well, without even thinking

about this stuff.

But… we need to be better than that.

Commercial in Confidence

It’s not so bad

For 80% of decisions, things work pretty well, without even thinking

about this stuff.

But… we need to be better than that.

Commercial in Confidence

Simple software engineering concepts

We can apply some well known and reasonably straight forward software engineering concepts to this problem.

But be aware there is “No silver bullet”. (Fred Brooks, 1986, No Silver Bullet — Essence and Accidents of

Software Engineering)

Commercial in Confidence

Get everything in one place

The first (obvious?) thing to do is to get everything in one place – condition monitoring and maintenance data.

This removes the expensive effort performed by the planner.

NOTE: That database will have a HUGE amount of data in it.

Commercial in Confidence

Data warehousing

A specific structure of database to:

Handle big data (a lot of history).

Doesn’t need to be used for real time alerting.

Optimized for big queries over large data sets, not individual transactions.

Commercial in Confidence

Data warehousing Transactional (Traditional)

Database Data warehouse

Data sources

Operations, Users, Transactions

Other databases & systems

Purpose To control/capture data transactions as they happen

To consolidate data for

analysis and reporting over time

Features

& Structure

Optimized for transaction

processing.

Fast inserts/updates per

transaction.

Normalized structure.

Fewer database indexes.

Optimized for querying &

reporting.

Faster queries, over larger

sets of data

De-normalized structure.

More database indexes.

Commercial in Confidence

Just doing this 1. Reduction in effort – just consolidating

the data in one spot saves ~50% of the planners time.

2. Much more complex queries can be defined, specific for each business need.

3. Historical repository for analysis.

Commercial in Confidence

Data cleansing. Just chuck it out.

Another (obvious?) technique.

We have lots of data anyway.

Per-instrument rules applied.

This applies to maintenance data too!

Commercial in Confidence

Example – removing a wheel profile error

Commercial in Confidence

Rates of data removed Device type # Train reads

# invalid train reads

% invalid train reads

Laser wheel profile

3399 417 12.2%

RailBAM 11010 0 0%

Wheel Impact Load detector

7315 3 0.04%

Hot Box Detector

3497 21 0.6%

*Laser devices are heavily effected by dust and sunlight in heavy haul *RailBAM cleans its own data

Commercial in Confidence

Total Records 27062

Records deleted as invalid

entries by a user

(open for a period of time)

462 1.7%

Records in an invalid state 493 1.82%

Records that are in a “work

started” state for more than

6 weeks

44 0.16%

Records requesting work

that is already in the system

(duplicates)

1901 7%.

Maintenance data errors

Commercial in Confidence

Trend based decisions Now we leverage the power of the

consolidated data warehouse.

To make informed decisions no individual measurement can be relied on.

Tools that facilitate trending are built on top of the warehouse.

Commercial in Confidence

Trending tools Simple:

• Counts of events

• Database queries that aggregate information for each component

• Running scores that are calculated by the warehouse

• Visualisation tools, where required.

Commercial in Confidence

Trend graph example

Commercial in Confidence

Conclusions (1) 1. Simple (but not easy) software techniques

can be applied to give us higher confidence in our decision making.

2. Build a data warehouse containing both condition monitoring data and maintenance information.

3. The presence alone of the data warehouse itself greatly reduces time spent in maintenance planning.

Commercial in Confidence

Conclusions (2) 4. Apply the principles of Data Cleansing,

and Trending of measurements to the warehouse, to significantly reduce errors in maintenance planning.

5. This in turn reduces cost:

• Reduction in over-maintenance, or

• Reduction in production delay risk.

Commercial in Confidence

We made it