Increasing productivity for machinists: Modern optimization approaches for peak and non-productive times

Part 1: Organizational Perspective — OEE & Machine Data Collection

Jörn Steinberg, oee.ai

Why OEE management in mechanical engineering?

According to an industry analysis, machine-cutting plants generate on average only around 40% of their available time to create value — 60% is due to losses. Compared to FMCG industries (e.g. filling), OEE-centered management is still not widely used in mechanical and plant engineering. The potential is correspondingly great: Anyone who goes from 40% to 60% OEE increases productivity by 50% — without investing in new systems.

OEE in machining: The three degrees of loss

Availability level: Setup, technical defects, missing production orders, load carriers, staff shortages.

Performance level: Often underrated. Employees slow down feed or speed. Even more common in machining: incorrect default values from work preparation. Performance factors of 130-200% are not uncommon — a clear sign that target times do not reflect reality.

Quality level: Waste that is detected either directly on the machine or later in the measurement laboratory.

Typical OEE values in machining without systematic management: 30 — 40%. Realistic optimization target: approx. 60%.

Data collection: How do you get the data?

Depending on the age and equipment of the complex, there are three routes:

• Plug & Play Retrofit: An external IoT gateway is being retrofitted — suitable for older systems, no intervention in the control system.
• Integrated retrofit: OPC UA server or industrial PC, delivered by specialists.
• Native: Modern systems with their own server architecture (e.g. via MQTT) — connection in around an hour.

Failures that the machine itself does not know (breaks, staff shortages, environmental disturbances) are recorded using tablets with an adaptable fault pattern catalog. Planning data comes from ERP or PPS systems: which order, in which order, at what speed.

All data converges in a cloud application (European data center), is browser-based retrievable and can be forwarded to other systems via an API interface.

What results from this?

• Real-time cockpit with configurable widgets: OEE, quantity history, heartbeat line
• Loss analysis by shift (early, late, night)
• Waterfall loss: Where exactly is OEE lost?
• Pareto diagram: main causes of malfunctions — e.g. NC programming errors, set-up
• MTTR (Mean Time to Repair) and MTBF (Mean Time Between Failures)
• Andon board above the system, smartwatch notification for the maintenance technician

Why does OEE optimization fail so often?

Two core problems from practice:

Missing reasons for loss: An OEE figure alone is not enough. Without knowing where the OEE is lost, there is no lever for improvement.

Outdated data: If data is not discussed until the next day in a shop floor meeting, no one can react anymore. Online data, on the other hand, creates an immediate behavioral effect: The visible Andon board above the system alone measurably increases productivity — without any technical changes.

A management culture that works with data is also crucial: A shift manager who asks why a set-up process took 52 minutes instead of 30 minutes today achieves more effect than any technical measure. This requires an undeniable, jointly accepted database — no room for discussion about data quality.

Part 2: Technical perspective — prime time optimization in machining

Ivan Jofkov, Pionik/Institute for Machining, Dortmund

Prime time vs. offtime

Value creation only occurs during peak time — when the chips fall. Non-productive time (tool change, workpiece change, set-up) is necessary but does not add value. The goal: to convert the maximum amount of available shift time into prime time.

Optimizing idle time is worthwhile when the proportion is really high — e.g. due to too many tool changes or long start-up movements (more relevant in mass production). In most cases, however, the greater potential lies in prime time optimization.

Rapid threading: punch-step process

A prime example of radical prime time reduction is axial thread forming (Punch Step), developed together with EMUGE and Audi (series start in 2014 at the Göhr engine plant):

Instead of classic tapping with acceleration, synchronized feed and reversing, the thread is formed in half a turn of the tool — helical entry, molding, helical extension in rapid motion. Result: 22 conventional threaded holes vs. 52 with Punch Step in the same amount of time.

The result at system level: In a production cell with 8 machines, the number of machines required was significantly reduced — less space, less investment, less energy consumption.

Energy efficiency: Why giving more gas is better

Only 5 — 15% of the electrical energy used by a machine tool goes into the actual cutting process. The rest is accounted for by base load (hydraulics, control, cooling), coolant pumps and auxiliary functions — regardless of whether the machine is running fast or slow.

This results in a counterintuitive recommendation: Use tools at their limits, don't spare them. Tool costs only account for 2 — 4% of the total costs. If you cut the processing time in half with higher feed rates, you save more energy despite higher forces than by driving carefully.

Specifically, using deep drilling with minimal quantity lubrication as an example: At a feed rate of 0.1 mm/U, 24 kJ of energy is required per hole and a component temperature of over 100 °C. With a feed rate of 0.5 mm/U, the energy consumption falls to approx. 16 kJ — and the component temperature to below 40 °C. Less heat input into the component means less thermal distortions and better dimensional accuracy.

Thermal component distortions as an underestimated problem

The following applies in particular to dry processing and MMS: High temperatures at one processing point can influence the dimensional accuracy at a completely different point of the component — through thermal deformation during clamping. The deviations are only visible after cooling on the measuring machine. Tool and process selection as well as a clamping concept are crucial to counteract this effect.

From the Q&A session

Is peak time optimization also worthwhile for small batches? Yes — but with a different focus. For small batch sizes and expensive components, process reliability is paramount: The first part should be an I/O part. “More careful” does not automatically mean slower feed — this can even be counterproductive during certain operations (e.g. corner milling with a high wrap angle). Roughing and finishing can be considered separately.

Why do many companies have OEE but not optimize? Two main reasons: First, there are no reasons for loss — an OEE figure without breakdown provides no starting point. Second, valuable time is often spent discussing data quality rather than the process. Real-time data that is visible and undeniable for everyone solves this problem. In addition, many companies are trying to save tenths of a second on feed, while there are hours of organizational losses in the same shift. The priorities are not right.

‍