The ProcessMiner blog

FeaturedPulp & paper8 min read

How a paper mill saved $600K a year in chemistry costs

Real-time quality prediction and autonomous chemistry control let one mill trim additive dosage while holding spec — six figures of recurring savings, no capital project required.

Read the case study

Case studies

Quantified customer outcomes

Tissue & towel mill cuts chemistry use 25%

Predictive dosing held wet-strength quality on target while reducing chemical additive consumption by a quarter across grades/SKUs.

Pulp & paper5 min read

Board mill lowers chemistry dosage 18%

Continuous quality prediction and centerlining trimmed additive dosage 18% without sacrificing sheet strength.

Pulp & paper4 min read

Board mill optimizes Kymene consumption 25%

Targeted wet-strength resin control reduced Kymene use 25% while keeping quality inside spec.

Pulp & paper5 min read

Board producer reduces chemistry consumption 14%

A 14% reduction in chemical spend, achieved by acting on predicted quality rather than lab lag.

Pulp & paper4 min read

Wastewater plant lifts sludge dryness to 20%

Optimized polymer dosing raised sludge cake dryness from ~13% to 20%, cutting hauling and disposal costs.

Wastewater6 min read

Injection molder cuts scrap rate 25%

A plastic injection-molding manufacturer used deep-learning quality prediction to reduce scrap by 25%.

Plastics5 min read

First autonomous chemistry control on a tissue machine

A milestone deployment: closed-loop chemistry control running live on a tissue line — the origin of ProcessMiner's physical-AI approach.

Pulp & paper7 min read

Data science

Technical deep dives — our backlink engine

LSTM autoencoder for anomaly detection in Keras

A widely cited walkthrough of building an LSTM autoencoder for rare-event detection in multivariate time series.

Technical12 min read

Building autoencoders in Keras

Companion tutorial covering autoencoder fundamentals and implementation for dimensionality reduction and reconstruction.

Technical10 min read

A rare-event dataset from a pulp-and-paper process (Part 1)

The multivariate time-series dataset behind the rare-event series — widely referenced across the ML community.

Technical9 min read

Rare-event classification in multivariate time series (Part 2)

Modeling approaches for extreme class imbalance, applied to a rare-event downtime dataset.

Technical11 min read

Sequence embedding for clustering and classification

Turning variable-length sequences into fixed embeddings for downstream clustering and classification tasks.

Technical10 min read

Estimating non-linear correlation in R

A practical method for quantifying non-linear relationships between process variables.

Technical8 min read

Wastewater

Industrial wastewater

Polymer dosing in wastewater treatment: getting it right

Why precise, data-driven polymer dosing is the difference between compliant, cost-efficient dewatering and runaway chemical spend.

Is sludge cake density hurting your bottom line?

Small gains in cake dryness compound into major savings on hauling, disposal, and polymer. Here's how to find them.

Article5 min read

AI-driven process and quality optimization in wastewater treatment

How continuous quality prediction and control help modern treatment facilities hold quality and compliance under variable load.

Article7 min read

Driving innovation and compliance in wastewater treatment

Refreshed for today's regulatory landscape — PFAS, biosolids, and the case for predictive compliance.

Pulp & paper

Our founding vertical

Pulp & paper

The challenges and benefits of autonomous chemistry optimization

What it really takes to move from operator-led dosing to closed-loop chemistry control in a paper mill — and the payoff.

Article8 min read

Pulp & paper

Using AI to speed up paper production and lower costs

Where machine learning creates throughput and cost headroom on the paper machine — without new capital equipment.

Pulp & paper

The future of pulp & paper manufacturing

Where the industry is heading as AI, autonomy, and sustainability pressure reshape the mill floor.

Article7 min read

AI in manufacturing

Cross-industry education

Three common misconceptions about AI in manufacturing

The objections we hear most from skeptical operators and execs — and what the reality on the floor actually looks like.

Variability reduction: the metric that decides margin

Why tightening process quality variation — the core of Intelligent Centerlining — is the most reliable lever on manufacturing margin.

Article5 min read

Bridging the skilled-labor gap in manufacturing

As experienced operators retire, AI captures and scales their judgment — helping newer teams run the line with confidence.

The evolution of statistical process monitoring

From SPC control charts to machine-learning process monitoring — what changed, and what it means for quality teams.

Article8 min read

Plastics

Injection molding

Plastics

Deep learning in plastics manufacturing: cutting scrap and defects

How predictive quality models help injection molders and extruders catch defects before they become scrap.