137: Chapter 137 Engineering Master

Jiang Lin sat in the shadow of the old Populus euphratica forest behind the Stone House, with a self-fired pottery cup by his hand, containing sea buckthorn leaf tea slowly boiled using a solar stove.

The tea was very light in color, carrying a slightly turbid yellowish-green, with a faint shimmer of precipitated plant wax floating on the surface.

At the first sip, it was astringent, followed by an extremely faint acidity, and finally a trace of bitterness forced out of the vegetation by boiling water, slowly crawling up along the throat.

This thing could hardly be called delicious, and to some extent, it could even be said to be challenging the baseline of human taste buds.

But Jiang Lin had long been used to it.

In the time flow of the Wasteland, sensory enjoyment was the cheapest and most unnecessary redundancy.

Besides, sea buckthorn was not originally planted for tea.

They were the vanguard of life launching an attack against the inorganic defense line on this dark red Wasteland.

He picked up the pottery cup and took another gulp with his head tilted back.

The astringent sensation slowly dispersed at the base of his tongue, acting like a stimulant, irritating his somewhat sluggish nerves caused by staring at the screen for a long time.

Looking out from this position, those few plots of barely surviving crops from back then could no longer be seen.

Spread out around the Stone House was a whole, well-layered, and rigorously calculated artificial ecological island.

The innermost circle was a dark brown rotation field, and the soil color was much darker than the outer circle; that was an artificial humus layer mixed with a large amount of organic compost and crushed slag.

Potatoes, sweet potatoes, Pumpkin, Soybean, and sorghum were cut into irregular patches by stone ridges thirty centimeters above the ground.

Further out was a shrub defense line composed of sea buckthorn, goji berries, tamarisk, and false indigo.

Their root systems intertwined underground like barbed wire, firmly gripping every inch of barren particles.

These plants inherently possessed drought-resistant and alkali-resistant foundations, with thick leaves and low transpiration, making it easier for them to stand firm on this Wasteland than ordinary crops.

Further away, Populus euphratica, jujube trees, mulberry trees, and elm trees crisscrossed into a forest.

Those Populus euphratica trunks were as rough as dried and cracked rocks, growing upward in a nearly tragic posture, forcefully pressing the dark red Wasteland outside the boundary of the oasis.

The wind blew from the northwest direction, its momentum first shaved off by the outermost sea buckthorn and tamarisk, then passing through the layers of interception by the thick branches of the Populus euphratica forest, and finally falling near the Stone House, leaving only a low rustle carrying dead leaves.

On the coffee table in front of him, the opened laptop split the screen into several windows.

It backed up three Real World supplier evaluation summaries, a pending-confirmation item list for the No. 7 abandoned tunnel test window, and a duplicated version of key data from Xu Man's G-01 Hengtai blind test damage report.

Jiang Lin put down the pottery cup and picked up a POM foot end just dismantled from the right front leg of G-01.

The bottom of the foot end was ground with deep and shallow scratches by the wet cinder and rusted angle steel on the Hengtai thirty-seven-degree gravel slope, and the two-millimeter-deep horizontal scratch groove on the outside looked particularly ferocious, left by brushing against the rusted nails in the broken wood of the narrow channel during the blind test.

FOOT-RF-03, right front foot, domestically produced POM part.

After the Hengtai blind test, the wear of the bottom contact surface showed a non-linear asymmetric distribution.

The outer scratch groove depth was 2.1 mm, accompanied by edge plastic yield; the inner normal wear depth was 0.3 mm.

Cause deduction: Under non-periodic gait, when the right front foot laterally probed to cross obstacles, due to the transfer of gravitational potential energy, it bore an extra lateral load exceeding 1.4 times the design value. The crystallinity of the domestic ordinary standard POM molecular chain was insufficient, and its impact resistance grade and notched impact strength could not cope with instantaneous high-frequency shear force.

The G-01 in the Real World, that prototype anticipated by countless capital and gazes, ran through the Hengtai blind test.

At this moment, it was hung on the maintenance rack by him, waiting for spare parts to arrive before reassembly.

It's just that the realistic situation was that the domestic POM suppliers were still in the evaluation stage of PPT presentations and sample dragging.

For Japanese suppliers whose performance met the standard, the delivery date of their special modified samples was scheduled for September.

As for the core reducer accuracy and lifespan problem, no domestic precision machining factory willing to take orders for reverse repair had been found yet.

The retired old engineer from Boston Dynamics introduced through Teacher Guo Jianye's connections had indeed been contacted, with an excellent resume and rich experience.

But the other party's schedule was filled with plans to enjoy family life with grandchildren, going to the Northeast to see his grandson, and would not return to Jiangcheng until early September.

In the Real World, everything was waiting—waiting for the supply chain, waiting for scheduling...

But Jiang Lin possessed the Wasteland World and time, and did not want to wait.

This was also the fundamental reason why he brought in G-01, which had undergone various cruel stress tests.

The Hengtai blind test had proved one thing.

G-01's underlying software architecture, including that non-periodic gait algorithm based on whole-body dynamics control, passive compliant fault-tolerant mechanism, and millisecond-level foot contact force closed-loop control, was extremely capable.

But the hardware problems exposed by G-01 were equally clear.

The embrittlement and uneven wear of foot-end materials; the precision decay and life collapse of harmonic reducers under alternating loads; the lateral shear deviation of the foot-end sensor protective layer on complex contact surfaces.

In tests like the Hengtai blind test that amplified working conditions to the limit, these problems pointed to the same root.

G-01 was essentially a Wasteland verification prototype, a monster manually put together to survive in this harsh environment, not a commercial engineered product that met ISO standards and had mass production feasibility in the Real World.

Each of its subsystems, when designed, never considered the Real World's supply chain, cost, mass production technology, EMC, electrical safety, environmental adaptability, reliability life testing, and a full set of third-party tests required for industrial site admission.

If he wanted the Low Entropy Workshop to survive in the Real World and turn G-01's technical route into a product line that could continuously generate cash flow, he had to do something that had to be done and thoroughly mastered in both the Wasteland and the Real World.

That was engineering dimensional reduction and reconstruction.

Iterating G-01 from a single verification prototype full of personal heroism colors into a modular, standardized universal engineering chassis architecture that could smoothly migrate to different industrial scenarios.

Jiang Lin checked G-01's chassis once, then returned to the workstation.

He created a new root directory.

[ G01_Platform_Engineering ].

Beneath it were divided into two subdirectories.

[ G01-Production (Mass Production Engineering Solution) ] and [ G-Variants (Multi-Scenario Variant Pre-research) ].

He opened the first one first.

The problems exposed by the Hengtai blind test in the Real World had already been clearly listed in the review report using fishbone diagrams and Failure Mode and Effects Analysis (FMEA) tables.

Now he wanted to tackle these three problems one by one in the Wasteland with materials and time.

The problem of the foot-end material was essentially a problem of polymer material formulation and molding thermal history.

In the Real World, although domestic standard POM had fair tensile strength, the flexibility of its macromolecular chains was poor, leading to high notch sensitivity and insufficient impact resistance grade.

Imported DuPont Delrin series special POM had excellent performance, but it was expensive and had delivery cycle risks brought by geopolitics.

Calculated according to the supplier's public parameters and material grades, the carbon nanotube-modified POM from Japanese suppliers might approach the current target performance package, but the actual measurement results still had to wait for the samples to arrive for verification.

These were the commercial supply chain constraints of the Real World.

But in the Wasteland, he was his own supply chain.

He could use small hot-press molds, molded samples, and tensile testing machines to perform physical spline tests in his crude processing room, and then use MPS to perform Monte Carlo parameter searches on injection molding temperature, holding pressure, cooling rate, and filler ratio.

Starting from the fourth Wasteland survival cycle, this Stone House was no longer just a shelter used to escape sandstorms and the extreme cold of the night.

Lathes, milling machines, simple hot-press molds, small resistance furnaces, bench fatigue testing machines, solar power systems, material sample racks, and that set of MPS (Manufacturing Execution and Process Search) engineering tool chains patched up round after round by Jiang Lin had long turned this place into a crude but usable small material and machining experimental station.

Jiang Lin walked over to the spare parts wall on the right side of the workbench.

Dozens of scrapped or to-be-tested foot ends were neatly stacked on the shelves here.

He scanned the paper labels on them from left to right.

The earliest batch was the simplest pure POM turned parts, without adding any inorganic fillers, and without undergoing any graft modification.

The label recorded: [ Pure substrate, no modification. Test environment: Wasteland rocky area. Result: Driving less than 100 kilometers, friction coefficient rose sharply, end face worn bald, scrapped ].

The second batch was modified POM added with 20% chopped glass fibers.

Glass fiber endowed the material with extremely high rigidity, and the hardness rose linearly, but what followed was a fatal increase in brittleness.

When testing on gravel ground, every time the foot end smashed into angled rocks, micro-cleavage fractures would occur at the edges, with an extremely high brittle fracture rate, just like being gnawed by a dog.

Moving further right was the mixed formula series he gradually introduced and iterated using the orthogonal test method starting from the eighth Wasteland cycle.

Carbon fiber powder was added to improve wear resistance and thermal conductivity.

A specific proportion of elastomer toughening agents such as TPU was introduced to absorb impact energy.

He tried to melt and compound POM substrates of different molecular weights to balance processing fluidity and final mechanical properties.

Behind each batch of new formulas corresponded a long and boring failure record, parameter fine-tuning, and re-hot-press molding in the workstation database.

He took down a foot end from the wall with the label [ POM-CF-07 | Carbon Fiber Modified POM | Ninth Wasteland Late-stage Verification | Cumulative Mileage Approx. 206km | Status: Slight Wear, Usable ], turned it over against the light, and looked at the wear patterns on the bottom.

Holding this foot end, Jiang Lin returned under the magnifying lamp, turned it over, and carefully observed the wear patterns on the bottom.

Under 20x magnification, the wear of the contact surface showed a very uniform matte texture.

There was no local deep-pit peeling, no brittle cracking caused by edge stress concentration, and even less uneven wear caused by the uneven stiffness of the material itself.

Carbon fiber formed a good micro-skeleton in the matrix, effectively dissipating frictional heat and preventing local melting of the polymer surface.

On the harsh terrain of the Wasteland, which was full of radioactive gravel, extreme temperature differences, and sharp weathered rocks, this foot end hard-carried G-01's dead weight and ran for nearly two hundred kilometers, yet its status was still usable.

In contrast, that Real World domestic POM foot end used in the Hengtai blind test had run for less than twenty kilometers under simulated working conditions, and a two-millimeter deep groove had already been scraped on the outside, which could be said to be a total collapse.

The generational gap of material formulations was thus laid bare on the table.

But POM-CF-07 had a problem: its pedigree was too wild.

This formula was concocted by Jiang Lin with his own hands in the Wasteland Stone House, like an alchemist.

The carbon fiber powder inside was cut off alive with an angle grinder by the mechanical platform at his Seventh Canopy Station, and then pulverized with a grinder.

As for the toughening agent inside, it was a product obtained by purifying the sap of a certain mutant plant unique to the Wasteland, adding polyurethane extracted from abandoned plastic bottles, and undergoing a crude chemical modification reaction.

This formula full of Wasteland cyberpunk style could run here because Jiang Lin himself controlled every tiny variable from raw material acquisition to hot-press molding.

He knew which furnace had a temperature two degrees higher, which batch of carbon fiber particles was a few micrometers thicker, and could rely on his feel to compensate during processing.

However, once returning to the Real World, no regular supplier could or would be willing to directly reproduce the chemical composition, mixing ratio, molecular weight distribution, and even molding process parameters of this formula.

Chemical plants in the Real World could not possibly analyze the molecular formula of a mutant plant sap.

What he needed to do now was technological translation.

Reverse-derive the macroscopic mechanical characteristics exhibited by these uncopiable handmade formulas in the Wasteland, and translate them into a Materials Engineering Specification that polymer material suppliers in the Real World can understand, execute, and comply with national or American standards.

Jiang Lin put the POM-CF-07 back onto the spare parts rack, pulled a notebook out of his pocket, and flipped to a new page.

[G-01 Foot-End Material: Wasteland Engineering Verification Results and Real World Mapped Transformation]

1. POM-CF-07 (Wasteland Version Carbon Fiber Modified POM): Comprehensive performance baseline.

The wear-resistant volume, falling weight impact work, and bending fatigue life are all significantly superior to commercially available domestic parts and Japanese standard parts.

Task: Convert the macroscopic physical parameters of this formula (hardness, tensile modulus, notched impact strength) into a target performance package while stripping away Wasteland-specific components.

2. Pure POM: The friction coefficient is too high, and the wear rate is large.

It is only suitable for indoor flat laboratory environments or low-strength regular terrain.

It is strictly prohibited for use in Hengtai mines and similar unstructured high-friction scenarios.

3. Glass Fiber Reinforced POM: Stiffness meets the standard, but brittleness is at a disaster level.

The microscopic fracture rate at the edges of gravel terrain exceeds 80%.

Eliminate from the candidate pool.

4. Ultra-High Molecular Weight Polyethylene (UHMWPE): Once used as a candidate.

Wear resistance is extremely excellent, but the elastic modulus and compressive stiffness are severely insufficient.

Under the dynamic heavy load of G-01, the foot-end compressive deformation is excessive.

This macroscopic deformation will directly contaminate the tiny tactile feedback signal of the foot-end contact force sensor, causing distortion in the WBC state machine solution.

In scenarios requiring precise contact force closed-loop, it is absolutely inapplicable.

Real World Constraint Analysis: There are many suppliers in China with the capability of carbon fiber / aramid fiber modified blending and pelletization, and the technical threshold is not high.

The real dilemma faced by the Low Entropy Workshop lies in quality control and discourse power.

What we need are industrial-grade materials with inter-batch performance fluctuations of less than 5%, complete traceability, the ability to withstand dynamic impact wear from high-humidity cinder and gravel, and capable of providing SGS or equivalent third-party testing reports.

Due to the extremely small initial order volume, major domestic factories are unwilling to take orders for prototyping, and the quality control of small factories cannot be trusted.

Engineering Transformation Strategy: Plan A — Formulate a strict target performance parameter table, look for medium-sized new material enterprises in China with research and development willingness, and bypass the minimum order quantity threshold in the name of joint research and development.

Plan B — Compromise with Japanese suppliers, adopt their ready-made high-impact modified POM models, bear the high sample fees and long delivery periods in the early stage, and ensure the offline release of the first batch of engineering prototypes.

The deduction of the material part ended, and the second hurdle was the most expensive and fragile heart in the electromechanical system.

The harmonic reducer.

After the Hengtai blind test, Xu Man disassembled and maintained the Real World version of G-01 in the laboratory.

When she removed the joint motor of the left-middle leg, she checked the output shaft of the reducer with a dial indicator.

The result was shocking: the radial runout jumped directly from the factory standard of 0.03 millimeters to 0.08 millimeters.

Subsequent internal industrial CT scans showed that a network of microscopic cracks invisible to the naked eye had already appeared in the cup-bottom transition area and tooth root portion of the flexspline, the core component of the harmonic reducer.

The log file of the data recorder revealed the reason.

In the third round of the blind test, when G-01 was retreating urgently through the culvert terrain, due to the instantaneous lag of the center-of-gravity algorithm under extreme terrain, the single left-middle leg instantaneously bore a peak impact force exceeding the designed rated load by about 30%.

Xu Man was holding the report at that time, frowning as she asked him: "No matter how strong our control algorithm is, it cannot make up for the physical limits. Should we consider switching to imported reducers?"

Imported Harmonic Drive, or specific high-end Leaderdrive models from the domestic first tier, have better materials and more precise tooth profiles, which can of course immediately alleviate the problems of precision loss and life collapse.

But what lay before the Low Entropy Workshop was an economic account that could not be balanced.

The unit price of these high-end goods is extremely high. A hexapod robot has eighteen or even more active joints. If all of them use high-end reducers, the cost will skyrocket.

More crucially, the delivery cycle of imported brands is as long as eight to twelve weeks, and they have extremely rigid hard requirements for the minimum order quantity of batch orders.

If G-01 is to enter small-batch trial production in the future, before obtaining huge financing, the cost of the reducer alone will directly eat up nearly a third of the entire machine BOM cost.

Jiang Lin woke up the workstation screen and opened the 3D CAD software SolidWorks.

He called up the complete assembly model of the harmonic reducer at the G-01 joint.

Dragging the mouse, he exploded and expanded the three core sub-parts that determined the fate of the reducer—the flexspline, circular spline, and wave generator—one by one, and zoomed in to the sectional view.

Then, importing the previous finite element analysis results, a color contour map of the contact stress distribution of the flexspline tooth surface under stress instantly appeared on the screen.

The dazzling red high-stress area was concentrated at the tooth root of the flexspline and the circular arc transition section between the cup body and the bottom.

Under extreme dynamic conditions of 30% overload, the wave generator forced the flexspline to undergo elliptical deformation. During the process of the flexspline gear ring meshing and disengaging with the circular spline, the tooth root endured severe alternating bending stress.

The contour map showed that the stress peak at this location had broken through the fatigue limit of ordinary domestic bearing steel materials.

This was not an algorithmic problem that Xu Man or Chen Yan could solve, nor was it a simple structural dimension design problem; this was the most fundamental problem of materials science and precision manufacturing technology.

The flexspline of a harmonic reducer is a contradictory body walking a tightrope between deformation and load-bearing.

It needs to maintain tens of thousands of hours of fatigue life without fracture under thousands of times of alternating bending stress per minute.

This has almost perverted requirements for material purity, heat treatment technology, and tooth surface treatment technology.

In reality, low-and-mid-range domestic flexspline models, in order to suppress costs, usually use conventional bearing steel or ordinary furnace batch materials. Material purity, inclusion control, and heat treatment consistency cannot be compared with high-end flexspline dedicated steel.

The heat treatment process tends to be conservative to avoid scrapping caused by excessive deformation, and the surface treatment process often fails to meet the nanoscale roughness requirements, resulting in poor anti-fatigue micro-pitting capabilities.

In the early engineering exploration stage of high-end legged robots, many top teams have encountered similar problems.

Their ultimate solution is often not simply and crudely throwing money to switch to more expensive vacuum remelted steel, but starting from the source of mechanical principles, through tooth profile topological correction, heat treatment curve optimization, and reconstructing and filtering the joint load spectrum at the control level, to actively reduce the stress concentration coefficient at the tooth root of the flexspline.

Jiang Lin gazed at those red high-stress patches on the screen.

The mathematical model of the double circular arc tooth profile was exceptionally clear in his mind.

Back then, in order to replicate the non-periodic hexapod mobile platform, he manually deduced all the meshing interference equations and curvature interference determination conditions of involute gears, cycloidal gears, and double circular arc gears optimized specifically for harmonic drives from beginning to end.

Theoretically, the double circular arc tooth profile allows the tooth profile to present a specific circular arc shape in the working section, which can achieve surface contact instead of line contact, greatly increasing the contact area, thereby exponentially reducing the tooth root bending stress.

However, the distance between a graceful curve on a mathematical drawing and a piece of steel in a factory workshop is separated by an entire bottomless chasm of craftsmanship.

The tooth profile machining of the double circular arc flexspline requires expensive high-precision five-axis CNC gear grinding machines produced in Switzerland or Germany within the real industrial system.

Not only that, but the dressing curve of the diamond grinding wheel used for grinding gears is by no means a standard circular arc in high school geometry, but a composite curve spliced together by two different radii of curvature, or even containing a very small transition straight line.

If one attempts to use an ordinary grinding wheel dresser in the Wasteland processing room for manual or simple numerical control dressing, its machining precision cannot even touch the threshold of the micron-level or even sub-micron-level tooth profile tolerance required by harmonic reducers.

Jiang Lin did not have a five-axis CNC gear grinding machine.

Therefore, from the day he deduced the equations, he had never held any illusions, nor did he intend to manually whip up a flexspline in the Stone House of the Wasteland that could reach the performance standards of commercially available commercial harmonic reducers.

His goal was very clear and very pragmatic.

Make a rough sample that can verify whether the direction of the stress-reducing tooth profile variant is correct.

Precision is insufficient, and gear meshing has interference?

Then lower the input speed of the test rig so that it does not jam due to overheating.

Life is insufficient, and there are many impurities that make it easy to break?

Then use the endless time of the Wasteland to stack up the test duration, using the sample size to offset individual differences.

The tooth surface machining is rough and covered with tool marks?

Then through long-term running-in tests, collect the wear debris data before and after running-in, inversely deduce the source of errors, and correct the theoretical model.

More importantly, he held a trump card in his hand that engineers in the Real World absolutely do not possess—MPS.

Jiang Lin switched out of SolidWorks and opened the interface of MPS-Kernel, the core toolchain built by him single-handedly on the workstation.

He created a new project process.

[G01_Harmonic_Flexspline_Tooth_Profile_Optimization (G-01 Harmonic Flexspline Tooth Profile Degraded Optimization Subroutine)]

He did not let MPS directly generate a set of standard double circular arc tooth profile machining codes.

A standard double circular arc tooth profile requires five-axis linkage gear grinding equipment and a high-precision grinding wheel dressing system, conditions he simply did not possess in the simple processing room of the Wasteland.

But he could utilize MPS's powerful search and heuristic algorithm framework to compromise on an optimal tooth profile parameter under the existing manufacturing process constraints.

Processing Equipment Constraints: The motion radius range of the grinding wheel dressing mechanism is set to 15mm-50mm; the maximum curvature change rate of the machinable tooth profile is limited within the hysteresis bandwidth that the existing stepper motor can respond to.

Material Property Constraints: Input the stress-life fatigue limit curve data of the batch of Wasteland high-carbon chromium bearing steel he tested earlier.

Working Condition Boundaries: Set the input torque spectrum, and forcibly add pulse-type extreme impact working conditions exceeding the rated load by 30%.

Solving Objectives: Under the premise of ensuring that the meshing contact stress does not exceed the material's allowable yield strength and the lower limit of the system's overall safety factor is not lower than 1.2, search for a variant tooth profile that is absolutely machinable on the existing simple lathe + homemade tooling, and can maximize the improvement of the flexspline's fatigue life.

In orthodox mechanical engineering textbooks, this belongs to unorthodox, wild methods.

No formal main engine factory would forcibly distort the optimal theoretical physical model just to accommodate lagging processing equipment.

But this is precisely a set of dimensionality-reduction attack optimization solutions tailor-made by the Low Entropy Workshop for survival, catering to the low manufacturing capabilities of the Wasteland.

Pressing the enter key, the candidate generator module inside MPS-Kernel began to operate at high speed.

In the terminal window on the right side of the screen, groups of candidate tooth profile geometric parameters began to scroll like a waterfall.

Every time a set of parameters with tiny distortions was generated, MPS would automatically call a simplified nonlinear finite element solver hard-core replicated by Jiang Lin based on the fluid mechanics and solid mechanics materials remaining in the Wasteland, to conduct a round of rapid contact stress coupling simulation in the background.

The simulation results were instantly fed back to the filter module.

Variants where the stress peak causes the safety factor to be lower than 1.2?

Directly eliminate and release memory.

Variants that meet safety conditions?

Retain and enter the memory pool to proceed to the next round of finer simulation with denser grids.

During this process, Jiang Lin deeply felt the worlds of difference between the Wasteland and reality in terms of R&D paradigms.

In the Real World, due to computing power resources, high trial-and-error costs, and lengthy supply chain cycles, this nearly exhaustive physical trial and error is simply impossible to do.

It is not that it is impossible algorithmically, but that the commercial schedule does not allow it.

A complete tooth profile optimization iteration—from CAD modification, CAE simulation, drawing issuance, mold or grinding wheel customization, precision machining of sample parts, surface treatment, and assembly, to final bench durability testing—takes at least several months in the Real World, even with expedited advancement.

If the first round of several hundred hours of durability testing ends with the fracture of the flexspline, engineers have to perform electron microscopy analysis on the fracture surface, modify the parameters, and run the simulation and trial production all over again.

This is another painful cycle measured in months.

This is why a new model of harmonic reducer in the Real World usually takes a long and arduous two to three years from preliminary project research to final mass production finalization.

But in the Wasteland, time lost its original scale.

He could directly cut a set of sample parts with obvious machining marks in the processing room of the Stone House, using the lathe and milling machine he had cobbled together, spending several days and nights.

Then, he would crudely install it onto the joints of the G-01 and drag it outside the Stone House to the non-periodic gait testing ground filled with weathered rocks, deep pits, and soft sandy soil, running it day and night for dozens or even hundreds of hours.

Listening to the abnormal noises when the gears meshed and watching the fluctuations of the current curves.

After the run, he would directly dismantle it and place it under a magnifying lamp to observe the shift of contact patches and the depth of wear patterns.

After obtaining real Physics feedback, he would immediately return to the workstation, modify the constraint parameters of the MPS, and then walk over to the lathe to make the next set.

The closed-loop speed was breathtakingly fast.

This was the greatest gift the Wasteland gave him.

The privilege of infinite iteration.

Before long, the MPS output the first set of recommended parameters after tens of thousands of internal iterations.

This was a heavily distorted double-circular-arc tooth profile variant.

The curve comparison chart on the screen showed that the MPS abruptly increased the curvature radius of the addendum arc by 2.3%, and counterintuitively elongated the transition section of the root arc downward by 1.5 millimeters.

The price of this visually somewhat clumsy modification was obvious.

The theoretical meshing contact ratio dropped precipitously from 1.8 of the standard double-circular-arc tooth profile down to 1.65.

This meant that during the transmission process, the number of pairs of gear teeth subjected to stress simultaneously decreased.

The transmission stability of the reducer during high-speed operation would inevitably experience a visible decline, and the NVH performance of the joints would worsen—meaning it would run noisier and vibrate more.

Yet the core benefit gained in exchange for this seemingly terrible set of parameters made Jiang Lin's eyes light up.

The simulation contour map showed that after sacrificing some stability, the fatal stress concentration factor at the root of the flexspline miraculously dropped by nearly 15%.

For the ordinary bearing steel material used in domestic mid-to-low-end flexsplines, near the critical point of the fatigue limit, even a mere 5% reduction in stress concentration could cause its corresponding macroscopic fatigue life to soar exponentially.

A 15% drop in stress meant that even if the material purity was subpar, its service life could be forcibly extended several times over.

Without any hesitation, Jiang Lin immediately saved the parameters, turned around, and walked into the processing room.

For nearly a week following that, harsh metal-cutting sounds and the screeching of grinding wheels continuously echoed from behind the Stone House.

Jiang Lin began making the first trial flexspline.

For the material, he selected a piece of high-carbon chromium bearing steel picked up from the Wasteland, similar to GCr15 in the Real World.

The internal structure of this steel ingot was dense, and it definitely counted as a high-tier scarce resource in the Wasteland.

He had stored and preserved it for several years, never willing to use it.

Now was the time for it to prove its value.

The machining process could be called a hand-to-hand combat with crude equipment.

Wearing protective goggles and relying on a self-made template device, he repeatedly fine-tuned the angle of the lathe tool, meticulously cutting out the cicada-wing-thin cup-shaped structure of the flexspline.

The most critical heat treatment step was completely a high-stakes gamble.

He did not have a vacuum quenching furnace.

The heat treatment relied entirely on that self-made resistance furnace.

The temperature gradient inside the furnace chamber was not very uniform.

To control the furnace temperature, he could only manually turn a huge voltage regulator knob, keeping his eyes fixed on the jumping numbers transmitted back by two industrial thermocouples.

When the steel was heated to a blazing cherry-red color, he clamped it out with iron tongs and plunged it into an iron drum filled with engine oil at maximum speed for quenching.

The subsequent tempering process was even more difficult.

For the tempering temperature and holding time, there were no manuals to refer to; it relied entirely on the empirical curve he had figured out bit by bit using muscle memory and intuition after destroying countless cutting tools and parts in the Wasteland.

After a long cooling period, this handmade flexspline with oxidation colors was finally placed on the workbench.

If this flexspline were placed in the cleanroom of any major precision gear manufacturer in the Real World, it would probably be thrown directly into the scrap bin by any quality inspection engineer without hesitation.

On its surface, tiny chatter marks left by the lathe feed could even be seen with the naked eye, and the end face was not flat enough either.

But Jiang Lin did not care at all whether it looked good.

He only cared about how long it could carry that heavy body inside the knee joint of the G-01 and keep running.

He smeared grease on the tooth surfaces, installed it into the harmonic reducer of the left-middle leg of the G-01, and tightened the flange bolts of the cross roller bearing with a constant torque wrench.

Powered it on, mounted a dial indicator, and re-calibrated the initial radial runout of the output shaft.

The indicator needle finally stopped at a number.

0.04 millimeters.

The data was slightly worse than the 0.03 millimeters of the initial factory state of original domestic ordinary parts, which was the natural disadvantage of manual machining precision.

But it was far better than the 0.08 millimeters of the scrapped state after Hengtai's blind test overload.

It was time to start running the long and tedious hybrid bench and field limit tests.

He set a rigorous automatic loop route for the G-01.

Regular concrete slab roads, simulated terrain filled with random potholes, non-repetitive irregular obstacle stairs, gravel slopes with an inclination of up to 35 degrees, loose sandy soil prone to foot slip, and slippery slopes splashed with waste oil and water.

He made the G-01 act like an indefatigable ascetic monk, falling, getting back up, climbing, and slipping backward on this site day after day.

Every time it reached the set phased mileage, he had to completely dismantle the reducer of the left-middle leg, wash the oil stains off the flexspline in a cleaning agent, place it under a high-magnification magnifying lamp, and carefully inspect the distribution of contact patches on the tooth surfaces and the trend of the wear patterns.

The first month.

The tooth surface contact patches of the flexspline showed a very healthy elliptical distribution, with no eccentric load phenomenon.

The root area was intact, with no signs of stress whitening whatsoever.

The stress-reduction effect of the variant tooth profile initially manifested.

The second month.

Due to rough surface machining, a tiny initial wear mark appeared on the root transition section.

Jiang Lin resisted the urge to replace the spare part, recorded the data, reassembled it, and let it continue running.

Because this kind of initial wear mark was often a normal phenomenon for materials to eliminate local high points during the running-in period.

By the third month, a miracle appeared.

The depth and width of that wear mark remained stagnant and extremely stable.

The most vexing fatigue micro-cracks still showed no signs of appearing.

This piece of bearing steel quenched with waste oil demonstrated astonishing toughness.

It kept running until the end of the sixth month.

This flexspline, born from a crude resistance furnace and a manual lathe, had accumulatively withstood over three hundred hours of dynamic load operation driven by the G-01's violent non-periodic gait.

When Jiang Lin dismantled it for the last time, performed a thorough surface cleaning on the workbench, and conducted non-destructive testing using a fluorescent magnetic particle flaw detector.

Under the ultraviolet light, the surface of the flexspline was clean, without a single fatal fluorescent line representing cracks.

Due to long-term running-in, the contact patches on the tooth surfaces had become as smooth and uniform as a mirror.

Measured again with the dial indicator, its meshing backlash had increased by less than 0.01 millimeters compared to the initial state, completely within the compensation redundancy range of the control algorithm.

He succeeded.

In an environment without top-tier equipment and top-tier materials, he had found a low-level optimization route for reducers that could work under the manufacturing constraints of the Wasteland.

This route could proudly announce that, at least in the first batch of engineering prototype stages of the G-01, the Low Entropy Workshop would not have to be completely choked by the delivery times and quotes of imported high-end reducers.

Sitting in front of the workstation, Jiang Lin created a new document and poured all his hard work of these six months into it.

[G01_Harmonic_Flexspline_Fabrication_Process_v1]

This was not just a document, but a martial arts cheat sheet brought back from the Wasteland to the Real World.

It detailedly recorded the material substitution attributes and spectrum analysis results of the Wasteland bearing steel.

The cooling rate curve of the quenching liquid, the specific values of the tempering temperature steps, and the holding time.

The mechanical structure diagram of the grinding wheel dressing tooling and the quantitative description of the operation methods.

The complete DXF coordinate system parameter file of the "Double-Circular-Arc Variant Tooth Profile" optimized by the MPS.

All mesh division boundary condition settings for contact stress and thermal-coupled simulation.

The complete time-series database of current, runout, and wear volume from the 300-hour high-intensity bench and field comprehensive test.

This document contained too many imprints of the Wasteland and crazy actions that defied Real World commercial common sense, but he only needed to extract the core conclusions from it.

"Adopting a double-circular-arc variant tooth profile scaled up by a specific ratio can reduce the peak stress concentration at the tooth root by 15% at the cost of a small reduction in the contact ratio. Under this topological structure, using domestic conventional high-carbon chromium bearing steel can complete over 300 hours of crack-free durability verification under extreme impact working conditions."

This conclusion could be directly transformed into rows of rigorous data, seamlessly integrated into the "Mass-Production Version Reducer Technical Specification" that the Low Entropy Workshop was about to issue to domestic machining plants.

By then, suppliers in the Real World would only need to program gear grinding according to the coordinate parameters on the specification and inspect acceptance according to the specified heat treatment hardness.

They would never need to know that this set of parameters—which saved them from material scrap and greatly improved the yield rate—was crafted by a lonely engineer in the gale-force winds of the Wasteland using a handmade flexspline and time.

At this point, the core cost-reduction route for the reducer was verified.

Solution: Double-circular-arc variant tooth profile topological optimization + 15% reduction in tooth root stress concentration.

Machining constraint requirements: The grinding wheel dressing curve must be strictly executed according to the optimized parameter DXF. It is estimated that at least two to three medium-sized gear factories in China currently possess this five-axis machining capability, and their production capacity is not saturated.

Material strategy: During the initial trial production and prototype stages, the flexspline material can temporarily use domestic conventional high-carbon chromium bearing steel directly (extremely significant cost reduction); subsequent to large-scale mass production or entering high-premium military/special industries, it can be smoothly upgraded to imported vacuum remelted steel without changing the structure to enhance absolute lifespan.

Wasteland bench and field verification duration: 300h+ (good condition, no fatigue failure observed).

With two heavy burdens unloaded, only the last hard nut to crack remained.

The lateral shear deviation of the foot-end sensor.

The cross-verification report of sensor data and state machine produced by Chen Yan after staying up for several nights in the Real World pointed directly to the soft spot of the G-01 control system and provided an extremely crucial quantitative conclusion.

The proud foot-end state machine of the G-01 had an extremely high consistency of over 99.7% between its contact force feedback signal and the calibration results of the laboratory's precision six-axis force sensor on regular surfaces with uniform roughness such as concrete floors and asphalt roads.

However, once the G-01 stepped onto irregular surfaces filled with wet mud, cinder, and sloping gravel like the Hengtai mine, the contact force reading data would exhibit a systematic positive deviation of about 3%.

3% sounded negligible, but in a dynamic balance system, this 3% error would be transmitted layer by layer through contact force distribution, support phase judgment, and joint torque inverse solution, ultimately leading the machine body to misjudge the ground load-bearing capacity.

Chen Yan pointed out in the report that the Physics root cause of the deviation was preliminarily determined to be the lateral shear micro-slip of the outermost protective layer of the sensor.

When the foot tip tilts and steps onto a slippery or loose contact surface and begins to apply normal force, a microscopic relative sliding—invisible to the naked eye yet truly existent—occurs between the protective layer material and the surface of the internal thin-film pressure sensor due to shear stress.

This micro-slip causes distortion in the force-bearing area and force vector of the sensor, which in turn outputs an erroneous voltage signal.

The engineering advice given by Chen Yan at the time was very straightforward.

Either find a thinner, harder material with an extremely small coefficient of friction to replace the existing protective layer and eliminate slippage at the Physics level.

Or, add an extra layer of lateral shear non-linear compensation algorithm in the control loop of the underlying state machine.

Jiang Lin walked to the other side of the spare parts wall and forcefully pulled out a dusty plastic storage box from the very bottom.

Inside were the thin-film pressure sensors, PVDF piezoelectric films, and matching protective layer material samples he had accumulated.

He cleared a space on the workbench, spread out the materials from the box one by one, and picked out the three most potential candidate materials from them.

First: Polyimide film.

It is extremely thin, only 0.025 millimeters. This material is widely used in the aerospace field, boasting high hardness, high temperature resistance, and excellent wear resistance.

Second: Ultra-thin stainless steel foil.

Thickness is 0.01 millimeters. Its hardness is impeccable, and its shear resistance is extremely strong, but its drawback is poor flexibility. Once macroscopic bending occurs, it is extremely prone to plastic deformation.

Third: A composite coating that Jiang Lin concocted himself through alchemy in the Wasteland.

It was formulated through chemical cross-linking using a polymer waxy substance extracted from the surface of radiation-resistant plants in the Wasteland, combined with epoxy resin recycled from waste circuit boards.

The thickness of this coating can be adjusted arbitrarily through the number of sprays, and its flexibility is excellent, but its wear resistance is completely an unknown.

For the next two solid months, Jiang Lin cut and backed each of the three materials, adhered them to the surface of the sensors, and then installed them on the right front foot of G-01 used for testing. On the set test route, he conducted high-intensity comparative tests on regular and irregular surfaces one by one.

The performance of the polyimide film showed extreme polarization.

On regular surfaces such as the flat mud and steel plates in the laboratory, its performance was nothing short of stunning. The hard surface greatly suppressed shear slippage, and the deviation of the state machine directly dropped to within 0.5%.

However, on irregular surfaces, when the foot tip of G-01 stepped on a weathered gravel with sharp edges in the Wasteland, although the hardness of the film itself was sufficient, as a polymer material, its bending stiffness still could not compete with metal.

Under the local ultra-high pressure generated by sharp objects, the film would produce minute elastic depressions.

This depression caused the contact area of the conductive particles inside the sensor to increase abnormally, making the contact force data read by the sensor abnormally large.

In short, it did not slip, but it was poked out of shape.

Taken together, the system deviation on irregular surfaces stubbornly lingered around 2%.

The Physics characteristics of the stainless steel foil were extremely fierce.

The sufficiently thin thickness and the rigidity of the metal allowed both micro-slippage and local elastic depressions to be well suppressed, whether on flat ground or gravel.

Test data showed that the overall deviation was brutally reduced directly to a staggering 0.8%.

If looking only at the data, this was the optimal solution.

But engineering practice immediately slapped Jiang Lin hard across the face.

Its flexibility was too poor.

During the dynamic process where G-01 took high-frequency steps and the rubber pad at the foot tip inevitably underwent macroscopic bending, the stainless steel foil experienced alternating bending stress.

Having been installed and run for less than fifty hours, Jiang Lin found under the microscope that dense metal fatigue cracks had already appeared on the edges of the foil, and some areas even began to peel off.

It was extremely easy to damage the sensor body itself, and it was completely unsuitable for long-term use as a consumable under complex working conditions.

The last one tested was the Wasteland-homemade composite coating.

Jiang Lin adjusted the thickness of the coating back and forth four times. After experiencing a few failures where early wear wiped it out completely, he finally locked the thickness into an extremely narrow process range.

About 0.08 millimeters thick.

At this thickness, the surface hardness of the coating was between polyimide and stainless steel, enabling it to effectively resist local depressions.

Thanks to the flexible segments of the plant waxy substance, its flexibility was far better than that of stainless steel, and it would not crack from brittleness due to the bending of the foot tip.

After running continuously for over a hundred hours of field road conditions, Jiang Lin extracted the background log file.

The deviation value stably lingered around 1.2%.

Although it could not compare to the stunning 0.8% of the stainless steel foil, the durability performance of the coating was excellent.

More importantly, it possessed incredible restorability.

If the coating was locally worn, it only needs to be re-sprayed with a spray can on-site. After standing to cure for half an hour, it could recover as good as new.

Jiang Lin gazed at the three data curves on the screen, organized the test data of the three materials into a table, and wrote it into his notebook.

[Foot Tip Sensor Protective Layer—Physics Material Candidate Comparison and Engineering Choices]

Scheme A (Polyimide film): Regular surface deviation < 0.5%, irregular surface deviation about 2%. Limitations: Under local high points and high pressure, the film undergoes microscopic elastic depression, leading to distortion in the calculation of the sensor contact area (too large).

Scheme B (Stainless steel foil): Lowest contact force analysis deviation, full operating conditions < 0.8%. Limitations: Metal fatigue threshold is too low, flexibility is extremely poor, fatigue cracks are extremely prone to occur at the edges during dynamic walking, leading to overall tearing. Maintenance cost is extremely high, and it will not be adopted in engineering.

Scheme C (Wasteland-homemade composite coating): Comprehensive deviation under full operating conditions is about 1.2%. Excellent flexibility, no fatigue cracks seen, good durability, and possesses on-site rapid spray repair characteristics. Fatal limitations: The coating cross-linking and curing process is difficult to standardize outside a cleanroom, and its unique plant wax-based formula is difficult to directly and massively reproduce within the current supplier system of the Real World.

Comprehensive engineering recommendations—

First mass-production version of G-01 in the Real World: Must adopt [Scheme A (Polyimide film)] as the Physics substrate, while introducing the [lateral shear software compensation algorithm] in the underlying driver to smooth out that 2% irregular surface deviation.

[Scheme C (Coating)] is retained as a backup for phase II pre-research technology. After obtaining funds later and having the ability to lead chemical material partners to intervene, the feasibility of its industrial mass production will be evaluated again.

Just like the problems of the first item, foot tip materials, and the second item, reducers, the optimal solution in engineering has never been the theoretically most flawless one.

Rather, it is the compromise solution among all parties under technological ideals, the upper limit of Wasteland materials, Real World supply chain barriers, and the dual constraints of strict cost and time, which has the minimum cost and can be truly implemented and executed.

Polyimide film has a 2% error?

Since Physics materials have limits, use algorithms—a virtual material—to make up for it.

Stainless steel foil has superb Physics performance but will crack and have a short life?

Then abandon it on the engineering route and leave no future troubles.

The Wasteland-customized coating works well in the Wasteland, but suppliers in the Real World cannot figure it out?

Then seal it away, leaving it for a future with stronger resource mobilization capabilities.

The essence of engineering is not blindly pursuing a full-score answer on the blueprint, but finding the compromise solution among countless flawed answers that allows the machine to truly live as it walks off the production line and continue operating in harsh environments.

Once the direction was clear, what followed was pure mathematics and code time.

He spent two full weeks deriving the mathematical model of the lateral shear compensation algorithm specialized in treating micro-slippage, and organized it into a tightly structured independent technical document.

The ingenuity of operator algorithms lies in the fact that it does not attempt to change the fact of the Physics deformation of the polyimide film.

Instead, it utilizes the contact angle data fed back in real time by the six-axis sensor at the foot tip, the instantaneous load reverse-calculated by the joint motor through the current loop in the current posture, the high-frequency micro-slip vibration signal captured by the body inertial measurement unit, and the ground friction coefficient interval estimated in real time within an extremely short time window.

These four dimensions of data are input into a non-linear disturbance observer to reversely and in real time calculate the possible slippage amount and depression degree occurring in the protective layer in the shear direction.

Finally, this calculated slip variable is quietly added into the contact force judgment threshold of the state machine as a dynamic bias parameter.

This algorithm does not change the original Physics voltage readings uploaded by the sensor; it only changes the confidence interval for the core state machine to judge whether this step is truly stepped on solidly and whether the next step can exert force to kick off the ground.

This is not about burying one's head in the sand and fixing erroneous readings into correct ones, but letting the brain of G-01 clearly know how much moisture is contained in this reading reported by the underlying sensor under the current contact angle, material, and friction conditions, and how much discount should be applied to it as a basis for exerting force.

After the code was written, it was burned into the main control chip of G-01.

It was brought back to the gravel slopes of the Wasteland for field testing.

The algorithm demonstrated breathtaking dominance in the high-intensity dynamic tests in the Wasteland.

By utilizing mathematical compensation, it forcibly compressed the otherwise unacceptable 2% Physics deviation of the polyimide film to within 0.6%.

This precision was already superior to the 0.8% deviation performance given by the stainless steel foil under current test conditions.

At this point, the engineering implementation and verification of the three core subsystems were all declared complete.

Jiang Lin packed all the test data tables, SolidWorks assembly drawings, MPS-generated tooth profile DXF parameter files, heat treatment temperature curves, and that extremely precious compensation algorithm C++ source code scattered across the workstation into an encrypted super folder.

Renamed: [G01_Production_Engineering_v1.0_Final].

This was not a ready-made file that could be thrown directly without changes to Chen Yan or Xu Man.

It contained too many raw data remnants whose sources could not be explained.

But this was a technical master template sufficient for Jiang Lin to filter and translate in his mind into Real World supplier bidding specifications, quality inspection acceptance standards, and underlying algorithm testing task books.

With it, the Low Entropy Workshop would no longer be a makeshift team relying only on creativity and prototypes in reality, but a hardcore technology company truly mastering core cost-reduction manufacturing processes.

After the closed-loop of the mass-production engineering solution, his long-tensed nerves finally reached their limit.

Jiang Lin forcibly gave himself a two-day vacation.

Holding a hoe ground from waste steel plates, he weeded the perimeter jujube trees and amorpha fruticosa.

Or climbing onto the roof of the Stone House, carrying a bucket of filtered neutral water, he used a rag to dust that array of solar photovoltaic panels.

When mental labor was excessive, this nearly mechanical physical labor turned into psychological massage instead.

He didn't need to think about anything in his head; he just needed to stare at a blade of grass, a piece of soil, and a withered Moss blown over by the wind right in front of him.

Two days later, the consumption of physical strength made his spirit sharp again.

Jiang Lin sat back in front of the workstation and opened [G-Variants (Multi-Scenario Variant Pre-research)].

The mass-production version could only keep G-01 alive.

Whereas the variant architecture determined how many kinds of battlefields it could enter in the future.