Category: Engineering

Why isn’t ChatGPT inside our refrigerators?

by philg

Some years ago people envisioned a refrigerator that would track contents via RFID and alert a consumer to being low on milk or whatever. Making this a reality would have required cooperation among all of the companies that make packaged food (to add the RFID tags) so of course it never happened.

A human can inventory a fridge. Anything a human can do ChatGPT can do better, or so we’re told. If a fridge costs $15,000 (see Sub-Zero refrigerator with R600a owner’s review) why can’t it use a handful of inexpensive video cameras to look at everything going in and out in detail? It can make some good guesses about quantities, e.g., every time the eggs are removed there will be three fewer eggs remaining in the carton (refine this guess after some experience in a household as to when the carton stops being returned to the fridge (assume this means the egg count is zero)). The in-the-fridge AI could email with a list of expired stuff to throw out and a list of stuff to buy. It could email at 3 pm every day with a suggestion for what to cook for dinner given the ingredients present in the fridge, adding critical items via an Instacart order if approved.

“New AI-powered fridge technology generates recipes based on diet, food on its shelves” (Deplorable Fox) describes a Samsung fridge introduced at CES 2024, but it turns out to be not that smart:

The fridge’s technology also reportedly enables users to add expiration dates for items purchased, and the refrigerator will alert them once that expiration date is near.

Why is it the human’s job to read expiration dates off the packages? Why can’t the brilliant AI do that? Let’s give some credit to Samsung, though, for including an epic 32-inch TV on the $4500 fridge:

So the Samsung fridge is missing the Instacart ordering support, I think, as well as the automation of ferreting out expired food.

Full post, including comments

Could Elon Musk be right about reusing the truss steel that fell into Baltimore Harbor?

by philg

Elon speaks on a topic for which he has few apparent qualifications, the classic mark of a fool, April and otherwise:

An example of idiocy, as the “trust the experts” crowd says? The experts themselves disagree with Musk. A Florida-based bridge engineer:

(My comment on the above: Sell the old truss on eBay? “Dropped once. Never snapped.” (see also, re: World War II, “French Rifle for Sale: dropped once, never fired”.))

I wonder if both Elon Musk and the Florida engineer could be correct.

If you want to build something to last 50 years and have a big safety margin and you don’t care how long it takes to build, the civil engineer is right. What if it needs to last only 10 years (while an adjacent replacement is built using conventional techniques) and you’re willing to compromise on aesthetics? Maybe Elon is also right. Pick up and reuse as much of the old bridge as possible. Do a new structural analysis of the old design to see where doublers and other structural enhancements are needed given possible weakness of some of the elements. Instrument it with strain gauges everywhere. Then patch it up and let all of the self-driving truck companies pull heavy trailers onto it. Check the strain gauges. If everything is consistent with the analysis under a real-world load, open it to cars, but not heavy trucks. In the #AbundanceOfCaution department, maybe close the bridge if there are exceptional winds (1 or 2 days per year).

Given the fact that it is possible to drive around this bridge and the American hunger for perfect safety and security and the construction industry’s reluctance to do anything unconventional, my suspicion is that the Florida engineer is correct and Musk’s idea could never work in practice. But Musk’s idea might be a good one if there were more urgency regarding the rebuild, for example.

Full post, including comments

Key Bridge collapse and electric aircraft

by philg

Dali, the Singapore-flagged container ship that brought down Key Bridge in Baltimore, presumably had multiple redundant power systems, yet apparently suffered a total loss of power that may have contributed to the bridge strike:

Obviously this is a sad day for the families of those who perished in the collapse, but the accident raises a question regarding electric aircraft. Many of them can’t glide. Absent a Cirrus-style ballistic parachute, total loss of power means that everyone on board will die. The typical design has three independent electrical systems and the calculated chance of a total failure is 1 in 1 billion or less. Yet the same calculation was likely done for the Dali, a much more expensive machine, and total failure appears to have happened nonetheless.

I wonder if this worse-than-calculated performance favors winged designs such as Beta’s.

The “super drone”-style designs can’t autorotate as a helicopter can/must and they can’t glide on wings like an airplane because they don’t have wings.

Full post, including comments

History of failed attempts to build houses cheaper

by philg

Loyal readers may recall that one of my pet obsessions is why the manufacturing techniques that have made cars and widgets cheaper can’t be applied to housing. Why can’t, at least, the house have plug-in bathrooms, kitchens, and utility rooms so that all of these items can be refreshed cheaply with factory-built rooms after 20 years?

A side effect of our failure to come up with a way to build houses at a lower cost is the “affordable housing crisis” that advocates for population growth via low-skill immigration like to decry (see Immigration and rent are both at all-time highs).

“Why Do We Build Houses in the Same Way That We Did 125 Years Ago?” (New York Times; non-paywalled version) digs into this question:

In 1969, the federal government announced that it would hand out millions of dollars in subsidies to companies willing to try something new: build houses in factories.

It didn’t work. Big companies, including Alcoa and General Electric, designed new kinds of houses, and roughly 25,000 rolled out of factories over the following decade. But none of the new home builders long survived the end of federal subsidies in the mid-1970s.

Last year, only 2 percent of new single-family homes in the United States were built in factories. Two decades into the 21st century, nearly all U.S. homes are still built the old-fashioned way: one at a time, by hand. Completing a house took an average of 8.3 months in 2022, a month longer than it took to build a house of the same size back in 1971.

As with most innovations, the central planners believe that central planning (“government help”) is necessary:

The tantalizing potential of factory-built housing, also known as modular housing, continues to attract investors and entrepreneurs, including a start-up called Fading West that opened a factory in 2021 in the Colorado mountain town of Buena Vista. But Fading West, and similar start-ups in other parts of the country, need government help to drive a significant shift from handmade housing to factories. This time, there is reason to think it could work.

How much can be saved?

Fading West says houses from its factory can be completed in as little as half the time and at as little as 80 percent of the cost of equivalent handmade homes, in part because the site can be prepared while the structure is built in the factory. A 2017 analysis by the Terner Center for Housing Innovation at the University of California, Berkeley, found similar savings for the construction of three- to five-story apartment buildings using modular components.

If we adjust for the inevitable startup hype factor… the 80 percent is probably 115 percent of what a tract house developer spends when building 25-100 houses at a time and 95 percent of what it would cost to build one house via the traditional method.

What do people who don’t get government money for their factory-built house startup say?

Factory home builders have struggled to streamline construction. [Brian Potter, a senior infrastructure fellow at the Institute for Progress, a nonpartisan think tank focused on technological innovation] spent several years looking for ways to make housing construction more efficient, an effort he narrated on a fascinating blog, before concluding that significant progress wasn’t likely. “Almost any idea that you can think of for a way to build a single-family home cheaper has basically been tried, and there was probably a company that went bankrupt trying to do it,” Mr. Potter told me.

The depressing conclusion: If you believe in fairy tales, single-family houses could potentially come down in price by 15 percent (the land underneath won’t be reduced in cost by 20%!) as an absolute outer limit. If the American population is to grow, therefore, people are going to live in smaller and crummier houses unless they develop valuable work skills.

Full post, including comments

Single-stage versus variable-speed air conditioning dehumidification performance

by philg

After an exciting summer packed with three blower motor failures in three 6-year-old Trane single-speed air conditioning systems, the transformation of our house into a showcase for variable-speed communicating Trane/American Standard equipment is complete.

For background, see the folllowing:

The most humid part of our house was the upstairs. This contains two big bedrooms served by a 3-ton A/C for a calculated Manual J demand of 2.1 tons. Relative humidity was 58-62 percent with a TEM6 variable-speed air handler and a single-stage condenser.

Step 1 was replacing the condenser with a variable-speed “communicating” condenser that sends digital information back to the air handler over a two-conductor cable. Trane says that this new condenser is a match for the 6-year-old TEM6 so long as an adapter relay panel is installed. What they don’t say is that the result is a brain-dead system in which the air handler always runs at the same blower speed regardless of what the compressor speed is. Compared to the 6-year-old single-stage A/C, there was no reduction in humidity from this arrangement.

Step 2 was replacing the (working perfect with a new blower) TEM6 air handler with a top-of-the-line TAM9 air handler. Humidity immediately plummeted to a reasonable 51 percent on a wet hot Florida day with hours of rain, an 87-degree high, and humidity as high as 95 percent.

What does #Science say about this result? “Dehumidification performance of a variable speed heat pump and a single speed heat pump with and without dehumidification capabilities in a warm and humid climate” (Kone and Fumo 2020; Energy Reports):

the variable speed mode was able to maintain relative humidity between 50% to 52% on summer days. In the single-speed with enhanced dehumidification, a slightly less effective humidity control was achieved on summer days with the mode keeping the relative humidity between 53% to 55%. In the normal cooling mode, which resembles a conventional system, the humidity levels were controlled between 55% to 60%. In the shoulder season, the variable speed and enhanced dehumidification modes maintained the relative humidity between 55% to 58% and 53% to 56% respectively. In the shoulder season, the normal cooling mode kept the indoor relative humidity near or above 60%.

In going from single-stage to variable-speed, #Science found a reduction in humidity from an average of 57.5% to 51% (middle of the ranges given), or 6.5%. My data, consistent from a Govee sensor set and a $300 Airthings monitor, was 8-10% reduction in the relative humidity reading. The ground floor of the house still feels and measures less humid (40-50% depending on the location), but walking upstairs no longer feels like entering a steam room.

It’s tough to find objective data from anywhere else. Carrier is the only company, I think, that offers any numbers:

The Trane stuff has an emergency dehumidification capability in which it will run the heat strips as the same time as the A/C. Carrier also might have something like this (their commercial systems have a “reheat” mode that might do something similar, but using only the coil and not the resistive heat strips).

It is unclear from the Carrier page if they’re talking about using an extreme measure to dehumidify or just running the variable-speed in an optimized manner.

I’m also unclear what they mean by “400 percent more moisture” removed. If a single-stage system is removing 1 gallon of water, the variable-speed system removes 5 gallons when outside temp and thermostat temp are held constant? That doesn’t seem plausible. If it is hot and humid outside, the system has to remove a huge amount of water just to do its basic job (since cooling outside air will almost immediately result in 100% relative humidity and condensation).

If relative humidity is linear in the amount of water vapor, a properly sized single-stage system has already removed more than half the water that was originally present in the air (since cooling resulted in 100% relative humidity and the house ended up at 50% humidity). As great as Carrier may be (they’re headquartered only about two miles from our house here in Palm Beach County!), I don’t see how they can remove 5X the amount of water compared to a system that removes half of the water available.

(Why didn’t we get Carrier? We already had Trane gear and thought that we might be able to preserve at least some of it (we weren’t). Also, the Carrier dealer who came out to quote the project refused to deal with our house because of a splice where the wires exit the house near the condenser, claiming that their communication wouldn’t function properly.)

I can’t figure out why single-stage A/C continues to be the standard here in the U.S. Everyone in Asia has variable-speed equipment (all of the mini splits are variable-speed). Assuming a constant thermostat setting, a single-stage system is the correct size for only one outdoor temperature. Why wouldn’t people be willing to pay a little more for a system that can run at the correct speed for whatever temp Climate Change happens to dish out at any given hour on any given day? Is it that it is impossible to explain to consumers what a dumb idea single-stage A/C is? (Maybe it makes sense in Arizona, though, where there isn’t any humidity to begin with?)

Full post, including comments

Progress in electric bicycles?

by philg

Extremely loyal readers may remember that I previously reviewed a 2013 Trek electric bicycle:

  • 52 lbs. for XL frame size
  • $2100
  • 250 Wh battery
  • 250 watt motor

It’s been 10 years. Let’s check in to see how much better today’s electric bikes are. Behold, the Trek Verve+ 2:

How much better is this than the 2013 bike?

  • 51.5 lbs. for M frame size (i.e., heavier)
  • 2,850 Bidies (BLS says that $2,100 in 2013 is equivalent to roughly 2,800 Bidies today, so this is about the same when adjusted for official inflation)
  • 400 Wh battery
  • 250 watt motor

The 2023 bike should be better balanced, due to the battery being in the middle, and it has hydraulic brakes. On the other hand, if the battery dies, the old bike’s 21-speed drivetrain will likely be superior to the new design’s 9-speed (presumably lacks the low gears you’d want to pedal yourself and a ponderous electric bike back to the garage).

I’m shocked at how little progress has been made. I would have guessed that, at the $2100 price, the weight would have come down to 40 lbs. and the battery capacity would have doubled to 500 Wh. Maybe if we’d put $20 trillion into electric bike engineering instead of coronapanic lockdowns, payouts, subsidies, etc.? Or are the bike engineers running up against the laws of physics and chemistry?

From my 2015 review:

What about the new stuff? It seems as though the 900-lb. gorilla of the bike world, Shimano, has entered the market with the Shimano Steps system, which is what Trek is using on their latest models. This may prove the point of Crossing the Chasm (that the innovators often don’t end up as market leaders because products that appeal to hobbyists and early adopters don’t necessarily appeal to the mainstream).

My bike is a regular Trek city bike to which they added some Bionx components, much as a consumer might have done in his/her/zir/their garage. What happened to Bionx when Shimano and Bosch moved in? A 2018 article:

Electric-assist and retrofit electric motor company Bionx has gone bankrupt and its assets are being sold off.

After cornering the North American electric-assist retrofit market, Bionx suddenly closed its doors and laid off all workers in February 2018, just at the start of the busy Spring season in the bicycle industry.

Apparently, the financial failure of the company is related to a deal with General Motors, in which Bionx was to produce electric bicycles for the auto-maker at a cost of $1000/ea. After finding that the bicycles would actually cost $1400/ea to build, Bionx defaulted on the contract and went into receivership shortly thereafter.

In the Department of Never Take Investment Advice from Philip, this is what I thought would happen to Tesla. They fiddled around with standard Li-ion batteries and electric motors. As soon as they’d proven that the market existed, the companies that were experts at making great cars would swoop in and take away all of Tesla’s customers because the cars around the batteries/motors would be so much better. The UK’s Car magazine, however, recently did a comparison test and BMW’s i4 M50 was ranked #3, Hyundai’s Ioniq 6 was #2, and the best electric sedan was the Tesla 3. In other words, Tesla figured out how to make a good car in less time than it took BMW to figure out how to mount some batteries and a motor into a car.

Related:

Full post, including comments

A visit to BETA Technologies in Burlington, Vermont (eVTOL aircraft)

by philg

Earlier this month, I stopped into BETA Technologies, a $1 billion (financing) electric aircraft baby that has been growing in the unlikely crib of the Green Mountains. With offices and labs in Burlington (KBTV) and flight tests across the lake at Plattsburgh (KPBG), the company is pushing ahead on making all aspects of battery-electric aviation practical.

Much of the company’s effort seems to have gone into making better electric motors. Cooling is a challenge for a motor that puts out 200+ hp continuously and there have been multiple iterations of design. The 3D printers were all running when I visited while mechanical engineers labored at desktop PCs.

Does it fly? Yes! In fact, a test pilot told me about going more than 300 nm on one charge. The company is working on two aircraft at the same time:

The CTOL version on the left (“conventional takeoff and landing”) might be more interesting for the general aviation crowd. Why pay $1 million for a new piston-powered airplane that is trying to shake itself and you apart with vibration and deafen you and your passengers with noise when you can cruise in smooth quiet electric comfort? BETA is hoping for certification in 2025 (which means 2027?) and is also working on the ground support infrastructure to make these aircraft practical transportation solutions. Charging will supposedly take about one hour, which is inferior to refueling time, but my host posted out that electric aircraft don’t waste any time in startup/runup/shutdown. The company has a Pipistrel electric two-seater and he demonstrated that it is up and running within a few seconds after flipping four switches.

One area where BETA might have less certification challenges than competitors is that they’re not trying to create a fully autonomous aircraft. In the VTOL version, one of the four seats is for a pilot with a powered-lift type rating on his/her/zir/their certificate (maybe the CTOL version can be flown by a pilot with a single-engine land rating?). On the other hand, if a commercial operator orders 100 of these, the operating will have to fight United Airlines for 100 pilots.

Just outside their engineering hangar is an example of what the ground support station would look like. The left cube is a GPU that can be hooked up to run cabin heat or A/C. The center is for a massive charging cable to top up the 800V battery. The right cube is for cooling the battery (during charging).

The company has a “study hall” where local kids can come in to learn about how battery packs, inverters, and three-phase AC motors work.

There is also a non-motion sim right by the front door:

I came away impressed with the company’s spirit and cooperative energy.

What’s the competition? Boeing-owned Wisk had a booth and a demonstration flight at Oshkosh this year:

Considering that it has the same seating capacity as a Cessna 172, the Wisk machine is enormous. It the electric future is more efficient, why does the efficient vehicle take up four parking spaces? And dare anyone ask how much it will cost to put together this much carbon fiber and plastic?

The BETA eVTOL works like a DJI drone. The rotors are fixed, but may spin at different speeds. A pusher propeller at the rear can then push the machine to cruise at 120 knots or more. Wisk takes a leaf from the Boeing V-22 Osprey, which cost $30 billion in pre-Biden money to develop, and tilts the motors as necessary.

If Wisk can achieve its engineering, certification, and production goals, the customers won’t have to worry about hiring pilots: the tilt-rotor is fully autonomous.

Related:

Full post, including comments

Hunting down an air handler fan that is running too slowly and growing mildew (breaker panel power monitor)

by philg

I’ve been trying to reengineer the air conditioning in our house to match the new(ish) reduced cooling load after a hurricane low-E glass window retrofit by the previous owner (see ChatGPT is almost as bad at home maintenance as I am). Before I downsized the system, however, I decided that I had better make sure that the theoretical Manual J calculations of an 8.5-ton demand were correct. The goal was to see what percentage of the time the 12-ton current system (divided into three condensers/air-handlers) was running on hot days (e.g., when the NYT says South Florida is facing EXTREME DANGER).

I decided to install an inductive current monitor in the circuit breaker panel that could watch all three air handler breakers, specifically the Emporia Vue 2. This is supposed to be easy to install oneself and I have a Ph.D. in Electrical Engineering… so I decided to hire an electrician to do it properly. It took him less than one hour and he never shut off power to the panel, as the instructions suggest.

The software is reasonably good, but setup operations take longer to complete than you’d expect. Budget perhaps 30 minutes to get it all connected to WiFi and then to rename the ports. Here’s our Air Handler 3, a 3-ton system, on a day that was 125 degrees (NYT) or 91 degrees (Google/Apple). We can see that the variable-speed air handler (sadly, connected to a one-speed condenser) ramps up to about 500 watts and also that it is running most of the time (the calculated current demand for the upstairs was just 2.3 tons).

Here’s the a 5-ton air handler:

It’s drawing only 100 watts. Notice that I called it “AH2Try2” because I replaced the probe (myself!) and connected it to a different port because I assumed that the Emporia device was bad.

The installation guide for the Trane TEM6 air handler says that it should be drawing at least 500 watts:

I found that the unit was sweating on the outside and, opening it up, mildewing on the inside. The A/C contractor did the following:

  • replaced the blower (covered under warranty by Trane)
  • took the air handler apart and cleaned it
  • pumped out the refrigerant and cut the evaporator coil out and brought it down to the side yard and cleaned it thoroughly
  • cleaned out the air handler interior
  • replaced the plenum
  • replaced a failed UV sterilizer that had been in the old plenum with a REME HALO

With the new fan in place, power consumption went up to over 700 watts and the cabinet stopped sweating.

Given that air handlers are hard-wired, I don’t know of any other way to verify that they’re working properly. The regular A/C service guys don’t measure airflow carefully. And the power monitor is fun to have for investigating random appliance power consumption questions. Our 20-year-old last-legs KitchenAid refrigerator is consuming only 75 watts, for example.

What if you don’t want to spend $250-ish, including an electrician’s time? You can spend $thousands to replace your whole breaker panel and/or all of the breakers with “WiFi breakers”. Span will sell you a panel for $4,500 (plus the breakers?). Can you guess where this new company is located?

Eaton, which has been making panels for about 100 years, sells individual WiFi breakers that can report consumption and also be reset remotely. These seem to cost about $250 each, but if you already have an Eaton panel the installation could be cheap and simple.

Leviton makes a comprehensive system, but it will require replacing your panel(s). The panels themselves from Leviton seem to be cheap (less than $200). Once that’s done, an individual breaker can be as cheap as $54. Our electrician put this system in his own house and likes it.

Speaking of breakers, how long do they last in your experience? Our panels have spent 20 years in a hot/humid garage and the Cutler Hammer breakers inside don’t seem to be happy about it. Especially if a big one trips or is toggled it will tend to require replacement.

Full post, including comments

Biking without bike infrastructure: the Netherlands

by philg

In Danish happiness: bicycle infrastructure I described the Danish system of road/curb/bike path/curb/sidewalk. What if a significant percentage of a society used bicycles for transportation, but nobody bothered to build infrastructure? That’s the Netherlands!

I recently visited a friend in Delft, a university town south of Amsterdam. There are no generally no curbs at all in the downtown area. The road is informally divided into car/bike/pedestrian, but these divisions can change depending on what exactly is sticking out from a house, possibly forcing pedestrians into the bike area, or whether a truck is trying to use the road.

The risk of injury has ballooned in the last few years due to the popularity of cargo bikes and electric bikes. Instead of getting hit by a 200 lb. person-bike combo going 8 mph you’ll get hit by a 400 lb. person-small person-groceries-bike combo going 15 mph. “Trouble in cyclists’ paradise: Amsterdam accused of favouring pedestrians” (Guardian 2021) describes the increasing conflict between walkers and bikers in Amsterdam.

There aren’t as many collisions as you’d imagine, but pedestrians are required to be constantly mindful. This works for the Dutch, but tourists are frequently wandering casually into near-collisions with cyclists. What the cyclists have gained is balanced by a loss of mental peace and capacity among pedestrians.

Here’s a narrow street designed for pedestrians in The Hague:

The bicycle is being used for transportation, not recreation, so it might be whipping by these pedestrians at 10-20 mph. Here are the two transportation modes interacting in Delft:

Maybe those white boxes are supposed to delineate between walking and biking? Or maybe there are two lanes for opposite directions? I didn’t figure it out.

Just a few of the bikes parked near the Amsterdam Zuid secondary station:

What if you choose “neither” but don’t have a car and/or don’t want to pay what my rich local friend said were insanely expensive parking fees? Take the tram!

My take-away: the Danes did it right.

Full post, including comments

Test deep submersible designs for 2-3X the number of dives that will carry humans?

by philg

Department of Fighting the Last War… let’s talk about ideas that could have prevented the Titan tragedy.

The potential for failure of a pressure vessel is something that aviation has been dealing with since at least the 1930s (Boeing 307). The cycles of pressurization and depressurization are known to cause metal fatigue and, sometimes, lead to catastrophic failure. Certification authorities, such as the FAA, require structural analysis to certify a cycles or hours limit and this may be extended once there is more experience with the airframe. Running the cycles up to 90,000 (many short hops) contributed to the failure of Aloha Airlines Flight 243, a Boeing 737.

The reasonable standard of safety for deep-sea exploration and tourism is lower than for commercial airline travel. However, what about a rule that you have to build at least two of each design and use one as a sanity check on the pressure vessel design and ability to tolerate cycling fatigue? Don’t send humans on a 50th dive in a machine unless its sister ship has done at least 100 or 150 robot-only dives to the same depth. (If additional protection from fatigue-related failures is desired, increase the number of sister ships and the multiplier for how many dives they must survive compared to the human-occupied dives.)

Note that this procedure wouldn’t guarantee safety. Portions of the de Havilland Comet were subjected to 16,000 simulated cycles and the finished design nonetheless suffered a catastrophic failure with passengers on board (though the airplane that failed had some construction method differences from the prototype that was tested). But it is better than finding out the cycle limit with humans on board. And every machine that goes to the ocean floor will have a cycle limit.

Full post, including comments